July 2026 Fed Decision: Why a Rate Hike Is Back on the Table as Big Tech Earnings and Nvidia’s $750 Billion Deals Collide

Last updated: July 28, 2026, 11:30 a.m. CEST (5:30 a.m. ET)

The Federal Open Market Committee announces its interest rate decision on Wednesday, July 29, at 2 p.m. Eastern time, and for the first time in more than two years the direction of the surprise runs upward rather than downward. Federal funds futures tracked by the CME FedWatch tool put the odds of no change at roughly 66% as of Monday, July 27, which leaves something close to a one-in-three chance that Chair Kevin Warsh’s Fed raises rates for the first time since 2023. Economists surveyed by FactSet and polled by CBS News overwhelmingly expect a hold at the current 3.50%–3.75% target range. The market is not so sure.

That uncertainty is the story. It is also colliding with the single densest week of the corporate calendar: Microsoft and Meta Platforms report on Wednesday, Apple and Amazon on Thursday, and 177 S&P 500 companies file second-quarter results across five sessions. The June personal consumption expenditures price index — the Fed’s preferred inflation gauge — arrives on Thursday morning, one day after the decision. And running underneath all of it is a financing story that has become impossible to separate from the macro one: Nvidia spent the past four days attaching its balance sheet to something in the region of three quarters of a trillion dollars of artificial-intelligence infrastructure, and the credit market noticed.

The short answer to the question most readers are asking: the Fed will most likely hold on July 29, but this is a genuinely live meeting, the statement and press conference matter more than the decision itself, and the market’s real anxiety is concentrated at the long end of the Treasury curve and in the debt now funding the AI build-out. What follows is the evidence behind that assessment, the numbers that support it, and the credible arguments on the other side.

Key Takeaways

• Main development: The FOMC meets July 28–29, 2026, with the federal funds target range at 3.50%–3.75% and market pricing implying roughly a one-in-three probability of a quarter-point increase — an unusual level of doubt heading into a decision day.
• Key figures: June consumer prices fell 0.4% on the month and rose 3.5% from a year earlier, the largest monthly decline since April 2020; core CPI was flat on the month at 2.6% year over year. But producer prices for final demand were still up 5.5% over twelve months, and the May PCE price index — the Fed’s target measure — ran at 4.1% headline and 3.4% core.
• Market response: The S&P 500 fell 0.6% in the week ended Friday, July 24, its second consecutive weekly decline and the longest losing streak since late March, leaving the index 2.6% below its June 2 record close and up 8.3% for the year. Nvidia closed down 4.99% at $196.51 on Monday, July 27.
• Why it matters: A hiking Fed changes the discount rate applied to the most expensive part of the equity market at exactly the moment that part of the market is issuing record amounts of debt to fund capital spending. The 30-year Treasury yield has spent its longest stretch above 5% since 2007.
• What comes next: Decision and Warsh press conference Wednesday, July 29; Microsoft and Meta results the same afternoon; Apple and Amazon Thursday, July 30; June personal income and outlays, including the PCE price index, Thursday, July 30, at 8:30 a.m. ET.

What the Fed is actually deciding on July 29

Start with where policy stands. The FOMC has held the federal funds target range at 3.50%–3.75% since its March meeting. The June 17 decision — Kevin Warsh’s first as chair — kept it there, and reporting on that meeting described the vote as unanimous. If the committee holds again on Wednesday it will be the fifth straight meeting without a change.

What makes this one different is that the distribution of plausible outcomes has become lopsided in a way it has not been since 2022. Through the first months of 2026, futures markets were priced for one to two cuts. By mid-June, after the war between the United States and Iran pushed crude oil to levels not seen since 2022, that had inverted. The U.S. Bank Asset Management Group’s account of the June meeting noted that the Summary of Economic Projections released that day carried a median expectation of one to two hikes in 2026, with nine of eighteen participants penciling in at least one increase.

Warsh himself did not submit a dot. That was deliberate, and it is the single most important thing to understand about how this Fed communicates.

There is a second, quieter reason the July meeting is awkward. The June PCE report lands on Thursday, July 30 — the morning after the decision. Given how sharply energy prices reversed during June, that report is likely to show a meaningful step down in headline inflation. A committee that raised rates on Wednesday and then watched its preferred inflation gauge decelerate on Thursday would look, at minimum, poorly sequenced. A committee that held and then saw a soft print would look vindicated. The calendar itself argues for patience, and the FOMC is well aware of it.

Between now and the September meeting the committee will receive two CPI reports, two PPI reports, two PCE reports and two employment reports. That is an unusually thick data run for a six-week gap, and it gives any member inclined to wait a defensible reason to do so.

Three inflation gauges, three different stories

The reason this decision is hard is that the United States currently has three respectable measures of inflation pointing in materially different directions.

Consumer prices, released July 14, looked like relief. The Bureau of Labor Statistics reported that the all-items CPI fell 0.4% on a seasonally adjusted basis in June and was up 3.5% before seasonal adjustment over the previous twelve months. The monthly decline was the largest since a 0.8% drop in April 2020, at the onset of the pandemic shutdowns. Economists polled by LSEG had looked for a 0.1% monthly decline and a 3.8% annual rate. Core CPI, which strips out food and energy, was unchanged on the month and up 2.6% from a year earlier, against forecasts of 0.2% and 2.8%. In May, the same series had printed 0.5% monthly and 4.2% annual.

That is a genuine downside surprise, and it is worth being precise about what caused it. Energy did the heavy lifting. Shelter, the largest single component of the index, rose just 0.1% on the month after a string of firmer readings. Core services were flat on the month while still running 3.2% higher than a year earlier.

Producer prices tell a rougher story. The BLS reported that the index for final demand fell 0.3% in June — but was up 5.5% over the twelve months ended in June. Final demand goods dropped 1.4% on the month, the largest decrease since a 1.9% decline in July 2022, with final demand energy down 6.4%. Final demand services still rose 0.2%. A 5.5% annual rate at the producer level, alongside a 3.5% consumer rate, describes an economy where cost pressure has been absorbed into margins and pipelines rather than fully passed through — or where the pass-through is still ahead.

Then there is the measure the Fed actually targets. In the May personal income and outlays report published June 25, the Bureau of Economic Analysis put the PCE price index up 0.4% on the month and 4.1% from a year earlier. Excluding food and energy, it rose 0.3% on the month and 3.4% over twelve months. Core PCE had climbed from 3.0% in December 2025 to 3.3% by April. Whatever the June figure shows on Thursday, the starting point is a core rate roughly 140 basis points — 1.4 percentage points — above the 2% objective.

Three points of interpretation matter here, and they are frequently muddled in coverage.

First, a slowdown in inflation is not falling prices. June’s headline CPI declined on the month, but the index still sits 3.5% above where it stood a year ago, and the price level itself has not reversed. Households do not experience disinflation as relief; they experience it as prices rising more slowly from an already-higher base.

Second, CPI and PCE are not interchangeable. They use different weights, different treatments of medical costs and different substitution assumptions. PCE typically runs below CPI. That the gap has inverted in some recent months — with PCE at 4.1% against CPI at 4.2% in May and a much wider spread on the core measures — reflects how concentrated the shock has been in energy and energy-adjacent categories.

Third, and most consequential for Wednesday: the Fed’s mandate is written against PCE, and the June PCE report has not been published yet. The committee will be voting with a partial information set that its own preferred gauge will complete twenty-four hours later.

The Warsh Fed: less guidance, more institutional review

Kevin Warsh took over a central bank that had spent fifteen years teaching markets to read its forward guidance, and he has spent his first two months unteaching them.

The June statement was dramatically shortened and stripped of language signaling a bias toward future cuts. Warsh declined to contribute a projection to the dot plot — an unusual step for a sitting chair, and one that removes the single most closely watched line in the SEP. In his first press conference he described the committee as “unanimous and unambiguous” in its commitment to fighting inflation and, by U.S. Bank’s count, used the phrase “price stability” twelve times. Treasury yields rose during that press conference: the two-year climbed 16 basis points and the ten-year six basis points on the day, while the S&P 500 fell 1.2% and the Russell 2000 fell 0.8%.

He also announced five internal task forces, covering Fed communications, balance sheet policy, data sources, productivity and employment, and the inflation framework itself. Taken together, that is the agenda of someone who intends to change how the institution works rather than simply steer it.

For investors, the practical consequence is uncomfortable: the Fed has reduced the amount of information it gives markets at precisely the moment markets most want information. Pricing a one-in-three hike probability into a decision is, in part, a function of that. When a central bank refuses to pre-commit, the distribution of outcomes widens mechanically.

It is worth being careful about what Warsh has and has not said. He has repeatedly described inflation as too high and told Congress that FOMC members have no tolerance for persistently elevated inflation. He has not promised a rate increase at any specific meeting, and no reading of his public remarks supports treating a July hike as signaled. Analysts who describe the next move as “most likely up” are making a probabilistic judgment about the reaction function, not reporting a commitment.

The long end is the real constraint

If the policy rate were the only thing that mattered, the past six weeks would have been quiet. The pressure has been further out the curve.

The 30-year Treasury yield has traded above 5% on 27 days so far in 2026, including a run of twelve consecutive sessions through July 22 — the longest such streak since 2007, when the yield held above that level for 50 days. It reached an intraday peak of 5.19% on May 19, its highest since the summer of 2007, and was sitting near 5.14% as of July 22. The ten-year has pushed to roughly 4.71% at points this month, its highest since January 2025.

Two forces are doing the work, and they are only partly about the Fed.

The first is inflation compensation. Investors lending for thirty years into an economy where core PCE has been above target for an extended stretch, and where an energy shock has demonstrated how quickly headline inflation can re-accelerate, want more yield for the privilege.

The second is supply. Treasury issuance and the fiscal outlook sit behind a good deal of the term premium rebuild, and market participants have noted that some of the recent move in short-dated yields reflects budget-cycle anxiety as much as policy expectations. That distinction matters. A two-year yield rising because traders expect a hike is a monetary signal. A two-year yield rising because dealers are bracing for heavy bill and coupon supply is a fiscal one, and the Fed cannot fix it by holding rates.

A third force has been added this year, and it is new: corporate issuance from the largest technology companies. The hyperscalers have moved from being net cash generators with fortress balance sheets to being among the largest investment-grade issuers in the market, funding data-center construction that does not pay for itself within a normal capital-budgeting horizon. That supply competes with Treasuries for the same pool of duration-tolerant buyers.

The practical effect is that long-term borrowing costs across the economy — mortgages, corporate refinancing, commercial real estate, project finance for power generation — have stayed elevated even though the Fed has not moved in four months. Officials have limited tools against this. They can talk about it. Jawboning a long end that is worried about the persistence of inflation and the size of the deficit is unlikely to accomplish much.

How a rate move would actually transmit

A quarter-point change in the federal funds target is a small adjustment to the rate at which banks lend reserves to each other overnight. It matters because of what it implies, and because of what it drags along with it. It is worth separating the mechanical effects from the expectational ones, because coverage routinely conflates them.

Deposit and money-market rates move quickly. Money market funds, high-yield savings accounts and short-dated Treasury bills reprice within days of an FOMC move, because their yields track the overnight rate closely. Savers see the effect first. Whether banks pass a full quarter point through to retail deposits depends on competition for funding, and historically the pass-through on the way up has been slower and less complete than on the way down.

Credit card and variable-rate consumer debt reprices on the prime rate. The prime rate typically moves in lockstep with the fed funds target, so variable-rate card balances, home equity lines and many small-business credit facilities reset within one or two billing cycles. For a household carrying a revolving balance, a quarter point is a modest but immediate increase in monthly cost.

Mortgages do not follow the Fed. This is the most persistent misconception in personal finance coverage. Thirty-year fixed mortgage rates track long-dated Treasury yields and mortgage-backed security spreads, not the overnight policy rate. That is precisely why the 30-year Treasury’s extended stay above 5% matters more to housing affordability than anything the FOMC does on Wednesday. It is entirely possible for the Fed to hike and for mortgage rates to fall, if the market reads the hike as evidence that inflation will be contained. It is equally possible for the Fed to hold and for mortgage rates to rise on supply concerns.

Corporate borrowing costs depend on which part of the curve a company uses. Firms funding with floating-rate bank debt or leveraged loans feel a policy move immediately. Firms issuing ten- or thirty-year bonds — which describes the technology companies currently funding data centers — care far more about the long end and about credit spreads. A company terming out debt at thirty years is exposed to term premium and inflation expectations, not to the overnight rate.

The dollar and equity valuations respond to expectations, not the move itself. By the time a decision is announced, the market has priced most of what it expects. What moves the dollar and long-duration equities is the gap between the decision and the expectation, plus whatever the statement and press conference imply about the path from here. That is why a hold accompanied by hawkish language can tighten financial conditions more than a hike accompanied by a signal that it is the last one.

This is also the reason the Warsh Fed’s reduced guidance is more than a stylistic preference. In a framework where guidance does the work, a central bank can tighten conditions without moving rates. Removing guidance means the policy rate has to do more of the lifting — or means conditions stay looser than the committee would like, which is precisely the tension several participants have flagged.

For readers assessing their own exposure: the relevant question is not “will the Fed hike” but “which rate am I actually exposed to.” Someone with a fixed mortgage and cash savings is positioned very differently from someone with a home equity line and a portfolio concentrated in long-duration growth equities. None of this constitutes advice about any individual’s circumstances, and rates quoted here reflect market levels as of late July 2026.

The rest of the world is already tightening

One piece of context routinely missing from American coverage: the Federal Reserve would not be moving first.

Global central banks spent 2025 easing. That has reversed. The European Central Bank and the Bank of Japan both raised rates in June 2026 in response to energy-driven price increases, and market expectations point to increases from the Bank of England and the Bank of Canada at some point this year. On a global net basis, the count of hikes minus cuts has swung back toward tightening for the first time since the 2022–2023 cycle.

That matters for two reasons. The first is currency. If the Fed holds while other major central banks tighten, the interest rate differential narrows and the dollar tends to weaken, which raises the domestic price of imported goods — including energy — and works against the disinflation the committee is hoping to see. The second is signaling. A Fed that hikes into a global tightening cycle is doing something ordinary. A Fed that hikes alone would be making a much larger statement about how it reads the U.S. inflation problem.

The energy shock is not an American phenomenon. It is a global one, transmitted through the same chokepoint, and it is producing broadly similar policy responses in economies with very different labor markets and fiscal positions. That commonality is itself evidence that the current inflation impulse is primarily a supply shock rather than a demand one — which is the strongest analytical argument available to those on the committee who would prefer to look through it.

The counterargument, and it is a serious one, is that central banks have already spent this decade’s credibility looking through a supply shock. The 2021–2022 experience taught policymakers that “transitory” is a claim about duration that cannot be verified in advance, and that the cost of being wrong is measured in years of above-target inflation and a much larger eventual tightening. Warsh has been notably direct about the limits of that patience. Reuters commentary published on July 28 framed the problem crisply: central banks cannot credibly see through this many overlapping inflationary risks at once — energy, tariffs, shipping, and a capital-spending boom running at full capacity.

Oil, Iran and the variable the Fed cannot control

The single largest input into the 2026 inflation problem is not tariffs, wages or housing. It is the price of crude, and the price of crude has been set for months by military decisions in the Persian Gulf.

The arithmetic is stark. West Texas Intermediate front-month futures began the year near $57 a barrel, peaked around $113 in April, and had fallen back to roughly $76 by mid-June. They then climbed again through July as the conflict escalated. Brent was trading around $100 in the week of July 20. On Monday, July 27, after a third consecutive night without American strikes, CNBC reported Brent falling below $90 a barrel. Reported closing marks for that session varied noticeably across data providers, which is itself a symptom of how thin and headline-driven the market has become; readers comparing figures across outlets should check whether a given number is an intraday level, a front-month settlement or a rolled contract price.

What triggered the move was a pause, not a resolution. The United States had conducted strikes on Iranian targets on thirteen consecutive nights before halting. Iran indicated it would refrain from attacks so long as the United States did the same. NBC News reported that the suspension followed warnings from the president’s advisers that the military was running short of viable targets. No public statement from the White House explained the decision at the time.

Two things about that pause deserve emphasis for anyone modeling the inflation path.

First, the ceasefire is informal and reciprocal, which makes it fragile by construction. Nothing has been signed. Israeli Prime Minister Benjamin Netanyahu landed in Washington on Monday, July 27 and was scheduled to meet President Trump at the White House on Tuesday, July 28, with Iran the stated focus. He also traveled for the funeral of Senator Lindsey Graham, who died last week. Netanyahu has been publicly clear that in his view the campaign should not end while Iran retains nuclear capability. Whether the pause survives that meeting is a genuinely open question, and it is a question with a direct line to the June and July CPI prints.

Second, and less discussed, the shipping disruption has proved stickier than the headline oil price. Vessel traffic through the Strait of Hormuz has collapsed. Lloyd’s List Intelligence recorded 53 transits in the week through July 20, down 66% from 157 the week before; tankers and gas carriers specifically fell to 30 crossings from 90. Before the war, an estimated 120 to 140 vessels crossed daily, roughly half of them oil tankers moving in the region of 20 million barrels a day. At least nine ships have been attacked since July 6 as Iran has pressed vessels to route through its territorial waters.

That lag is the crux. Insurance rates, charter availability and crew willingness do not normalize the day the bombing stops. Even in a scenario where diplomacy succeeds, pipelines and refineries damaged during the campaign take time to bring back, and rebuilt infrastructure takes longer still. A Fed forecasting energy prices for the back half of 2026 is forecasting a shipping-risk premium as much as a barrel count.

There is a conspiracy theory circulating in trading desks and on social platforms that the strike pause was timed to soften oil prices ahead of the FOMC meeting, giving the committee room not to hike. It should be treated as what it is: an unsupported claim with no identifiable source, and one that would require an implausibly precise chain of assumptions about market reactions. Bloomberg’s Michael McKee, asked about it on air on July 27, said he did not believe there was truth to it. The more prosaic explanation — that the campaign had reached the limits of its target set and its munitions stockpile — is better supported by the reporting.

The busiest week of earnings season: $24 trillion reports

Set the Fed aside for a moment. This week would be significant on the corporate calendar alone.

Microsoft reports fiscal fourth-quarter results on Wednesday, July 29, alongside Meta Platforms’ second quarter. Apple’s fiscal third quarter and Amazon’s second quarter follow on Thursday, July 30. Those four companies together account for a substantial share of the S&P 500’s market capitalization — Bloomberg Television put the figure at roughly 17% on July 27 — and FactSet counted 177 index members scheduled to report during the week.

The framing that has taken hold on trading desks is that capital expenditure guidance, not earnings, is the number that moves these stocks. That framing is correct, and Alphabet demonstrated it a week early.

The Alphabet distortion: why the “record” earnings season is not what it looks like

Here is a statistic that will be repeated all week and that almost every reader will misunderstand.

As of July 24, with 27% of the S&P 500 reported, FactSet put the blended second-quarter earnings growth rate at 37.9%. If that held, it would be the index’s fastest growth since the third quarter of 2021. Companies were beating estimates by 39.3% in aggregate, against a five-year average beat of 7.0%.

Both numbers are almost entirely an artifact of one company.

Alphabet reported GAAP diluted earnings per share of $9.11 against a consensus of $2.88. That actual figure included a gain of $98 billion. Strip Alphabet out and the aggregate earnings surprise falls from 39.3% to 12.6%, and the blended growth rate falls from 37.9% to 25.9%. FactSet’s John Butters flagged this explicitly, which is more than can be said for a great deal of the coverage that has recycled the 37.9% figure without qualification.

This is exactly the distinction that separates useful earnings analysis from headline-chasing. A $98 billion gain recognized in a single quarter is a real accounting event with real balance-sheet consequences, but it is not operating performance, it does not recur, and treating it as evidence that corporate America is compounding earnings at nearly 40% is a category error.

The underlying picture, once cleaned up, is still strong. Excluding Alphabet, roughly 26% blended earnings growth would mark the second consecutive quarter above 20% and the seventh consecutive quarter of double-digit growth. Revenue growth of 13.2% would be the fastest since the second quarter of 2022, with all eleven sectors growing. Eighty-six percent of reporters have beaten on earnings, against a five-year average of 78%. Eighty percent have beaten on revenue. These are good results by any historical standard.

But the composition matters. Ten of eleven sectors are growing; health care is the exception. Energy, communication services, information technology and materials lead. Energy’s contribution is a direct function of the same oil price that is complicating the Fed’s job — a reminder that “strong earnings” and “benign macro” are not the same thing, and can in fact be opposites.

Valuation, meanwhile, has not run away. The forward twelve-month price-to-earnings ratio stood at 20.1 based on the close of Wednesday, July 22 — above the five-year average of 19.9 and the ten-year average of 19.0, but below the 20.4 recorded at the end of the second quarter. Citi Wealth’s Kate Moore made a related argument on Bloomberg Television on July 27, noting that 2026’s equity gains have been driven by earnings rather than multiple expansion, which she treats as a reason to stay fully allocated to equities with an overweight to U.S. large caps. That is a defensible read of the data. It is also a house view from a firm with an interest in clients remaining invested, and readers should weigh it accordingly.

Alphabet’s capex raise was the template — and the warning

Alphabet’s July 22 report set the pattern the rest of the week will be measured against.

Revenue came in at $119.8 billion, up 24% year over year and the twelfth consecutive quarter of double-digit growth. Google Cloud grew 82%. On any conventional reading, an outstanding quarter.

The stock fell anyway. Alphabet raised full-year 2026 capital expenditure guidance to $195–$205 billion from a prior range of $180–$190 billion — an increase of roughly $15 billion at the midpoint — with approximately 60% earmarked for servers and 40% for data centers and networking equipment. Shares fell about 3.65% in after-hours trading to $329.43 and declined 7.1% across the week.

Management’s explanation was straightforward and, on the evidence, credible: the spending increase reflects demand it cannot currently serve. Cloud backlog remains enormous relative to deliverable capacity. If a company genuinely cannot fulfill signed commitments, raising capex is the correct decision and refusing to would be the error.

The market’s response reflects a different calculation. Every incremental dollar of capital expenditure reduces free cash flow now in exchange for revenue later, and the depreciation schedule that follows compresses reported margins for years. Investors who were happy to fund the AI build-out when it was measured in tens of billions are running a different discount rate on it now that the four largest hyperscalers are collectively committing hundreds of billions annually and, increasingly, borrowing to do it.

Kelly Covley of Manning & Napier framed the two-sided risk on Bloomberg Television on July 27 in terms worth repeating. The obvious risk is that hyperscalers guide capex higher and the market punishes them, as it did Alphabet. The less obvious and potentially larger risk is the reverse: that capex guidance comes in conservative, and the enormous cohort of companies that have been benefiting from that spending — memory and storage manufacturers, chipmakers, utilities, industrials, data-center real estate — reprices downward at once. Commoditized suppliers face a double hit in that scenario, because falling demand triggers price competition on top of volume declines.

That asymmetry explains one of the stranger market patterns of recent weeks: on days when the hyperscalers rally, the chipmakers often fall, and vice versa. It is the market attempting to price a transfer of value between the entities writing the checks and the entities cashing them, without any confident view on the total size of the pot.

Reading the four reports: what actually matters in each

Aggregating Microsoft, Meta, Apple and Amazon into “Big Tech earnings” obscures four very different businesses facing four different questions. Here is what to look for in each, and why the headline EPS number is unlikely to be the thing that moves the stock.

Microsoft (fiscal fourth quarter, July 29)

Consensus estimates compiled by data providers cluster around $4.22 to $4.24 in earnings per share on revenue of roughly $87.5 billion to $87.7 billion. Those figures are analyst forecasts, not company guidance, and they vary by source.

The number that matters is Azure. Specifically: growth rate, whether management again describes itself as capacity constrained, and what fiscal 2027 capital expenditure commentary implies. Microsoft has spent several quarters telling investors it cannot serve all the demand it has, which is simultaneously the best possible statement about demand and the reason capex keeps rising. A quarter in which Azure growth decelerates and capex rises would be the worst combination — it would suggest spending is running ahead of monetization rather than chasing it.

The second thing to watch is the split between capital expenditure funded from operating cash flow and capital expenditure funded from debt or finance leases. Microsoft generates enormous cash. The proportion it is choosing not to fund internally is a signal about how large the build has become relative to even its resources.

Meta Platforms (second quarter, July 29)

Consensus sits near $7.18 to $7.24 per share. Meta is the cleanest test of the market’s tolerance for capex because it has the least obvious near-term revenue attached to its AI spending. Microsoft and Amazon sell cloud capacity; the spending has a customer. Meta’s AI infrastructure primarily serves its own products.

Full-year 2026 capital expenditure guidance of $125 billion to $145 billion is already an enormous range, and the width of it is itself informative — it suggests management does not know where in that band it will land. If the range moves up, investors will ask what changed. If Reality Labs losses widen alongside it, the operating margin question becomes acute.

The bull case for Meta has always rested on advertising revenue growth funding everything else. The specific thing to check is whether ad revenue growth is still comfortably outpacing total expense growth. If it is not, the capex story becomes a margin story.

Apple (fiscal third quarter, July 30)

Consensus is around $1.86 per share. Apple is the odd one out in this group: it is not building hyperscale AI infrastructure and it is not a meaningful buyer of Nvidia systems. Its exposure to the AI cycle runs the other way — as a buyer of components whose prices are being driven up by everyone else’s spending.

Conventional DRAM prices rising in the region of 93% to 98% quarter over quarter in the first quarter is a direct hit to Apple’s bill of materials across every product line. The gross margin line and management’s commentary on component costs are therefore more informative than usual. So is anything said about supplier qualification — Apple has signaled it wants to diversify beyond the three incumbent memory manufacturers, and CXMT’s listing has made that a live commercial question rather than a theoretical one.

Tariffs and the September product cycle will also feature. Neither is likely to be the swing factor this quarter.

Amazon (second quarter, July 30)

Consensus is around $1.85 per share. Amazon carries the most complex version of the same question the others face, because AWS capital spending, retail fulfillment investment and free cash flow interact in ways that make the headline numbers hard to read.

Watch three lines. First, AWS revenue growth against backlog — the gap between contracted commitments and delivered revenue is the clearest read on whether capacity is the constraint. Second, trailing twelve-month free cash flow, which Amazon reports prominently and which has swung dramatically as capex has risen. Third, the operating income split between AWS and the retail segments, because the market has been valuing Amazon increasingly as an infrastructure company with a retail attachment rather than the reverse.

Amazon’s absence from the open-weights letter is a small but genuine question mark, given its cloud position and its investment in Anthropic. Whether it comes up on the call is another matter.

The common thread

All four are being asked the same underlying question, which Alphabet answered a week early and got punished for: how much are you spending, and when does it produce cash? The companies have a coherent answer — demand exceeds capacity, and underbuilding is the greater risk. The market’s discomfort is not that the answer is wrong. It is that the answer has been the same for eight consecutive quarters while the number attached to it has roughly tripled, and the depreciation from the earlier tranches is now arriving in the income statement.

Nvidia’s $750 billion week: separating what is signed from what is reported

Over four days spanning July 24 to July 27, Nvidia was attached to headline numbers totaling somewhere north of three quarters of a trillion dollars. Almost every subsequent write-up has aggregated those figures into a single total. That aggregation obscures more than it reveals, because the individual components have wildly different levels of legal and financial certainty.

Here is the breakdown, sorted by how much is actually confirmed.

Confirmed by primary source: SK Group, $500 billion-plus

On July 24, Nvidia and South Korea’s SK Group jointly announced a “$500-billion-plus comprehensive partnership”. The two components: SK Telecom will build an AI factory of up to 2 gigawatts using Nvidia’s DSX platform and Vera Rubin accelerated computing powered by SK hynix HBM4, with the first facility planned to come online in 2027; and SK hynix enters a long-term partnership with Nvidia to secure and co-develop next-generation AI memory.

The detail that matters, and which was largely absent from secondary coverage: the two sides signed letters of intent, not definitive agreements. A letter of intent is a statement of commercial intention. It is not a purchase order, and the $500 billion figure is a characterization of a multi-year, multi-party ambition rather than a contracted sum flowing in any single direction. SK Group Chairman Chey Tae-won framed it as an effort to make Korea “a global hub that drives AI innovation”; Jensen Huang described building “a new generation of AI factories.”

Anyone modeling Nvidia revenue off that headline is modeling a press release.

Confirmed by primary source: Safe Superintelligence

On July 27, Nvidia and Safe Superintelligence Inc. — the research company founded by former OpenAI chief scientist Ilya Sutskever — announced a long-term strategic partnership including an Nvidia investment in SSI. Bloomberg reported the investment at $5 billion. The arrangement gives SSI access to the Vera Rubin platform and, per the companies, will let SSI increase its compute by an order of magnitude, with collaboration on future Nvidia platforms.

Nvidia framed the deal partly as a research-access arrangement, saying it had obtained rare visibility into SSI’s closely held work. For a company that has raised at enormous valuations while publishing almost nothing, that validation cuts both ways.

Reported, not confirmed: the Ohio backstop and the chip financing

This is where the largest numbers live, and where the least is settled.

The Wall Street Journal and Bloomberg reported on July 27 that Nvidia is in discussions to guarantee as much as $250 billion of lease payments so that OpenAI can occupy a 10-gigawatt data center being built in southern Ohio by SB Energy, the power arm of SoftBank. Separately, Nvidia is reported to be considering financing up to $350 billion of OpenAI’s purchases of Nvidia chips tied to the same project. Total project cost including silicon has been reported above $500 billion, which would make it the largest data center project ever proposed.

The site is a decommissioned uranium enrichment facility on federal land roughly 50 miles south of Columbus. Power is expected to come from a new $33 billion natural gas plant pledged by Japan. Neither Nvidia nor OpenAI commented on the reports. People familiar with the talks have cautioned that terms are unsettled and the arrangement could collapse.

The commercial logic is not mysterious. OpenAI is not profitable and does not carry an investment-grade credit rating. A developer trying to raise construction debt against OpenAI’s covenant alone would pay punitive spreads or fail to raise at all. With Nvidia standing behind the lease obligations, the borrower is effectively financing against Nvidia’s balance sheet — a company with a market capitalization near $5 trillion and, per its most recent quarterly results, a profit margin around 75%.

That last line is the one to sit with. Nvidia’s own disclosed lease-guarantee ceiling is $3.5 billion. A $250 billion commitment would be roughly seventy times that. Whatever the merits of the transaction, it would represent a step-change in the kind of financial risk Nvidia carries — from a company that sells hardware into a company that underwrites the ability of its customers to pay for hardware.

What “circular financing” actually means, and the honest case on both sides

The phrase gets thrown around loosely. It is worth defining before arguing about it.

Circular financing, in this context, describes an arrangement where a supplier provides capital — equity, debt, or a credit guarantee — to a customer, and the customer uses that capital, directly or indirectly, to buy the supplier’s products. The supplier books revenue. The customer books an asset. Both balance sheets grow. Nothing is necessarily improper about this; vendor financing is a legitimate and ancient practice, common in aviation, telecommunications equipment and capital goods generally. The question is always one of degree, disclosure and what happens under stress.

The case that this is fine

Nvidia’s own argument, articulated repeatedly by executives and echoed by analysts, has three parts.

First, there is no quid pro quo. Nvidia does not contractually require recipients of its capital to spend it on Nvidia silicon. OpenAI is free to buy from AMD, to deploy its own custom accelerators — which it is developing — or to use non-Nvidia cloud capacity. Support for the ecosystem is not the same as a tied sale.

Second, demand genuinely exceeds supply. This is the strongest part of the argument, because it is externally verifiable. Alphabet’s cloud backlog and its capex raise, Microsoft’s repeated commentary about capacity constraints, and the extraordinary run in memory pricing all point to a market where the binding constraint is the ability to build, not the willingness to buy. Nvidia’s position is that it is greasing the wheels of construction, not manufacturing demand.

Third, the alternative is worse for everyone. If financing constraints slow the build-out, the bottleneck becomes capital rather than physics, and a six-month digestion period would ripple through the entire supply chain. Bloomberg Intelligence’s Mandeep Singh made this point on July 27: a pause, even a short one, would be materially damaging for Nvidia, and the company is spending to prevent one.

The case that this is a problem

The skeptical argument is not that any single deal is unsound. It is about what the structure does to information.

When a supplier underwrites its customer’s ability to buy, the market loses the ability to distinguish independent demand from supported demand. Revenue that would previously have been evidence of end-market appetite becomes, in part, evidence of the supplier’s own risk appetite. That is a real degradation in signal quality, and it arrives at a moment when investors are trying to judge whether AI capital spending is justified by future cash flows.

The credit market registered this immediately. Prices for credit default swaps on Nvidia bonds — instruments that function as insurance against default — recorded their largest intraday increase since they began trading actively in November, according to ICE Data Services figures cited by Bloomberg. Nvidia’s equity fell 4.99% on Monday to close at $196.51. Investor Michael Burry, who has been publicly skeptical of the AI financing complex, posted “around and around we go.”

There is also the question of concentration. Nvidia had previously taken an approximately $30 billion stake in OpenAI and committed roughly $10 billion to Anthropic. Jensen Huang suggested in March that those might be the last of the company’s largest equity checks, citing both companies’ expected public listings. The Ohio talks suggest the mechanism has shifted rather than stopped — from equity into guarantees, which are cheaper to extend, harder to see in headline financials, and correlated to exactly the same underlying risk.

Layer on SoftBank. The Japanese group signed a $40 billion unsecured bridge facility in March 2026 — its largest dollar-denominated borrowing — to fund a $30 billion follow-on investment in OpenAI through Vision Fund 2. The loan carries a twelve-month maturity expiring March 2027 at roughly 6.14%, based on an initial margin of about 250 basis points over SOFR. SoftBank drew the first $10 billion tranche on April 1, with further tranches scheduled for July 1 and October 1. On July 27, Bloomberg reported that 21 new lenders had joined the syndicate, taking roughly $7 billion of the facility; First Abu Dhabi Bank, GIC and Standard Chartered each took close to $1 billion.

So the picture is: SoftBank borrows against its balance sheet to buy OpenAI equity; SoftBank’s energy arm builds the data center; Nvidia guarantees the lease so the data center can be financed; OpenAI occupies it and buys Nvidia chips, possibly with Nvidia financing. Each link is individually defensible. The chain as a whole is a set of highly correlated exposures held by a small number of very large entities, and it has been assembled in under two years.

A more skeptical reading than either extreme: the deals are probably economically rational for each participant, and the systemic concern is not fraud or delusion but correlation. Everything works so long as AI compute demand keeps growing. If it pauses, the losses do not land in one place — they land in the equity of the model companies, the debt of the developers, the guarantees of the chipmaker and the loan books of the banks simultaneously.

The memory squeeze and the arrival of a Chinese competitor

If there is one physical bottleneck that matters more than any other right now, it is memory, and the same Monday that Nvidia’s financing web made headlines, the memory market got its most significant new entrant in a decade.

ChangXin Memory Technologies — CXMT — surged 466% in its Shanghai trading debut on July 27, closing at 49 yuan against an IPO price of 8.66 yuan after touching 55.03 yuan intraday. At that close the company was valued at roughly 3.3 trillion yuan, approximately $488 billion, making it the most valuable company listed on China’s A-share market and displacing Industrial and Commercial Bank of China. The offering raised 57.92 billion yuan, about $8.6 billion — Asia’s largest IPO of 2026.

CXMT is the world’s fourth-largest DRAM producer, with roughly 7.67% of the global market as of 2025. The context for the enthusiasm is a memory market in acute shortage: DRAM contract prices rose in the region of 93% to 98% quarter over quarter in the first quarter of 2026.

Two things follow, and they are frequently conflated.

The first is the high-bandwidth memory market, where SK hynix, Samsung and Micron effectively constitute an oligopoly and where Nvidia’s HBM4 requirements are the binding constraint on GPU output. CXMT is not a near-term factor there. HBM is a fundamentally harder product with tighter qualification cycles and deeper process requirements.

The second is conventional DRAM, where prices have gone vertical precisely because HBM production is absorbing capacity. That is where CXMT is relevant, and that is where Apple enters the story. Apple does not need HBM. It needs enormous volumes of conventional memory, and it has signaled an intention to diversify away from the incumbent three. Bloomberg Intelligence’s Mandeep Singh argued on July 27 that if Apple were able to qualify CXMT as a supplier, it would be a significant event for the memory complex, because the cash flows currently being generated by DRAM scarcity would come under pressure. Micron has publicly pushed back on Apple’s approach.

For now this is a qualification question, not a market-share fact, and it should not be reported as though CXMT has already displaced anyone. What is verifiable is that a Chinese DRAM manufacturer just became China’s most valuable listed company on the strength of an AI-driven memory shortage, and that equipment makers exposed to Chinese price competition — lithography and deposition suppliers in the United States and Europe — traded lower on the news.

The open-weights letter, and a correction worth making

On July 24, Jensen Huang made his first post on X. It linked to an open letter arguing that open-weight AI models are essential to American AI leadership — that they strengthen cybersecurity, accelerate competition, widen access across industries and support national sovereignty alongside closed frontier models. The letter drew an explicit analogy to the open-source software movement of the 1980s and urged policymakers not to impose premature restrictions.

The initial signatory list ran to roughly two dozen companies, including Meta, Microsoft, IBM and Perplexity. It notably did not include OpenAI, Google or Anthropic, which prompted a widely repeated interpretation — voiced on Bloomberg Television on July 27 and elsewhere — that the three companies best positioned to monetize proprietary models had declined to endorse a document that would erode their advantage.

That interpretation was accurate about the original list and had already been overtaken by events. Forbes reported on July 25 that signatories had doubled to about 50 within a day, with OpenAI and Google among the additions. Amazon and Anthropic remained absent.

The revised picture is more interesting than the original one. Anthropic’s absence is consistent with its long-stated positions on frontier model safety and controlled deployment. Amazon’s is harder to read and has drawn less attention than it deserves, given the company’s position in cloud infrastructure and its investment in Anthropic.

The strategic subtext is not subtle. Nvidia sells accelerators. A world with many competing model developers, open weights and low switching costs consumes more accelerators than a world with three dominant labs that eventually design their own silicon — which OpenAI, Google and Amazon are all doing to varying degrees. Advocating for openness is entirely consistent with Nvidia’s commercial interest, and pointing that out is not a criticism of the argument’s merits. It is context a reader should have.

The competitive backdrop is real. Huang has publicly argued that China is producing more AI researchers than any other country and will produce excellent AI technology. Multiple industry participants have observed that a substantial share of U.S. developer traffic — measured by tokens processed — now flows to Chinese open-weight models rather than American ones. That claim is difficult to verify independently and depends heavily on how usage is measured, but the direction of travel is not seriously disputed, and it is the animating anxiety behind the letter.

Nvidia’s own numbers, and what its August report has to answer

Nvidia does not report with the rest of Big Tech. Its fiscal calendar puts the next quarterly release in late August, which means the market will spend a month speculating about financing arrangements the company has not confirmed before it hears anything official.

What is already on the record is worth restating, because it explains both the confidence and the concern. Nvidia’s most recent quarterly results showed a profit margin in the region of 75%. Its market capitalization has been hovering near $5 trillion. Those are the economics of a monopoly supplier in a shortage market, and they are what make the financing commitments plausible in the first place: a company converting three quarters of revenue into profit has unusual latitude to extend credit.

The disclosure question is where this gets uncomfortable. Nvidia’s fiscal first-quarter 2027 Form 10-Q discloses total lease-guarantee exposure capped at $3.5 billion. If the reported Ohio arrangement is executed anywhere near the scale described, that figure becomes obsolete by roughly two orders of magnitude, and Nvidia’s contingent liabilities disclosure becomes one of the most closely read sections of any technology filing.

Three questions the August report and the accompanying call will need to address, whether or not management wants to:

• Customer concentration. Nvidia has disclosed in recent filings that a small number of direct customers account for a large share of revenue. As the company extends financing across that same customer set, concentration risk and credit risk stop being independent.
• How supported demand is accounted for. If Nvidia guarantees a lease that enables a data center to be built, and that data center houses Nvidia systems purchased by a company Nvidia has invested in, the revenue is real and recognized under normal rules. Investors will nonetheless want the components broken out, and management’s willingness to provide that breakdown will be informative in itself.
• The order book versus the letters of intent. The SK Group announcement was formalized through letters of intent. Announced partnership values and contracted backlog are different things, and the gap between them across Nvidia’s growing collection of national and corporate AI agreements is now large enough to matter.

Jensen Huang’s framing of the opportunity, delivered in an interview on July 27, is that computing is shifting from roughly a billion human users to a world of “100 billion agents and billions of robots” using computers, and that the chip industry is therefore not large enough to support the computer industry being built on top of it. It is a genuinely interesting thesis about the direction of demand. It is also, unavoidably, the thesis of the person selling the equipment, and it is unfalsifiable on any timeframe relevant to a quarterly earnings model. Treating it as a vision statement rather than a forecast is the appropriate discount.

The more testable version of the same claim is the one his industry keeps making in numbers: cloud backlog, capacity utilization, memory pricing, and lead times. Those are all currently consistent with genuine scarcity. They are also all lagging indicators of a build cycle, and they were consistent with scarcity in 2000 and 2007 too, right up until they were not.

The part of the market that has not repriced: corporate credit

Equity investors have spent two weeks arguing about AI capex. Credit investors have mostly not joined in, and that divergence is the most underexamined feature of the current market.

Investment-grade spreads remain close to their tightest levels in fifteen years. Equity multiples compressed over the first half of 2026 as earnings outgrew prices; credit spreads did nothing comparable. Citi Wealth’s Kate Moore described this on July 27 as a disconnect between the equity and corporate credit environments, and said her firm prefers to take risk in equities rather than credit for that reason. That is a judgment, not a fact, but the underlying observation — that credit is not pricing the concerns equities are pricing — is straightforwardly visible in spread data.

Layered on top of that is a structural argument that has been gaining ground. Akhil Bansal, head of asset-backed finance at Carlyle Global Credit, published research this month contending that public fixed income can no longer be relied upon to deliver income, stability and capital preservation simultaneously. His central data point: the correlation between public fixed income and equities ran at approximately 0.13 over the 2010–2020 period and approximately 0.63 from 2020 to today. Speaking on Bloomberg Television on July 27, he attributed the shift primarily to the inflationary regime — inflation punishes stocks and bonds at the same time, which breaks the hedging relationship that the classic balanced portfolio depends on.

There is a second mechanism at work, and it is specific to this cycle. Corporate bond portfolios have become concentrated in the same sectors that dominate equity indices. Alphabet and Meta sit in communication services or consumer cyclical buckets in credit indices rather than technology, which understates the true concentration. Once data-center financings and chip-related structures are traced back to their ultimate credit support, a substantial and growing share of the investment-grade market rests on the obligations of a small group of very large technology companies — the same companies that dominate the S&P 500.

An investor holding a conventional 60/40 portfolio may therefore own the AI capital cycle twice: once through equity market capitalization weights, and once through credit index weights. That is a diversification problem that will not show up in a standard risk report, because the sector labels disguise it.

Bansal’s proposed alternative is asset-backed finance — private credit secured by diversified pools of assets, where the cash flows derive from consumer receivables, equipment, royalties, energy assets and similar collateral rather than from corporate earnings. His argument for the liquidity premium required is specific: hundreds of basis points, in his view, to compensate for the illiquidity of the asset class.

That claim deserves scrutiny rather than acceptance. Carlyle sells asset-backed finance products, and Bansal opened his research by disclosing exactly that, which is to his credit. But it is also true that a good deal of what is marketed as diversified private credit ultimately traces back to the same investment-grade counterparties. Bansal acknowledged as much on air: cut through many chip and data-center financings and the ultimate support is a lease or cash-flow obligation from an investment-grade corporate. Secured by an asset is not the same as uncorrelated with the AI cycle, and investors should ask precisely that question of any private credit manager pitching diversification in 2026.

Carlyle’s own positioning is instructive on this point. The firm has emphasized energy and natural gas exposure — acquiring mature cash flows and hedging them — on the view that power will be the next constraint on AI. That is a bet on the same theme through a different asset. Whether it constitutes diversification or repackaging depends entirely on what breaks.

New plumbing: single-stock futures, prediction markets and perpetuals

Three structural changes to how Americans can take financial risk landed within a week of each other, and together they say something about where the retail trading market is heading.

CME single stock futures

CME Group launched single stock futures on Monday, July 27, having announced the products on June 30. The initial slate covers more than 50 large U.S. companies through 55 standard-sized and 22 micro-sized contracts. They are cash-settled against the closing price of the underlying stock and trade roughly 23 hours a day, five days a week on Globex — far beyond the 9:30 a.m. to 4 p.m. Eastern equity session. CME is distributing through more than 35 retail intermediary firms alongside its institutional channels.

The obvious appeal is leverage and hours. The less obvious appeal is access. Because the contracts settle in cash against a reference price, they permit exposure to companies whose shares an individual investor cannot easily buy — a point that applies directly to SpaceX, which appears on the initial list.

The candid framing is that this is a product designed to widen CME’s addressable market beyond its institutional core, and that it does so by giving individual traders a leveraged instrument that trades nearly around the clock. There are legitimate hedging uses: an investor wanting to reduce exposure to a single holding without triggering a taxable sale, or without navigating options greeks, now has a cleaner tool. There are also obvious risks. Futures carry margin obligations that equity ownership does not, near-continuous trading removes natural cooling-off periods, and the majority of retail participants in leveraged products historically lose money.

Fanatics buys its own exchange

On July 27, Fanatics and BGC Group announced an agreement for Fanatics to acquire Water Street Labs, LLC — a CFTC-registered Designated Contract Market — and CX Clearinghouse L.P., a registered Derivatives Clearing Organization. Financial terms were not disclosed.

The strategic point is control. Fanatics has been offering prediction market contracts through a third-party exchange. Owning a DCM and a DCO lets it list and clear its own contracts, set its own product roadmap and keep users inside its own application. The companies also said they plan to develop market data products combining prediction market activity with traditional financial data — which is the more interesting long-term angle, and the part that would bring institutional participants into a market currently dominated by retail flow.

That imbalance is the structural weakness of prediction markets today. Pricing in any derivatives market improves when both retail and institutional capital participate; concentration in one cohort produces wide spreads and poor price discovery. BGC’s involvement is a bet on fixing that. Whether sports-led prediction markets can attract serious institutional capital remains unproven.

Kalshi extends perpetual futures into metals

Kalshi filed with the Commodity Futures Trading Commission on July 21 seeking approval to list perpetual futures on gold, silver and platinum, with copper mentioned as a possible extension. The filing triggers a review window of 45 days.

Perpetual futures — contracts with no expiry, settled in cash, kept in line with spot through periodic funding payments — originated in offshore crypto markets and have processed enormous volume there. Kalshi launched crypto perpetuals in May and has reported $16.1 billion of trading volume since. The metals contracts would initially trade 24 hours a day, five days a week, matching the underlying markets rather than the 24/7 schedule crypto permits.

Kalshi’s chief risk officer Udesh Jha argued on Bloomberg Television on July 27 that the products are not primarily competitors to dated futures. His case: perpetuals track cash prices more closely because of the funding mechanism, eliminate roll costs and the operational friction that goes with them, and are cheaper as a result. He also pushed back on the “casino-fication” criticism by pointing to leverage levels, noting that standard S&P 500 futures offer roughly 15 times leverage while some of Kalshi’s crypto perpetuals offer six to eight times.

The leverage comparison is fair as far as it goes. It also sidesteps the substantive concern, which is not the notional leverage multiple but the combination of near-continuous trading, no expiry to force position review, and a user base skewed toward individual traders. A commercial hedger rolling dated futures does so on a schedule with a purpose. A retail trader in a perpetual contract has neither.

What the derivatives are actually saying about Wednesday

There is a neat symmetry in the fact that all of this new market plumbing arrived in the same week as one of the least predictable Fed decisions in years, because the two are connected. Federal funds futures — the instrument from which the widely quoted “one-in-three chance of a hike” is derived — are themselves a derivatives market, and the number they produce is routinely over-interpreted.

The CME FedWatch tool converts fed funds futures prices into implied probabilities of specific target-range outcomes. That conversion involves assumptions: that the effective federal funds rate settles near the middle of the target range, that the market is risk-neutral, and that the only possible outcomes are the discrete quarter-point increments the tool models. None of those assumptions is exactly true. Futures prices embed risk premia as well as expectations, and in a period of unusually wide outcome distributions, the gap between “the market expects” and “the market is pricing” widens.

What can be said with confidence is narrower and more useful. The market is not confident in a hold. Roughly a third of the probability mass sits on an increase, which is far more than at any point since 2023. Short-dated Treasury yields have moved in a way consistent with that pricing. And some portion of the move in two-year yields reflects budget and issuance concerns rather than policy expectations, which means reading the two-year purely as a Fed indicator overstates the hike signal.

Prediction markets add a second, independent read. Contracts on Fed decisions trade on regulated venues in the United States, and their pricing has occasionally diverged from futures-implied odds — sometimes because prediction markets attract a different participant mix, sometimes because they are thinner and more prone to single-participant distortion. Where the two disagree, the disagreement is information, but it is not a reason to prefer the less liquid market.

The practical instruction for a reader watching Wednesday: treat 66% and 34% as a description of uncertainty rather than a forecast. A one-in-three event happens one time in three. The market has correctly identified that this meeting is genuinely uncertain, and that is close to the limit of what any pricing exercise can tell you in advance.

Corporate cross-currents: a family fortress, an activist and a takeover that will not die

Underneath the macro noise, three corporate situations moved on July 27 that illustrate how differently governance structures allocate power.

Brown-Forman rejects Sazerac again

Brown-Forman’s board rejected a renewed unsolicited takeover proposal from Sazerac valuing the Jack Daniel’s owner at approximately $15 billion. Sazerac wrote to Brown-Forman shareholders and directors asking them to reconsider an all-cash offer of $32 per share — a premium of about 23% to the July 24 close. The company had rejected an earlier $15 billion approach in May.

The decisive factor is ownership. Wolf Pen Branch, LP, a vehicle representing Brown family members, holds a majority of Brown-Forman’s Class A voting shares — reported at roughly 60% — which means no outcome is possible without family consent. The family said Sazerac’s proposal does not align with its vision for the company’s future. Whatever the merits of the price, that structure functionally forecloses a hostile route and limits what activist pressure can accomplish.

The context is a spirits industry in contraction. Consumption is falling among younger cohorts, cannabis has taken share in several U.S. states, and premium American whiskey has been working through a supply overhang. Brown-Forman’s shares have fallen sharply from their peak. Company-specific issues compound the industry ones: a portfolio critics describe as narrow and an acquisition record that has not obviously created value.

There is also an antitrust question that has drawn less attention than the price. Reporting on the proposed combination has noted that Sazerac and Brown-Forman together would account for roughly 18% of the U.S. spirits market and close to 40% of American whiskey. That is the kind of concentration that invites extended review regardless of whether the boards agree.

H.B. Fuller, Ancora and the limits of “we told you so”

H.B. Fuller is pursuing the acquisition of U.K.-listed Advanced Medical Solutions Group in an all-cash transaction valued at approximately $942 million. Ancora Holdings, which disclosed a stake above 2%, opposed the deal publicly in late May, arguing that it contradicts H.B. Fuller’s stated commitment to pause acquisitions while reducing leverage to 2.5–3.0 times net debt to EBITDA, that it would push leverage above 4.0 times, and that the company lacks experience integrating a European medical products business. Ancora also pointed to total shareholder returns of roughly negative 25% during chief executive Celeste Mastin’s tenure, and said it was prepared to pursue a proxy contest.

Mastin’s counterargument, restated on Bloomberg Television on July 27, is that the medical adhesives strategy was laid out at an investor day in October and has been communicated consistently for years; that the company evaluates roughly 20 investment priorities annually against a buy-or-build test in a fragmented $80 billion industry; and that EBITDA margins have expanded by approximately 400 basis points over the course of the plan. She acknowledged deleveraging will be required after the AMS acquisition closes.

Both accounts can be partly true. A company can have flagged a strategic direction for years and still have said something specific about pausing deals that it then did not do. The distinction Ancora is pressing is between strategic intent and capital allocation sequencing, and it is a fair one. AMS shareholders vote on the transaction in mid-August.

Mastin’s operational commentary was more informative than the governance dispute. She said building adhesives posted 6% organic growth in the second quarter, well ahead of the construction industry; that durable goods exposure through engineered adhesives is under pressure globally; and that 99% of what H.B. Fuller sells in the United States is made in the United States, which limits direct tariff exposure while leaving the company importing raw materials. Asked to rank the disruptions, she was unambiguous: the Strait of Hormuz has had a larger impact on her business than tariffs, because global petrochemical supply and demand has been shifting rapidly and raw material sourcing has had to move between regions.

That is a useful data point from the real economy, and it cuts against the assumption that the Middle East conflict is primarily a financial-market story. For a global industrial company, it is a supply-chain story first.

Where the professionals disagree, and how to weigh them

The institutional views aired around this week’s decision cluster into three camps. All three are held by serious people. None is disinterested.

The constructive equity camp. Kate Moore, chief investment officer at Citi Wealth, argued on July 27 for a full allocation to equities with an overweight to U.S. large caps, on the grounds that 2026’s gains have come from earnings rather than multiple expansion, that fundamentals continue to come in strong, and that the worries currently weighing on sentiment — geopolitics, inflation, rates, AI sustainability — will fade. She expects the next Fed move to be a hike but does not expect it this week, and she is not concerned about recession. Her caveat is specific and worth noting: she thinks a pause in AI capital spending would be a shock to sentiment, and that the market has been range-bound for months pending a broader set of investors becoming comfortable adding risk.

The strength of this view is that it is grounded in verifiable data — earnings growth, revenue growth, beat rates and a forward multiple that has compressed rather than expanded. Its weakness is that it requires several independent worries to resolve favorably at once, and Citi Wealth’s business is served by clients remaining invested.

The cautious allocator camp. Kelly Covley, senior investment analyst at Manning & Napier, made the more nuanced argument: that earnings growth of this magnitude outside a recovery from recession is unusual, that investors are right to question earnings quality when so much capital expenditure sits between reported earnings and free cash flow, and that the greater market risk may lie in conservative capex guidance rather than aggressive guidance. She also flagged signs that investors are having a harder time absorbing the debt issuance coming out of the hyperscalers, and expects prolonged upward pressure on energy prices sufficient to keep the Fed hawkish.

This view has the merit of identifying a specific transmission mechanism — debt absorption — rather than gesturing at valuation. Its weakness is that “investors are having a harder time digesting issuance” is difficult to verify precisely, and spread data has not yet corroborated it.

The structural credit camp. Carlyle’s Akhil Bansal argues that the diversification premise underlying conventional fixed income has broken, supported by a correlation shift from roughly 0.13 in 2010–2020 to roughly 0.63 since 2020. His prescription is asset-backed finance with a liquidity premium measured in hundreds of basis points.

The correlation data is the strongest empirical claim any of the three camps has made, and it is consistent with what an inflationary regime should produce. The prescription is where interest enters: Carlyle sells the alternative, Bansal disclosed as much, and he conceded on air that much of what is marketed as diversified private credit ultimately rests on investment-grade corporate obligations. An investor persuaded by the diagnosis should test the remedy hard.

What none of these views claims, and what a reader should be sceptical of anywhere it appears: that anyone knows what the Fed will do on Wednesday, or what the four Big Tech reports will say. The useful function of institutional commentary in a week like this is to identify which variables matter, not to predict their values.

Historical comparisons: which analogy actually fits

Every cycle attracts analogies, and most of them are lazy. Three are worth examining seriously.

1994. The comparison most often reached for by those expecting a hike is the Fed’s aggressive tightening into a strong economy, which detonated the bond market and forced a global repricing of duration. The parallel is real in one respect: a market positioned for easing that has to reverse. It is weak in another. In 1994 the Fed surprised markets by moving early against inflation that had not yet appeared. In 2026 inflation has already appeared, has been above target for an extended period, and the committee has been transparently hawkish. The surprise element is largely absent.

1999–2000. The comparison reached for by AI skeptics. A capital-spending boom in a general-purpose technology, financed increasingly with debt and vendor credit, with suppliers extending financing to customers who could not otherwise buy. Lucent and Nortel financed carrier equipment purchases; when the carriers stopped, the receivables went bad and the suppliers collapsed. The structural rhyme is uncomfortable, and it is why the phrase “circular financing” carries the weight it does.

But the differences are substantial and should not be waved away. The 2026 hyperscalers are extraordinarily profitable businesses with enormous operating cash flow funding the majority of their capex from internal resources. Nvidia’s margins are not those of a telecom equipment vendor stuffing the channel. And crucially, the demand shortfall in 2000 was visible in falling utilization rates before it appeared in financials, whereas current evidence points to utilization running at capacity. The bear case requires demand to disappoint from here; it cannot point to it having already done so.

2006–2007. The comparison for the credit-market angle. Tight spreads, elevated long yields, a concentrated set of correlated exposures, and a structure in which risk was distributed in ways that were poorly understood until they were tested. The 30-year Treasury has not held above 5% this persistently since 2007, which is a coincidence of levels rather than a causal parallel. The more relevant echo is Bansal’s point about correlation: the assumption that a diversified portfolio is diversified is exactly the assumption that failed in 2008, and the mechanism was labeling, not fraud.

None of these maps cleanly. The honest summary is that the equity market looks like a late-stage capital cycle with genuinely strong fundamentals, the credit market looks complacent relative to the equity market, and the macro backdrop is an energy-driven inflation shock of uncertain duration. Those three things have coexisted before, and the outcomes have varied enormously depending on whether the shock persisted.

The economy underneath the market

It is easy to lose the actual economy in a week like this. Three observations keep it in view.

Growth is slowing, but not alarmingly. Regional Federal Reserve tracking estimates had current-quarter growth running near 1.7% as of late July. That is below trend but not recessionary, and it sits alongside financial conditions that remain unusually easy by most standard measures — a combination that gives hawks on the committee a straightforward argument. If conditions are loose and inflation is above target, the case for restraint does not depend on growth being strong.

Investment is the surprise. The capital spending story is not confined to technology. Overall investment in the economy has been running strongly, and the second-order effects of the AI build-out have propagated into utilities, industrials, construction and real estate. Kate Moore made this point on July 27 and it is worth taking seriously: a large share of the current cycle’s economic momentum is being generated by a small number of companies’ capital budgets. That is a source of strength and a concentration risk in the same sentence.

Economic surprises have been consistently positive. Data has come in ahead of forecasts across consumer and business measures through the back half of 2025 and into 2026. Two readings of that are available. The generous one is that the economy is more resilient than economists modeled. The less generous one is that forecasters have been persistently too conservative, which is a statement about forecasters rather than about the economy.

Moore’s conclusion — that she is not worried about the economy slipping into recession, but that a pause in AI capital spending would be a shock to sentiment rather than to output — is a reasonable synthesis. The distinction between a sentiment shock and an economic shock is real, and worth holding onto. Markets can fall a long way on sentiment while GDP keeps growing.

How markets actually traded into the decision

Monday, July 27 was a useful preview of how conflicted positioning has become.

Equities opened higher, with the S&P 500 up roughly 0.8% and the Nasdaq 100 up more than 1% in the first minutes of trading, led by the megacap names due to report. Ten-year yields fell about three basis points to roughly 4.65%. Brent crude slid as the U.S.–Iran pause held for a third night, dragging energy shares lower while airlines and cruise operators rallied. Small caps outperformed on the lower yields.

By mid-morning the move had reversed. The Nasdaq turned negative and was down roughly 0.8% about seventy minutes into the session. Semiconductor equipment makers — Lam Research and Applied Materials among them — fell on the CXMT listing and the prospect of cheap Chinese chipmaking equipment. Nvidia, initially fluctuating, ended the day down 4.99% at $196.51 as the Ohio financing reports circulated.

The pattern is worth naming precisely, because it recurs: on days when the companies writing the capital expenditure checks rally, the companies receiving them fall. That inverse relationship is not sustainable indefinitely — the two groups’ fortunes are ultimately linked — but it accurately describes a market trying to work out how the economics of the AI build-out will be split between buyers and sellers.

For the year to date, Nvidia was up roughly 4% against a 7.8% gain for the S&P 500 as of July 27, according to Benzinga Pro data. That is a striking underperformance for the company most identified with the AI trade, and it says something about how much of the story is already priced.

Risks that would change the picture

These are the specific, material risks worth monitoring rather than a generic list.

• A resumption of U.S.–Iran hostilities. The pause is informal and reciprocal. A breakdown would push crude and shipping insurance higher, reverse the June inflation improvement and remove the Fed’s principal reason for patience. This is the single largest macro variable.
• Hyperscaler capex guidance in either direction. Higher guidance risks a repeat of the Alphabet reaction in the four largest names within 48 hours. Conservative guidance risks a synchronized derating across memory, semiconductor capital equipment, utilities, industrials and data-center property.
• The Ohio financing failing to complete — or completing. If Nvidia formalizes a guarantee approaching the reported scale, the company’s credit profile changes materially and the disclosure question becomes urgent. If talks collapse, the market will ask who else can finance a project of that size.
• Long-end supply. Heavy Treasury issuance combined with record investment-grade issuance from technology companies competes for the same duration buyers. A poorly received auction would tighten financial conditions without any Fed action.
• Credit spread repricing. Investment-grade spreads near fifteen-year tights leave little cushion. Because credit index concentration mirrors equity index concentration, a technology-led equity drawdown would not be offset by bonds in the way historical correlations imply.
• Memory cost pass-through. DRAM contract prices rising at the observed pace feed into consumer electronics bills of materials. That is a margin question for Apple and an inflation question for the Fed, and it will show up with a lag.
• Policy on open models and chip exports. Washington is actively considering restrictions on Chinese AI models. Any rule that constrains open-weight deployment, or that changes export licensing, would alter both the competitive landscape and the demand forecast underpinning current capital plans.
• Concentration itself. Four companies reporting within 48 hours, together representing roughly 17% of the S&P 500, is an unusual amount of index-level risk compressed into two sessions.

What happens next: the confirmed calendar

Separating scheduled events from expectations:

• Tuesday, July 28: FOMC meeting begins. Prime Minister Netanyahu meets President Trump at the White House. Ford reports second-quarter results.
• Wednesday, July 29: FOMC decision at 2:00 p.m. ET, Chair Warsh press conference at 2:30 p.m. ET. Microsoft (fiscal Q4) and Meta Platforms (Q2) report after the close.
• Thursday, July 30: Bureau of Economic Analysis publishes June personal income and outlays, including the PCE price index, at 8:30 a.m. ET. Apple (fiscal Q3) and Amazon (Q2) report after the close.
• Mid-August: Advanced Medical Solutions shareholders vote on the H.B. Fuller acquisition.
• Early September: Kalshi’s 45-day CFTC review window on precious metals perpetual futures concludes.
• Before the next FOMC meeting: Two CPI reports, two PPI reports, two PCE reports and two employment reports.
• October 1, 2026: SoftBank’s third $10 billion tranche of its OpenAI bridge facility is scheduled to draw.
• March 2027: SoftBank’s $40 billion bridge facility matures.

Analyst expectations, which are not scheduled events: FactSet’s compiled consensus calls for S&P 500 earnings growth of 27.3% in the third quarter, 24.9% in the fourth, and 27.3% for calendar 2026. Those figures are estimates, they will move, and the second-quarter experience shows how much a single company’s accounting can distort an index-level aggregate.

A reader’s guide to the numbers you will see this week

A great deal of what gets published over the next four days will be technically accurate and practically misleading. Five distinctions do most of the work.

Percentage points are not percent. If the Fed raises the target range from 3.50%–3.75% to 3.75%–4.00%, that is an increase of 25 basis points, or a quarter of a percentage point. It is not a 25% increase; in proportional terms it is closer to 7%. Basis points exist precisely to avoid this confusion — one basis point is one hundredth of a percentage point.

GAAP is not adjusted, and neither is cash. Alphabet’s $9.11 in GAAP diluted earnings per share included a $98 billion gain. Companies routinely present adjusted figures that exclude items management considers non-recurring, and analysts’ consensus estimates are usually built on adjusted measures. Comparing an adjusted actual to a GAAP estimate, or the reverse, produces nonsense. When a headline says a company “beat by 300%,” check which basis is being compared before drawing conclusions.

Blended growth is not reported growth. FactSet’s blended rate combines actual results from companies that have reported with estimates for those that have not. It changes daily and it is heavily influenced by whichever large companies happen to have reported. The 37.9% figure circulating this week is a blended rate distorted by one company; the number that will eventually be recorded for the second quarter will look different.

Operating cash flow is not free cash flow, and capex is not an expense. Capital expenditure does not hit the income statement when it is spent; it hits the cash flow statement immediately and the income statement gradually, through depreciation. This is why a company can report record earnings and shrinking free cash flow in the same quarter, and why the AI capex debate is fundamentally a question about the timing mismatch between cash out and revenue in.

Announced value is not contracted revenue. Nvidia’s “$500-billion-plus” partnership with SK Group was formalized through letters of intent covering multiple entities and multiple directions of commitment. A reported $250 billion guarantee is a contingent liability under discussion, not cash. Announced deal values, letters of intent, memoranda of understanding, backlog, remaining performance obligations and recognized revenue are five different things, and the AI infrastructure story has produced headlines conflating all five.

One more, specific to this week: an inflation report that shows prices falling on the month while the annual rate remains at 3.5% is a report about the rate of change, not the level. Nothing has become cheaper. Coverage that describes June CPI as prices “coming down” is describing deceleration, and households will not recognize the description.

Frequently asked questions

Will the Fed raise interest rates on July 29, 2026?

Most economists expect the FOMC to hold the federal funds target range at 3.50%–3.75%, which would be the fifth consecutive meeting without a change. Federal funds futures implied roughly a 66% probability of no change as of July 27, leaving close to a one-in-three chance of a quarter-point increase. No Fed official has signaled a hike at this specific meeting.

What is the current federal funds rate?

The target range has been 3.50%–3.75% since the March 2026 meeting. It was held there at the June 17 meeting, Chair Kevin Warsh’s first.

Why would the Fed hike when inflation just fell?

Because the June improvement was concentrated in energy and the underlying rate remains above target. Headline CPI fell 0.4% on the month but is still up 3.5% over twelve months; producer prices for final demand are up 5.5% over the same period; and the May PCE price index, which the Fed actually targets, ran at 4.1% headline and 3.4% core. Officials have described inflation as broad enough and persistent enough to warrant a tightening bias rather than an easing one.

When is the June PCE report released?

The Bureau of Economic Analysis publishes June personal income and outlays, including the PCE price index, on Thursday, July 30, 2026, at 8:30 a.m. ET — one day after the Fed decision.

Which Big Tech companies report earnings this week, and when?

Microsoft (fiscal fourth quarter) and Meta Platforms (second quarter) report on Wednesday, July 29. Apple (fiscal third quarter) and Amazon (second quarter) report on Thursday, July 30. FactSet counted 177 S&P 500 companies reporting across the week.

Is S&P 500 earnings growth really approaching 40%?

No, not in any operating sense. FactSet’s blended second-quarter growth rate of 37.9% as of July 24 is heavily distorted by Alphabet, whose GAAP diluted EPS of $9.11 included a $98 billion gain against a consensus of $2.88. Excluding Alphabet, the blended growth rate is 25.9% and the aggregate earnings surprise falls from 39.3% to 12.6%.

What exactly did Nvidia announce, and what is only reported?

Confirmed by company announcements: a “$500-billion-plus” partnership with SK Group covering an SK Telecom AI factory and SK hynix memory supply, formalized through letters of intent on July 24; and a long-term partnership with and investment in Safe Superintelligence announced July 27, reported by Bloomberg at $5 billion. Reported but not confirmed by any party: discussions to guarantee up to $250 billion of lease obligations for an OpenAI data center in Ohio, and separately to finance up to $350 billion of OpenAI chip purchases. Those talks are described as early stage and could fall apart.

What is circular financing, and is Nvidia doing it?

Circular financing describes a supplier providing capital or credit support to a customer that then buys the supplier’s products. Nvidia’s position is that it does not contractually require recipients to purchase its chips, so no quid pro quo exists. Critics argue that the distinction between independent demand and supplier-supported demand becomes unreadable at sufficient scale. Both characterizations are defensible; what is not in dispute is that Nvidia has extended equity, commitments and potentially guarantees across a large share of its own customer base.

Why did Nvidia stock fall on July 27?

Shares closed down 4.99% at $196.51 following reports of the Ohio financing discussions. Prices for credit default swaps on Nvidia bonds recorded their largest intraday increase since they began actively trading in November, according to ICE Data Services data cited by Bloomberg — indicating the concern was about balance-sheet risk rather than demand.

What is CXMT and why did it surge 466%?

ChangXin Memory Technologies is China’s largest DRAM manufacturer and the world’s fourth largest, with roughly 7.67% of the global DRAM market as of 2025. It closed its Shanghai debut on July 27 at 49 yuan against an 8.66 yuan IPO price, valuing it near 3.3 trillion yuan — approximately $488 billion — and making it China’s most valuable listed company. DRAM contract prices rose in the region of 93% to 98% quarter over quarter in the first quarter of 2026 as AI demand absorbed capacity.

Why is the 30-year Treasury yield above 5% a problem?

It has held above 5% for its longest run since 2007, reaching an intraday 5.19% on May 19 — the highest since summer 2007. Long-term yields set borrowing costs for mortgages, corporate refinancing and project finance regardless of what the Fed does with the overnight rate. The move reflects inflation compensation and heavy supply, including record technology-sector issuance, and is largely outside the Fed’s direct control.

Did Brown-Forman accept Sazerac’s offer?

No. The board rejected a renewed proposal valuing the company at roughly $15 billion, or $32 per share in cash — a premium of about 23% to the July 24 close. Sazerac had been rebuffed on similar terms in May. Wolf Pen Branch, LP, representing Brown family shareholders, controls a majority of Class A voting shares and said the proposal does not align with its vision for the company.

Final assessment

The most likely outcome on Wednesday is a hold, and the most consequential part of the afternoon will not be the decision.

What matters is whether Kevin Warsh, having spent his first two months systematically removing forward guidance from the Fed’s toolkit, gives markets anything to work with on how the committee weighs an energy-driven inflation shock that is currently reversing against a core rate that has been above target for years. If he does not — and his track record suggests he will not — the market will keep pricing a wide distribution of outcomes, which means continued volatility in short-dated rates and continued pressure at the long end. That is a policy choice with market consequences, and it is worth watching whether the committee is comfortable with them.

The strongest evidence for patience is on the calendar. The June PCE report lands twenty-four hours after the decision, and two of everything arrives before September. A committee that wanted to hike had a cleaner opportunity in June, when energy prices were still rising, and passed.

The strongest credible concern runs the other way, and it is not really about the policy rate at all. It is that the AI capital cycle has quietly become a macro variable. Four companies’ capital budgets are driving a meaningful share of investment growth, generating a meaningful share of index earnings growth, absorbing a growing share of investment-grade issuance and — through Nvidia’s expanding financial commitments — introducing correlated exposures across equity, credit and private markets simultaneously. The credit market spent Monday repricing that risk in Nvidia’s default swaps while investment-grade spreads more broadly stayed near fifteen-year tights. Those two facts cannot both be right for long.

What changed this week is not the Fed’s reaction function. It is the visibility of the financing chain underneath the AI build-out. A month ago the debate was whether hyperscalers were spending too much of their own cash. Now the question is who is lending, guaranteeing and underwriting, and against whose balance sheet — and the answer increasingly runs back to a single chipmaker whose own filings disclose a lease-guarantee ceiling of $3.5 billion.

What remains genuinely uncertain: whether the Ohio arrangement is signed and on what terms; whether the U.S.–Iran pause survives the week; whether June PCE confirms the CPI improvement; and whether Microsoft, Meta, Apple and Amazon can raise capital spending without repeating Alphabet’s reaction. Any one of those resolving badly would change the picture. All four resolving well would leave the market roughly where it started, which is at a forward multiple of about 20 times with an unusually concentrated set of dependencies.

Watch three things this week in this order: the language in Wednesday’s statement about the balance of risks, the capex lines in the four Big Tech reports, and any confirmation — or denial — from Nvidia or OpenAI about Ohio.

This article is provided for general informational purposes and does not constitute financial, investment, tax, or legal advice.

Sources

• NVIDIA Newsroom — SK Group and NVIDIA Expand Strategic Partnership Across AI Factories and Next-Generation Memory, July 24, 2026
• NVIDIA Newsroom — Ilya Sutskever’s Safe Superintelligence Inc. and NVIDIA Announce Long-Term Strategic Partnership, July 27, 2026
• NVIDIA Corporation — SEC filings and investor relations
• U.S. Bureau of Labor Statistics — Consumer Price Index Summary, June 2026
• U.S. Bureau of Labor Statistics — Producer Price Index News Release, June 2026
• U.S. Bureau of Economic Analysis — Personal Income and Outlays, May 2026
• Federal Reserve — FOMC calendars, statements and minutes
• U.S. Bank Asset Management Group — Fed holds rates steady as new Chair Kevin Warsh commits to price stability, June 17, 2026
• CBS News — Will the Federal Reserve raise interest rates? What experts predict for July’s meeting
• FactSet Insight — S&P 500 Earnings Season Update, July 24, 2026
• Advisor Perspectives — S&P 500 Snapshot: Index Ends Choppy Week in the Red, July 24, 2026
• Fox Business — Consumer inflation cooled more than expected in June as gas prices fell
• Axios — Nvidia reignites “circular” financing concerns as it weighs OpenAI deal, July 27, 2026
• Benzinga — Nvidia Funds AI Frenzy: Timeline of its Circular Financing Deals, July 27, 2026
• CNBC — Chipmaker CXMT’s 466% market debut surge makes it the most valuable China-listed company
• CNBC — Strait of Hormuz traffic: renewed U.S.-Iran conflict chokes Hormuz, July 21, 2026
• CNBC — Oil prices slide, Brent crude below $90 as pause to U.S.-Iran hostilities appears to hold
• NBC News — Oil prices slide as U.S. and Iran pause strikes to give ‘space’ for diplomacy
• CNN — Why this time around, Netanyahu’s meeting with Trump might not go to script
• CME Group — CME Group to Launch Single Stock Futures on July 27, June 30, 2026
• Fanatics Inc. — Fanatics to Acquire Exchange and Clearinghouse from BGC, July 27, 2026
• Forbes — Nvidia Open Weights Letter Doubled To 50 Without Amazon And Anthropic, July 25, 2026
• Tom’s Hardware — Nvidia and 24 other companies sign open-weights letter as Washington weighs Chinese AI model ban
• Carlyle — Akhil Bansal, Head of Asset-Backed Finance, Carlyle Global Credit
• Business Wire — Ancora letter regarding opposition to H.B. Fuller’s acquisition of Advanced Medical Solutions, May 26, 2026
• BeverageDaily — Brown-Forman rejects Sazerac takeover bid, July 27, 2026
• The Japan Times — SoftBank’s $40 billion loan for OpenAI stake gets 21 new lenders, July 27, 2026
• Mining.com — Kalshi seeks approval to list perpetual futures tied to gold, July 2026
• Investing.com — Alphabet Q2 2026 earnings call transcript and capex guidance
• Marketplace — Yield on 30-year Treasury tops 5% for longest period since 2007, July 22, 2026

Affiliate disclosure: Businessfinance.news may earn compensation from qualifying actions completed through selected links on this website, at no additional cost to the reader. Affiliate relationships do not influence our editorial reporting, analysis, or conclusions.