How AI Investment Gains Distorted S&P 500 Earnings

Last updated: August 4, 2026, 5:30 a.m. EDT

The headline number from the second-quarter earnings season looked extraordinary: profits for the S&P 500 were growing 47.4% from a year earlier, the strongest pace since the post-pandemic rebound of 2021. Yet a large part of that apparent boom did not come from selling more cloud computing, advertising, software, retail goods, or subscriptions. It came from accounting gains tied to investments in other technology companies.

Amazon recorded $53.4 billion of non-operating pre-tax other income, primarily from its investments in Anthropic. Alphabet reported $99.0 billion of net gains on equity securities, driven largely by its holdings in SpaceX and another private company. Microsoft’s quarterly earnings were also affected by its investments in OpenAI, while its finance chief separately disclosed a $3.2 billion gain from Anthropic. These gains were real under U.S. generally accepted accounting principles, but most were not operating profits and did not represent cash generated by the companies’ main businesses during the quarter.

The effect on the market’s earnings statistics was enormous. FactSet calculated that the blended year-over-year earnings-growth rate for the S&P 500 reached 47.4% after Amazon reported. Excluding Alphabet and Amazon, that rate would have been 28.8%. The gap—18.6 percentage points—does not mean the rest of corporate America had a weak quarter. A 28.8% increase would still be unusually strong. It does mean that the index-level headline overstated the acceleration in recurring business profitability.

This distinction matters because investors, policymakers, business owners, and households often treat aggregate earnings growth as a signal about the health of the economy and the durability of stock-market valuations. When a handful of giant companies book tens of billions of dollars of mark-to-market gains, an index can appear to be producing a broad surge in profits even though much of the increase reflects revaluation of financial assets. The companies have not necessarily done anything improper. The challenge is one of interpretation.

The clearest way to read the quarter is to hold two ideas at once. First, Big Tech’s underlying operations were generally strong: Amazon Web Services accelerated sharply, Alphabet’s advertising and cloud businesses grew, and Microsoft’s cloud platform continued to expand. Second, the reported net-income figures were boosted by investment gains so large that conventional comparisons of earnings per share became misleading unless readers separated operating performance from portfolio revaluations.

Key Takeaways

• S&P 500 earnings growth: FactSet’s blended estimate for the second quarter of 2026 rose to 47.4%, compared with 23.2% at the end of the quarter before most companies had reported.
• Concentration effect: Excluding Alphabet and Amazon, the S&P 500 growth rate would have been 28.8%, still strong but far below the headline figure.
• Amazon: Net income rose to $62.6 billion, but the quarter included $53.4 billion of non-operating pre-tax other income, primarily from Anthropic-related investments. Operating income increased 43% to $27.5 billion.
• Alphabet: Net income rose 298% to $112.1 billion. A $99.0 billion gain on equity securities added $77.1 billion to net income after taxes and $6.26 to diluted earnings per share.
• Microsoft: GAAP net income rose 31% to $35.8 billion. Excluding the effect of OpenAI investments, Microsoft reported 22% growth in adjusted net income, while a separate $3.2 billion Anthropic gain remained an important quarterly item.
• Accounting reality: The gains comply with GAAP rules that require changes in the value of many equity investments to flow through net income, even when the shares have not been sold.
• Investor lesson: Revenue, operating income, cash flow, capital expenditure, and segment results offered a cleaner view of the companies’ core performance than headline net income alone.

What Actually Happened to S&P 500 Earnings

Corporate earnings seasons are measured with a moving blend of actual results and analyst estimates. Companies that have already reported contribute their actual earnings; companies that have not yet reported remain represented by consensus forecasts. That means the aggregate growth rate changes as each large company releases results.

At the end of June, before most second-quarter reports had arrived, FactSet’s blended estimate called for S&P 500 earnings to increase 23.2% from the same period in 2025. By July 31, after 61% of the index had reported, the growth rate had jumped to 47.4%. The shift was not simply the usual pattern in which companies beat conservative forecasts by a few percentage points. Two unusually large GAAP earnings surprises transformed the index calculation.

Alphabet reported diluted earnings per share of $9.11, compared with the $2.88 consensus estimate cited by FactSet. Amazon reported $5.75, compared with an estimate of $1.82. Those differences were so large because the reported figures included the investment-related gains. FactSet said Amazon alone accounted for 76% of the net dollar-level increase in S&P 500 earnings during the week in which it reported. Alphabet was the largest contributor to year-over-year growth for the quarter, with Amazon second.

The concentration is easier to understand when measured in dollars rather than percentages. Alphabet’s $112.1 billion of quarterly net income exceeded its year-earlier profit by roughly $83.9 billion. Amazon’s $62.6 billion exceeded the comparable period by about $44.4 billion. Together, those two companies added more than $128 billion of year-over-year net income. Because both are among the largest members of the index, their results had enough weight to shift the entire S&P 500 earnings-growth rate.

These numbers were not fictional. Alphabet’s holdings became more valuable, and Amazon’s Anthropic-related securities were remeasured upward after observable financing transactions. The economic value of the companies’ portfolios increased. Under GAAP, those changes belonged in earnings. But the gains were qualitatively different from profit earned through normal operations.

A cloud customer paying Amazon for computing capacity generates revenue, operating profit, cash receipts, and potentially recurring demand. An upward revaluation of Anthropic preferred stock produces accounting income without a customer transaction, without operating revenue, and generally without immediate cash. The investment may later be sold for more than its carrying value, validating the gain. It may also fall in value before sale, reversing part of the benefit.

Index-level earnings statistics usually compress those distinctions. A dollar of net income from advertising and a dollar of net income from a securities revaluation are both included in the aggregate. That is appropriate for a standardized accounting measure, but it can obscure the question investors often believe they are asking: how much more profitable were the underlying businesses?

The answer was still encouraging. Excluding Alphabet and Amazon entirely—not merely removing their investment gains—FactSet found that S&P 500 earnings were growing 28.8%. Eight sectors were reporting double-digit profit growth, and the earnings-surprise rate excluding the two companies remained above five- and ten-year averages. The quarter therefore was not a mirage. It was a strong earnings season made to look exceptional by a narrow set of revaluation gains.

Why GAAP Put the Gains in Net Income

The accounting treatment dates to changes adopted by the Financial Accounting Standards Board. Under Accounting Standards Update 2016-01, changes in the fair value of equity securities generally run through net income rather than accumulating in other comprehensive income. The objective was to create a more consistent measurement model and ensure that changes in the value of equity holdings were visible in reported earnings.

For publicly traded shares, the mechanism is straightforward. A company can observe the market price at the reporting date and adjust the carrying value of its investment. For shares in private companies, no continuously quoted price exists. GAAP therefore allows a measurement alternative for certain equity securities without a readily determinable fair value. A company may carry the investment at cost, less impairment, while adjusting it upward or downward when an observable transaction involving the same issuer provides evidence of a new value.

FASB’s 2018 technical update clarified that an observable transaction in a similar security may require adjustments for differences in rights and obligations. Preferred shares, convertible notes, common shares, and warrants are not automatically equivalent. Their liquidation preferences, conversion rights, voting provisions, sale restrictions, and seniority can produce materially different values even when issued by the same company.

That is why Amazon’s filing described a Level 3 valuation process rather than a simple multiplication of shares by a public price. Its Anthropic preferred securities were adjusted using information from financing rounds, but Amazon also considered the rights attached to each class, expected timing and type of liquidity events, anticipated offerings, and discounts for lack of marketability. Those inputs require judgment.

Alphabet’s situation included both private and public securities. SpaceX completed its public offering in June 2026, creating a visible market price for shares that previously had been difficult to value. Alphabet’s quarterly filing said $77.4 billion of its net gain came from non-marketable equity securities measured under the alternative, while another $21.4 billion came from marketable and other equity securities. The result was $99.0 billion of total equity-security gains in one quarter.

Recognizing these gains in net income has a defensible economic logic. Shareholders own the investment portfolio, and an increase in its value increases the company’s net assets. Excluding it permanently would understate the value created by successful capital allocation. Alphabet’s early investment in SpaceX, for example, became one of the most valuable venture investments ever held by a public company. Amazon’s strategic support of Anthropic may also produce substantial financial returns in addition to commercial benefits for AWS.

The problem arises when a period’s change in value is interpreted as repeatable operating performance. Fair-value gains are often volatile, sensitive to financing terms, and disconnected from the sales cycle of the parent company. They may dominate net income in one quarter and reverse in another. GAAP answers the question, “How did the company’s recognized net assets and earnings change under standardized rules?” It does not always answer, “How much did the core business earn from customers?”

For that reason, analysts frequently examine operating income, segment profit, cash flow, and adjusted earnings. Those measures can illuminate recurring performance, but they introduce their own risks. Non-GAAP definitions vary by company, and management teams may exclude recurring costs while describing them as exceptional. The correct response is not to replace GAAP with any single adjusted number. It is to reconcile the measures and understand what each one captures.

Amazon: A Strong Operating Quarter Hidden Beneath a Bigger Anthropic Gain

Amazon’s second-quarter results provide the clearest example of how an investment gain can overwhelm an otherwise impressive operating report. According to the company’s earnings release, net sales increased 20% year over year to $200.6 billion. Operating income rose 43% to $27.5 billion. AWS sales increased 37% to $42.2 billion, the cloud unit’s fastest growth in 18 quarters, while AWS operating income climbed to $16.6 billion from $10.2 billion.

Those figures would ordinarily have been the center of the earnings story. Instead, Amazon reported net income of $62.6 billion, up from $18.2 billion a year earlier, and diluted earnings per share of $5.75. The reason net income exceeded operating income by such a wide margin was $53.4 billion of non-operating pre-tax other income, primarily from Anthropic-related investments.

Amazon’s Form 10-Q disclosed that the company recorded approximately $50.5 billion of upward adjustments to Anthropic nonvoting preferred stock during the quarter and $62.8 billion during the first half of 2026. The adjustments reflected observable price changes associated with Anthropic financing transactions. Amazon also recognized gains when portions of convertible notes were converted into preferred stock.

The scale of the balance-sheet revaluation was striking. Amazon’s carrying amount for Anthropic nonvoting preferred stock increased from approximately $14.8 billion at the end of 2025 to $92.5 billion at June 30, 2026. Its Anthropic convertible notes were carried at an estimated fair value of $97.9 billion, including $92.0 billion of unrealized gains recorded in accumulated other comprehensive income. Equity investments in private companies—primarily Anthropic and OpenAI—had a combined carrying value of $122.3 billion at quarter-end, up from $16.2 billion six months earlier.

Those figures require careful reading because different instruments were accounted for differently. The preferred-stock adjustments flowed through other income and net income. Some unrealized gains on convertible debt accumulated in other comprehensive income until conversion or another triggering event. The parent company’s economic exposure therefore was larger than the amount that appeared in quarterly net income, while the timing of recognition depended on the type of security.

Amazon had also expanded its direct investment commitments. During the second quarter, it exercised an option to invest $5.0 billion in Anthropic Series H nonvoting preferred stock, reducing an available financing facility to $15.0 billion. The filing said Amazon invested $10.0 billion in Anthropic preferred stock during the quarter when all relevant transactions were included. It had separately invested $28.7 billion in OpenAI Series C preferred stock by June 30 and funded another $21.3 billion after quarter-end.

These investments were intertwined with commercial agreements. Amazon said AWS and Anthropic expanded an existing multiyear commitment by more than $100 billion over ten years, including obligations related to AWS chips. AWS and OpenAI had expanded a separate commitment by $100 billion over eight years. Such contracts strengthen Amazon’s position as a supplier of cloud capacity and custom silicon to frontier-model developers, but they also create a network in which Amazon is simultaneously an investor, lender or financing partner, cloud provider, and customer collaborator.

That does not make the transactions circular in a purely artificial sense. Anthropic and OpenAI need enormous amounts of computing capacity; AWS needs large, credible customers to justify data-center investment; equity financing gives the model developers capital to purchase infrastructure and fund research. Each contract can have commercial substance. The concern is that the relationships can make it harder to separate independent customer demand from demand supported by the same strategic capital providers that later recognize valuation gains.

Amazon’s Core Business Was Strong Even Without the Gain

Removing the Anthropic revaluation from the narrative should not lead to the opposite mistake of dismissing Amazon’s quarter. The company’s operating income grew by $8.3 billion from a year earlier. AWS accounted for $6.4 billion of that improvement. North American operating income increased to $9.1 billion from $7.5 billion, and the international segment remained profitable at $1.7 billion.

The most important operational development was AWS growth. A 37% sales increase represented a significant acceleration for a business already generating more than $40 billion of quarterly revenue. AWS also produced an operating margin of about 39%, calculated from $16.6 billion of operating income on $42.2 billion of sales. That combination of accelerating revenue and high profitability supported the market’s positive reaction to the report.

Amazon’s advertising business grew 26%, while subscriptions and retail operations benefited from faster delivery and continued consumer demand. The company’s results therefore contained a genuine improvement in sales, margins, and operating profit. The earnings distortion did not transform a weak quarter into a strong one. It transformed a strong quarter into an apparently historic profit explosion.

The cleanest measure for comparison was operating income because it excluded the non-operating securities gain. Even that measure is not perfect: it includes stock-based compensation, depreciation, restructuring effects, and the costs of investments that may produce future growth. It nevertheless captured the performance of Amazon’s operating segments far better than net income did.

Cash Flow Told a More Cautious Story

Amazon’s cash-flow figures showed the cost of building the infrastructure behind its AI ambitions. Operating cash flow increased 33% to $161.4 billion for the trailing twelve months. Yet free cash flow fell from an inflow of $18.2 billion to an outflow of $7.6 billion because purchases of property and equipment rose sharply. Amazon attributed the increase primarily to artificial-intelligence investment.

This divergence is central to the quality-of-earnings debate. The Anthropic revaluation increased reported profit without providing current cash. At the same time, Amazon was spending cash on data centers, chips, networking equipment, power, and other infrastructure. A reader focused only on net income would see a company generating more than $60 billion of quarterly profit. A reader focused only on free cash flow would see a business with a trailing-twelve-month outflow. Both figures were correct; they described different dimensions of the company.

Amazon raised its 2026 capital-spending expectation to approximately $220 billion. Management argued that demand remained greater than available capacity and that investments in AI, chips, robotics, and logistics would support future growth. The operating results from AWS provided evidence that spending was producing revenue. The negative free cash flow showed that the payback period was not immediate.

Investors responded more to the acceleration in AWS than to the accounting gain itself. Amazon shares closed 15.3% higher on July 31, their strongest one-day percentage advance in years, and the company’s market value later crossed $3 trillion. The gain in Anthropic contributed to the earnings surprise, but the durability of the stock move depended on expectations for cloud growth, pricing, margins, and future return on capital.

Alphabet: A $99 Billion Gain That Changed the Meaning of Net Income

Alphabet’s second-quarter report produced an even larger accounting effect than Amazon’s. Revenue increased 24% to $119.8 billion, operating income rose 30% to $40.8 billion, and the operating margin expanded to 34% from 32%. Those were strong results for a company of Alphabet’s size. Yet net income surged 298% to $112.1 billion, a rate that could not be explained by advertising, cloud, subscriptions, or any other operating segment.

The difference appeared in other income and expense. Alphabet reported $98.0 billion of net other income, compared with $2.7 billion a year earlier. The largest component was a $99.0 billion net gain on equity securities. In its earnings release, the company quantified the effect precisely: the equity-security gain increased income-tax expense by $21.9 billion, net income by $77.1 billion, and diluted earnings per share by $6.26.

That disclosure makes it possible to construct a more informative comparison. Subtracting the after-tax investment effect from reported net income leaves approximately $35.0 billion. Subtracting $6.26 from reported diluted EPS of $9.11 leaves $2.85. Those figures are not company-defined non-GAAP measures, and they should not replace the official results. They are simple reconciliations using Alphabet’s own disclosed after-tax effect.

On that basis, net income excluding the disclosed equity gain would have increased about 24% from the $28.2 billion reported a year earlier, while diluted EPS would have risen roughly 23% from $2.31. That performance closely matched the growth in the operating business and was far more representative of the quarter’s recurring economics than the 298% increase in reported net income.

Alphabet’s Form 10-Q showed how the $99.0 billion total was assembled. The company recorded $77.4 billion of unrealized net gains on non-marketable equity securities under the measurement alternative, $21.4 billion of unrealized gains on marketable and other equity securities, and $278 million of realized gains on securities sold. The filing said the quarter’s gains were primarily related to SpaceX and another private company.

SpaceX had completed its public offering in June, creating a market price for a holding that had previously been private. Alphabet disclosed that its marketable securities included tens of billions of dollars of SpaceX shares subject to sale restrictions. The transition from a private valuation to a quoted public price crystallized a massive accounting change even though Alphabet had not sold most of the position.

The investment was economically consequential. Alphabet had supported SpaceX years before the rocket and satellite company reached public markets. A successful early venture investment can legitimately become part of the parent company’s value. The issue was not whether the gain belonged to shareholders; it was whether the gain said anything about the quarter-to-quarter performance of Google Search, YouTube, Google Cloud, Android, or the company’s subscription businesses.

Alphabet’s Operating Results Were Strong on Their Own

Google Services revenue increased to $94.5 billion from $82.5 billion. Google Search and other advertising revenue rose to $63.3 billion, YouTube advertising reached $11.1 billion, and Google Cloud revenue accelerated 82% to $24.8 billion. Cloud operating income increased to $8.8 billion from $2.8 billion, demonstrating that a business once criticized for heavy losses had become an important source of profit.

Alphabet’s total operating income of $40.8 billion exceeded the prior-year figure by $9.5 billion. That increase came from the company’s actual businesses rather than its investment portfolio. Search remained resilient despite competitive pressure from generative AI. YouTube advertising and subscriptions expanded. Cloud benefited from demand for AI infrastructure, enterprise models, and core computing services.

Those gains matter because they answer the central strategic question facing Alphabet: can the company convert AI investment into revenue and defend the economics of search? The quarter provided positive evidence. It did not settle the issue, but an 82% increase in cloud revenue and 17% growth in Search and other revenue indicated that Alphabet’s operating engines were not dependent on the securities windfall.

The quality of the quarter looked less comfortable in cash flow. Alphabet generated $39.1 billion of operating cash flow but spent $44.9 billion on property and equipment, producing negative free cash flow of $5.9 billion. The company also raised its expected 2026 capital expenditure to a range of $195 billion to $205 billion. Those numbers showed the same tension visible at Amazon: accounting profits from strategic holdings were rising while cash demands for AI infrastructure were increasing even faster.

Alphabet’s shares initially fell after the report as investors focused on spending and free cash flow rather than the extraordinary GAAP EPS figure. That reaction was rational. A one-time or volatile valuation gain does not necessarily support a higher multiple on the operating business, while recurring capital expenditure can reduce the cash available for buybacks, acquisitions, or other uses.

The market later recovered alongside a broader technology rally, but the initial response illustrated that sophisticated investors were not simply treating $9.11 of EPS as equivalent to recurring earnings. The more relevant questions concerned cloud growth, AI monetization, infrastructure efficiency, advertising demand, and the durability of margins.

Microsoft: Smaller OpenAI Effects, but More Complicated Adjustments

Microsoft’s quarter was less dramatically distorted, but it demonstrated why a single “adjusted EPS” label may not remove all investment-related effects. For the fiscal fourth quarter ended June 30, 2026, Microsoft reported revenue of $90.0 billion, operating income of $40.6 billion, net income of $35.8 billion, and diluted EPS of $4.81. Revenue and operating income each grew 18%, while GAAP net income increased 31%.

Microsoft’s earnings release presented a reconciliation that excluded the impact of OpenAI investments. The OpenAI effect increased fourth-quarter net income by $480 million and EPS by $0.07. Excluding it, Microsoft reported adjusted net income of $35.3 billion, up 22%, and adjusted EPS of $4.74, up 23%.

The year-over-year comparison was affected by both sides of the investment cycle. In the prior-year quarter, OpenAI-related losses reduced net income by $1.6 billion and EPS by $0.21. The swing from a $1.6 billion loss to a $480 million gain improved GAAP growth by more than $2 billion. For the full fiscal year, OpenAI gains increased net income by $5.0 billion, while the prior year included $3.6 billion of losses. The year-over-year swing was approximately $8.6 billion.

Microsoft’s OpenAI-adjusted figures were useful because they made that effect visible. Yet they did not remove every venture-related gain. On the earnings call, Chief Financial Officer Amy Hood said the quarter included a $3.2 billion gain from Microsoft’s investment in Anthropic. She grouped that gain with other discrete items that collectively benefited diluted EPS by $0.27 relative to the company’s prior guidance.

Microsoft also said that, after adjusting for OpenAI, other income and expense was $2.8 billion, driven by the Anthropic gain. In other words, the company’s headline non-GAAP EPS excluded OpenAI but retained Anthropic. The result was not misleading—the reconciliation was clearly labeled—but it was not a pure operating measure either.

This illustrates a broader problem with company-specific adjustments. Management may exclude one investment because it has historically created volatility or because the company has established a formal reporting convention, while another economically similar gain remains in adjusted results. Comparisons across companies can therefore become inconsistent even when each company follows disclosure rules.

Microsoft’s Operations Were the Main Reason the Stock Rose

The investment effects were not the principal reason investors rewarded Microsoft. Microsoft Cloud revenue rose 27% to $59.3 billion, Azure and other cloud services grew 43%, and commercial remaining performance obligations increased 84% to $678 billion. Operating income increased to $40.6 billion, and operating cash flow rose 30% to $55.4 billion.

Microsoft’s adjusted operating performance exceeded expectations, and management forecast Azure growth of approximately 45% in constant currency for the following quarter. The company also said customer demand continued to exceed available capacity. Those disclosures supported the view that AI infrastructure was being monetized, not merely constructed.

Microsoft shares rose more than 15% after the report, adding hundreds of billions of dollars to the company’s market value. The size of the move reflected confidence in Azure growth, cloud demand, and future margins rather than the comparatively modest OpenAI gain. The quarter therefore provided a useful contrast with Amazon and Alphabet: investment-accounting effects mattered, but the market’s valuation decision still centered on recurring operations and guidance.

Capital spending remained substantial. Microsoft reported $41 billion of capital expenditures in the quarter, including finance leases, with roughly two-thirds directed to shorter-lived assets such as CPUs and GPUs. Cash paid for property and equipment was $35.8 billion, and free cash flow was $19.6 billion. Unlike Amazon and Alphabet, Microsoft remained strongly free-cash-flow positive during the quarter, though its infrastructure requirements continued to rise.

Microsoft’s disclosures also showed how strategic investing, customer contracts, and platform economics can overlap. OpenAI commitments influenced commercial bookings and remaining performance obligations. Microsoft therefore reported growth rates both including and excluding OpenAI in some sales metrics. This was a reasonable attempt to help investors distinguish broad customer demand from one exceptionally large partner, but it reinforced the complexity of the relationship.

Are the Profits “Fake”? No—but They Are Different

Calling the gains fake would be inaccurate. Amazon, Alphabet, and Microsoft owned financial interests that became more valuable. The companies followed accounting standards that require those changes to be recognized. Their auditors and financial-reporting systems treated the gains as part of GAAP income. Shareholders ultimately benefit if the investments can be sold or converted into valuable public shares.

The more accurate description is that the gains were non-operating, largely unrealized, and potentially volatile. Each word matters.

Non-operating means the gain did not arise from the company’s main revenue-generating activities. Amazon sells goods, advertising, subscriptions, and cloud services. Alphabet sells advertising, cloud services, subscriptions, devices, and other products. Microsoft sells software, cloud capacity, gaming products, and related services. Portfolio revaluations sit below operating income because they are not generated by those businesses.

Unrealized means the company had not sold the security and converted the gain into cash. An unrealized gain can later become realized at a higher or lower amount. It still represents economic value at the measurement date, but it is less certain than cash received from a completed sale.

Volatile means the value can move sharply between reporting periods. Public shares change every trading day. Private-company valuations can jump when a new financing round establishes a higher price, then remain stale until another observable transaction occurs. An IPO may produce a sudden remeasurement, followed by quarterly gains or losses as the public stock trades.

Investors should also avoid assuming that operating earnings are automatically “real” while investment gains are not. Operating income contains estimates, accruals, depreciation schedules, stock-based compensation, warranty assumptions, reserves, and revenue-recognition judgments. Cash flow can be affected by working-capital timing and supplier terms. Every financial measure has limitations.

The goal is not to identify one perfect number. It is to understand the source, repeatability, cash characteristics, and risk of each component. A recurring cloud subscription has different economics from an unrealized venture gain. Both can create shareholder value, but they deserve different valuation treatment.

How Investors Can Read Through the Distortion

A disciplined analysis begins with the income statement and then moves through the cash-flow statement, balance sheet, segment disclosures, and footnotes. The order matters because headline EPS is often the most compressed and least explanatory figure.

1. Start With Revenue

Revenue shows the amount earned from customers before most expenses. It is not profit, but it helps determine whether demand is expanding. Amazon’s 20% sales growth, Alphabet’s 24% revenue growth, and Microsoft’s 18% revenue growth indicated broad operating momentum. None of those figures included the investment gains.

Revenue quality still requires scrutiny. Cloud contracts may be long-term and usage-based. Advertising can be cyclical. Retail revenue may carry thin margins. Large commitments from strategic partners may not translate into immediate recognized revenue. Yet revenue remains a cleaner first step than net income when portfolio gains are unusually large.

2. Examine Operating Income and Segment Profit

Operating income separates the main business from interest, taxes, and most investment gains. Amazon’s operating income rose 43%, Alphabet’s 30%, and Microsoft’s 18%. Those figures confirmed that core profitability improved.

Segment results reveal where the improvement occurred. AWS produced most of Amazon’s operating-profit increase. Google Cloud expanded profit rapidly while Search remained the largest earnings engine. Microsoft’s Intelligent Cloud segment benefited from Azure growth. A company can report strong consolidated operating income while one important segment weakens, so the segment tables matter.

3. Reconcile Other Income

“Other income and expense” can contain interest income, interest expense, foreign-exchange effects, equity gains, debt-security changes, impairments, and miscellaneous items. In ordinary quarters, the line may be small relative to operations. In this earnings season, it was the main driver of net income for Amazon and Alphabet.

Readers should identify whether the gain was realized or unrealized, marketable or private, recurring or unusual, taxable or tax-deferred, and supported by a quoted price or a valuation model. The answers determine how much confidence to place in the amount and how likely it is to recur.

4. Compare Net Income With Cash Flow

Net income includes noncash items, while cash flow shows money moving through the business. The comparison was particularly revealing in 2026. Alphabet reported $112.1 billion of net income but generated $39.1 billion of quarterly operating cash flow and negative $5.9 billion of free cash flow after capital spending. Amazon reported $62.6 billion of net income while trailing-twelve-month free cash flow was negative $7.6 billion.

That does not imply a liquidity crisis. Both companies had enormous financial resources and access to capital markets. It does show that their record accounting profits did not translate into equally extraordinary cash generation.

5. Track Capital Expenditure and Asset Lives

AI infrastructure is capital intensive. Companies purchase or lease chips, servers, networking equipment, land, buildings, power capacity, and cooling systems. Those outlays appear in investing cash flow and are depreciated over time rather than expensed immediately.

Investors should monitor both the amount spent and the useful lives assigned to assets. Longer assumed lives reduce annual depreciation expense and increase current operating profit, while shorter lives accelerate expense. Rapid technological change raises the risk that some AI equipment becomes economically obsolete before the accounting life ends.

Microsoft disclosed that it was extending the estimated useful lives of data centers and office buildings from 15 to 25 years at the start of fiscal 2027. The company said the operating-income benefit would be minimal, while the larger effect would be a shift in lease classification. The disclosure was important because depreciation assumptions can influence comparisons across time and across companies.

6. Distinguish Company-Defined Adjustments

Microsoft’s OpenAI-adjusted results demonstrated both the value and the limitation of non-GAAP measures. The adjustment helped readers understand the OpenAI effect, but the Anthropic gain remained. Amazon historically reports GAAP EPS without a company-defined adjusted EPS figure, so consensus estimates often use GAAP. Alphabet quantified the after-tax gain but did not present a formal adjusted EPS number excluding it.

Analysts therefore need to create consistent reconciliations. A comparison that removes OpenAI from Microsoft but leaves Anthropic in Microsoft, removes nothing from Amazon, and treats Alphabet’s full GAAP EPS as recurring would not be economically coherent.

7. Value the Operating Business and Investments Separately

A sum-of-the-parts framework can be more informative when investment portfolios become large. The operating business can be valued using revenue, operating profit, cash flow, growth, and capital intensity. Strategic holdings can be valued separately, with discounts for taxes, liquidity restrictions, uncertainty, and the possibility of future losses.

This approach avoids two errors: ignoring valuable investments entirely and capitalizing volatile gains as though they were recurring operating earnings. It also makes clear that a company’s stock can serve partly as a proxy for private or newly public assets that investors cannot easily access directly.

The Strategic Logic Behind Big Tech’s AI Stakes

The investments were not conventional venture bets made solely for financial return. Amazon, Alphabet, and Microsoft operate the infrastructure, software, distribution channels, and enterprise relationships that frontier AI companies need. Their equity stakes sit inside larger strategic partnerships that can influence where models are trained, which chips are used, how products are distributed, and which cloud platforms capture the resulting demand.

Amazon’s relationship with Anthropic is the clearest example. Anthropic selected AWS as a primary cloud provider and has committed to using Amazon’s custom Trainium and Inferentia chips. Amazon gains a customer for computing capacity, a showcase for its semiconductors, access to Claude models through Bedrock, and potential appreciation in Anthropic’s equity value. Anthropic receives capital, infrastructure, and enterprise distribution.

Microsoft’s original partnership with OpenAI followed a similar strategic pattern. Azure supplied computing capacity; Microsoft integrated OpenAI models into products; OpenAI received capital and infrastructure; and Microsoft gained both commercial rights and investment exposure. By 2026, Microsoft had broadened its model strategy, adding Anthropic and emphasizing that Azure customers could choose among multiple frontier, open-weight, and in-house models.

Alphabet’s Anthropic relationship gave Google Cloud a major AI customer and a counterweight to OpenAI, while its SpaceX investment originated from a different strategic era. Google’s early support of SpaceX was connected partly to ambitions for global internet access. The eventual public-market value of the stake became far larger than the original investment and, by 2026, materially affected Alphabet’s financial statements.

Strategic investing can create several kinds of return simultaneously:

• Financial return: the equity stake appreciates.
• Cloud revenue: the investee buys computing capacity, storage, networking, and related services.
• Product access: the investor gains rights to distribute or integrate models.
• Customer acquisition: the partnership attracts developers and enterprises to the investor’s platform.
• Technical learning: operating frontier workloads improves infrastructure design and software tooling.
• Competitive positioning: the investment prevents a rival cloud provider from controlling an important model ecosystem.

That strategic logic helps explain why the investments can be worth more to a hyperscaler than to a passive financial investor. Amazon may accept terms that appear expensive if the relationship drives hundreds of billions of dollars of long-term AWS demand. Microsoft may value OpenAI partly through its effect on Azure, GitHub, Microsoft 365, and developer adoption. Alphabet may view Anthropic as both a customer and a source of model diversity.

The same logic complicates financial analysis. Revenue from the investee can boost the investor’s operating results, while financing supplied by the investor helps the investee pay for that revenue. An increase in the investee’s valuation then produces an investment gain for the investor. The cycle can be commercially valid, but outsiders must ask how much end-user demand exists beyond transactions among the participants.

When Strategic Partnerships Become Circular

“Circular financing” is often used loosely. Not every transaction between an investor and an investee is circular, and not every circular element is abusive. A bank can lend to a customer that uses the funds to purchase equipment. A manufacturer can finance a distributor. A cloud provider can invest in a software company that purchases cloud services. These arrangements can support productive investment.

The concern becomes more serious when reported growth depends heavily on capital supplied by the same companies recognizing the revenue or investment gains. Suppose a hyperscaler invests $20 billion in an AI lab, the lab commits much of that capital to the hyperscaler’s cloud services, and a new financing round at a higher valuation allows the hyperscaler to recognize a large paper gain. The hyperscaler has improved revenue, backlog, and net income, while the lab has more capital and a higher valuation. The economic durability ultimately depends on whether external customers pay for AI products at a level sufficient to support the infrastructure commitments.

Amazon’s filing disclosed more than $100 billion of expanded AWS commitments from Anthropic and another $100 billion expansion involving OpenAI. Microsoft’s remaining performance obligations were also influenced by OpenAI commitments, although the company said all sequential growth in commercial RPO during the quarter came from customers outside frontier-model companies. These disclosures were valuable because they allowed investors to distinguish broad enterprise demand from a small number of strategic counterparties.

The strongest bullish interpretation is that frontier labs are creating genuine demand for scarce computing resources. Their contracts are not accounting inventions; they require data centers, electricity, chips, engineering, and operating support. As AI products gain users, the labs may generate enough revenue to finance the commitments independently. Early strategic capital would then have accelerated a real market.

The skeptical interpretation is that the financing ecosystem can temporarily make both sides look stronger than they would under fully independent funding. Valuations rise because strategic investors are willing to provide capital at prices justified partly by cloud demand. Cloud demand rises because the investees have fresh capital. Portfolio gains then improve the investors’ reported earnings, helping support their own market values and capacity to invest more.

The available evidence does not establish that the transactions lack economic substance. It does establish that investors need more disclosure about customer concentration, financing terms, cancellation rights, minimum commitments, pricing, and the extent to which revenue is supported by strategic capital.

Private-Company Valuations Carry Special Uncertainty

Public-market prices are imperfect, but they are observable. Millions of buyers and sellers can react continuously to earnings, competition, regulation, and changes in interest rates. Private-company valuations are established less frequently and often through negotiated financing rounds involving a small group of investors.

A headline valuation from a private funding round does not necessarily mean every share is worth the same amount. New investors may receive liquidation preferences, anti-dilution protection, information rights, board rights, or other advantages. Existing common shares may be worth less. Secondary transactions may occur at discounts. Restrictions on sale can reduce economic value.

Amazon’s Level 3 valuation description acknowledged these issues. The company considered the rights and obligations of its nonvoting preferred stock, the existence of other share classes, expected liquidity events, possible offerings, and discounts for lack of marketability. Those judgments were necessary because an observable funding price could not simply be applied to every instrument.

Level 3 does not mean unreliable or improper. It means significant inputs are not directly observable in active markets. The valuation may be carefully prepared and audited, but it is more sensitive to assumptions than a quoted stock price. Small changes in discounts, conversion probabilities, or expected timing can produce large changes when the underlying position is worth tens of billions of dollars.

Private valuations can also be sticky. If a company raises money at a higher price, investors recognize an upward adjustment. If business conditions deteriorate but no new transaction occurs, the carrying value may not decline immediately unless impairment indicators are identified. The result can be asymmetric timing: large gains appear when financing rounds establish higher prices, while losses require a separate judgment that the investment is impaired.

An IPO changes that pattern. Once shares trade publicly, market prices create frequent remeasurement. Alphabet’s SpaceX stake moved from a difficult-to-observe private holding to a marketable security. That improved price transparency but introduced quarterly volatility. A fall in SpaceX shares can now reduce Alphabet’s net income even if Google’s operations perform exactly as expected.

The Rivian Precedent Shows How Quickly Gains Can Reverse

Amazon has already experienced the downside of this accounting model. When Rivian completed its public offering in 2021, Amazon recognized an $11.8 billion pre-tax valuation gain in the fourth quarter. The gain helped lift annual net income to $33.4 billion.

Rivian shares then fell sharply. In 2022, Amazon recorded a $12.7 billion pre-tax valuation loss on the investment and reported a $2.7 billion annual net loss. Amazon’s 2022 results made clear that the shift from profit to loss was heavily influenced by the Rivian revaluation, not solely by changes in retail or AWS operations.

The Rivian episode is a useful precedent for Anthropic, OpenAI, and SpaceX. A large gain can be economically meaningful and still reverse. The more valuable the position becomes relative to the parent company’s normal quarterly profit, the more volatile reported EPS becomes.

Suppose a public holding worth $100 billion falls 20% in a quarter. Before taxes and other offsets, the parent may recognize a $20 billion loss. That could erase much of the operating profit of even a large technology company. The underlying business might meet expectations, yet headline EPS could appear disastrous.

Private holdings create a different timing pattern but similar risk. A down round, impairment, failed IPO, or deterioration in expected liquidity can force a write-down. Convertible instruments can produce additional complexity when gains move between other comprehensive income and the income statement.

The Rivian comparison also shows why excluding gains but including losses—or excluding losses but including gains—would be inconsistent. Investors should examine both directions. A sensible normalized measure removes volatile revaluations symmetrically while preserving the investment portfolio as a separate component of value.

Why the S&P 500 Is Especially Sensitive to Mega-Cap Accounting

The S&P 500 is often described as a broad market index, but its earnings are increasingly influenced by a small group of enormous companies. Market-cap weighting determines index performance, while aggregate earnings calculations are also affected by the dollar profits of the largest constituents. When one company adds $80 billion of year-over-year net income, the effect can exceed the combined changes at dozens of smaller companies.

This concentration has several consequences.

First, index earnings growth can diverge from the median company’s experience. The typical S&P 500 member may report moderate growth while the aggregate rises sharply because of a few mega-cap firms. Equal-weighted earnings measures, sector medians, and company-count statistics can provide a broader view.

Second, sector growth rates can become difficult to interpret. Alphabet sits in Communication Services, Amazon in Consumer Discretionary, and Microsoft in Information Technology. Their portfolio gains therefore make those sectors appear more profitable even though the gains may be unrelated to communications, retail, or software demand.

Third, valuation ratios can be temporarily distorted. A trailing price-to-earnings ratio falls when reported earnings jump, making the market look cheaper. If the earnings include gains unlikely to recur, the lower multiple may exaggerate affordability. Forward multiples based on analyst estimates can be more useful, but they depend on whether analysts normalize future investment effects consistently.

Fourth, index-linked investors indirectly own the strategic venture portfolios of the largest constituents. A person buying an S&P 500 fund receives exposure not only to Amazon retail and AWS, Alphabet advertising and cloud, or Microsoft software and Azure, but also to their stakes in Anthropic, OpenAI, SpaceX, and other companies.

That exposure can be beneficial. Public-company investors gain access to private-market value creation that would otherwise be unavailable. It can also add risk that is not obvious from a company’s sector label or historical business model.

GAAP Versus Non-GAAP: The Wrong Debate Can Miss the Point

The quarter revived a familiar argument about whether GAAP or non-GAAP earnings provide the “true” picture. The better answer is that neither should be used alone.

GAAP offers standardized rules, audited statements, and comparability. It prevents management from simply excluding every unfavorable item. The investment gains were part of shareholder value and belonged in the official results under the applicable accounting rules.

Non-GAAP measures can isolate operating trends, but they are not standardized. A company may exclude restructuring charges that recur every year, stock-based compensation that is a real cost to shareholders, or acquisition expenses essential to its strategy. Adjusted EPS can therefore become more flattering than informative.

The Securities and Exchange Commission requires companies that present non-GAAP measures to reconcile them with the most directly comparable GAAP figure and prohibits misleading presentation. SEC guidance also emphasizes that non-GAAP measures should not be presented as superior substitutes for GAAP.

In this case, the most useful analysis does not require choosing sides. It requires a bridge:

1. Begin with GAAP net income and EPS.
2. Identify the disclosed investment gain or loss and its tax effect.
3. Calculate a consistent operating-oriented measure.
4. Compare that measure with revenue, operating income, and cash flow.
5. Value the investment portfolio separately rather than pretending it has no value.

Alphabet made the bridge relatively easy by disclosing the after-tax and per-share effect. Microsoft provided an OpenAI reconciliation but required readers to examine the call for the Anthropic gain. Amazon disclosed the pre-tax other income but did not publish a company-defined adjusted EPS. Greater consistency would improve market understanding.

What Better Disclosure Would Look Like

The companies complied with existing reporting requirements, and each provided important information. Still, the scale of the gains suggests that investors would benefit from more standardized detail.

A Reconciliation of Net Income Excluding Major Investment Changes

Companies could present a supplemental schedule showing reported net income, realized gains, unrealized gains, tax effects, and net income excluding those items. The schedule should remove gains and losses symmetrically and clearly state that it is non-GAAP.

Instrument-by-Instrument Exposure

When material, companies could disclose carrying value by investee and security type: common stock, preferred stock, convertible debt, warrants, and other interests. They could also explain which gains flow through net income and which remain in other comprehensive income.

Valuation Sensitivities

For Level 3 positions, a sensitivity table could show how changes in marketability discounts, expected liquidity timing, or conversion assumptions affect fair value. Exact proprietary details may be commercially sensitive, but broad ranges would help investors understand risk.

Commercial Relationships With Investees

Companies could quantify revenue, remaining performance obligations, financing commitments, and customer concentration associated with strategic investees. Microsoft already provided some growth rates excluding OpenAI. Similar disclosure across the industry would make comparisons easier.

Cash Realization and Sale Restrictions

Publicly traded holdings may be subject to lockups or other restrictions. Companies could disclose the portion immediately saleable, the tax cost of disposal, and any strategic reasons they do not intend to sell. A quoted market value is less liquid when the owner cannot or will not monetize it.

More disclosure would not eliminate volatility. It would make the source of volatility easier to model.

The Cash-Flow Counterpoint to Record Earnings

The same earnings season that produced extraordinary paper profits also revealed extraordinary cash requirements. Amazon, Alphabet, and Microsoft were spending at a scale once associated with national infrastructure programs. The investments included semiconductor purchases, data-center construction, power contracts, networking equipment, land, cooling systems, and long-term leases.

Amazon expected approximately $220 billion of capital spending in 2026. Alphabet raised its forecast to between $195 billion and $205 billion. Microsoft said its calendar-year spending expectation, adjusted for a lease-classification change, remained approximately $175 billion. Combined, those figures approached $600 billion for a single year.

The comparison with investment gains is revealing. Amazon’s Anthropic-linked other income and Alphabet’s equity-security gain improved earnings without requiring the companies to sell the assets. The infrastructure programs consumed or committed cash. Reported net income therefore rose faster than the amount of cash available after capital expenditures.

This does not necessarily signal poor capital allocation. Data centers can produce revenue for years. Cloud demand was accelerating, capacity remained constrained, and customers were signing large commitments. The companies had strong balance sheets and access to debt markets. The question was whether future operating cash flows would justify the scale and pace of construction.

Return on invested capital becomes more important as spending rises. Investors need to know not only how quickly cloud revenue grows but also how much incremental capital is required to generate that growth. A business that adds $10 billion of annual operating profit after investing $20 billion has different economics from one that requires $100 billion.

AI infrastructure also faces technological risk. Chips can become obsolete quickly, model architectures can become more efficient, and customers can shift among providers. Power and land retain value, but specialized hardware may have a shorter economic life. If demand disappoints, excess capacity could pressure margins and require impairments.

The bullish case is that the current buildout resembles the early cloud era: upfront capital creates a durable platform with high utilization, recurring revenue, and strong incremental margins. The skeptical case is that competitive spending will produce oversupply, falling prices, and lower returns. The investment-accounting gains do not resolve that debate. They can temporarily make the companies look more profitable while the answer remains uncertain.

What the Market Reaction Really Said

Stock prices did not react mechanically to the reported EPS figures. Alphabet’s shares initially fell after its results despite the enormous earnings beat. Amazon and Microsoft rallied sharply. The differences reflected investors’ judgments about underlying operations, capital efficiency, and forward guidance.

Alphabet’s report contained strong revenue and cloud growth, but investors focused on the higher capital-spending forecast and negative free cash flow. The $99 billion equity gain was already understood as nonrecurring. It did not answer whether the company could maintain margins while spending roughly $200 billion a year.

Amazon’s report offered a more favorable combination: AWS growth accelerated to its fastest rate in 18 quarters, operating income increased substantially, and management described persistent capacity constraints. The Anthropic gain boosted GAAP EPS, but the stock’s 15.3% advance was better explained by renewed confidence in AWS and Amazon’s custom-chip strategy.

Microsoft’s stock rose more than 15% after Azure growth exceeded expectations and management forecast further acceleration. The OpenAI gain was modest relative to operating income, and the Anthropic gain was a disclosed discrete item. Investors treated cloud demand and monetization as the main story.

By August 3, Amazon’s market value had crossed $3 trillion. The broader S&P 500 closed at 7,600.50, up 1.5% for the day and near a record, while the Nasdaq Composite gained 2.1%. The rally also benefited from lower oil prices and easing geopolitical concerns, so it would be inaccurate to attribute the market move solely to AI earnings. The Associated Press market recap showed that macroeconomic news contributed materially.

The market’s behavior offered an important lesson: professional investors were willing to look through portfolio gains when valuing the operating companies. The index-level earnings statistic remained distorted, but individual stock reactions were more discriminating.

Implications for Valuation

Valuation models commonly use earnings, free cash flow, or operating profit. Each produces different conclusions when non-operating gains are unusually large.

Trailing Price-to-Earnings Ratios

A trailing P/E ratio divides the stock price by the most recent twelve months of earnings. When a company records a massive one-time gain, the denominator rises and the ratio falls. The stock can appear cheaper even though recurring earning power has changed much less.

Alphabet’s quarterly EPS included $6.26 from the equity gain. Annualizing the full $9.11 quarter would be inappropriate, and even including the gain once in trailing earnings can depress the P/E for several quarters. Amazon’s $5.75 EPS created a similar effect.

Forward Earnings

Forward P/E ratios depend on analyst forecasts. Analysts may exclude future gains if they cannot be predicted, producing a more normalized denominator. Yet comparisons become difficult if actual GAAP results include gains while estimates exclude them. That mismatch explains why Alphabet and Amazon posted enormous “beats.”

FactSet noted that most analysts contributing Amazon estimates used GAAP EPS because Amazon historically reports only GAAP EPS. The estimates did not anticipate the full Anthropic revaluation, but the actual figure included it. The resulting surprise measured both operating performance and an unforeseen valuation event.

Enterprise Value and Operating Profit

Enterprise-value-to-operating-income or enterprise-value-to-EBIT measures can reduce the effect because investment gains sit below operating income. They must still account for the value of investments held on the balance sheet. Subtracting cash but ignoring a $100 billion strategic stake would overstate the effective price paid for the operating business.

Sum-of-the-Parts Valuation

A sum-of-the-parts approach can separate the operating company from investments. For Alphabet, an analyst might value Google Services, Google Cloud, Other Bets, and the SpaceX or private-company holdings independently. For Amazon, the framework might separate retail, AWS, advertising, logistics assets, and strategic AI investments.

Discounts are necessary. Taxes may be due when gains are realized. Lockups and marketability restrictions reduce liquidity. A strategic owner may be reluctant to sell. Private valuations can change. Applying the full quoted or financing-round value without adjustment can overstate realizable value.

Cash-Flow Valuation

Discounted cash-flow models avoid treating unrealized gains as recurring operating cash. They focus on future cash generated by the businesses. Strategic investments can be added separately at estimated net value. The method is conceptually cleaner but highly sensitive to assumptions about long-term margins, capital spending, growth, and discount rates.

The most defensible conclusion is not that Big Tech became suddenly cheap because earnings exploded. It is that the companies owned valuable assets while their operating businesses also improved. The two sources of value should be modeled separately.

Implications for Passive Investors and Retirement Accounts

Millions of Americans own the S&P 500 through retirement plans, index funds, insurance products, and brokerage accounts. They may not think of themselves as venture-capital investors, yet the index’s largest companies increasingly provide indirect exposure to frontier AI labs and other high-growth private businesses.

That exposure can diversify sources of return. A successful investment in SpaceX or Anthropic may add value without requiring an individual saver to qualify for a private offering. It can also create hidden concentration. The same AI ecosystem influences the operating revenue, capital spending, investment portfolios, and market valuations of multiple index leaders.

If AI demand continues to expand, the relationships can reinforce one another. Frontier labs purchase more cloud capacity, hyperscalers grow revenue, investment stakes appreciate, and rising market values increase the index. If demand weakens or valuations contract, the reversal can also be correlated. Cloud growth slows, utilization falls, portfolio values decline, and several large index constituents report weaker earnings at the same time.

Passive investors do not need to react to every quarterly accounting change. Long-term index strategies are designed to absorb company-specific volatility. They should understand, however, that broad-market exposure is not evenly distributed across independent industries. A large share of index value and earnings is linked to a connected technology and AI investment cycle.

Regulatory and Policy Questions

The accounting treatment itself is governed by established standards, and there was no indication that the companies violated them. The policy debate concerns whether disclosure keeps pace with the scale and complexity of strategic cross-investments.

Regulators may focus on several areas:

• Non-GAAP consistency: whether companies exclude gains and losses on a balanced basis and avoid presenting adjusted measures more prominently than GAAP.
• Fair-value controls: whether Level 3 valuations use reasonable assumptions and appropriately distinguish different securities.
• Related commercial arrangements: whether revenue and backlog tied to investees are disclosed clearly enough to assess concentration and collectability.
• Competition: whether cloud-and-model partnerships limit customer choice or reinforce dominant platforms.
• Systemic concentration: whether interconnected financing and infrastructure commitments create risks that spread across major index constituents.

More prescriptive accounting is not necessarily required. The existing rules captured the economic value changes. A targeted disclosure framework may be more effective than excluding the gains from GAAP. Investors need both the official number and enough information to normalize it.

Strongest Bullish Interpretation

The bullish reading begins with the operating results. AWS grew 37%, Google Cloud 82%, Azure 43%, Amazon operating income 43%, Alphabet operating income 30%, and Microsoft operating income 18%. These were not companies relying solely on financial engineering. Their core businesses expanded at rates that would be impressive for much smaller firms.

The investment gains may be evidence of strategic foresight rather than distortion alone. Alphabet’s early SpaceX position created tens of billions of dollars of value. Amazon’s Anthropic partnership positioned AWS at the center of frontier-model development. Microsoft’s OpenAI relationship accelerated the adoption of generative AI across its product portfolio.

Strategic stakes can lower customer-acquisition costs, secure anchor tenants for data centers, improve product access, and create optionality. The parent companies have the balance sheets to fund long-term projects that smaller competitors cannot match. If AI becomes a foundational layer of computing, the investments may produce both operating and financial returns for years.

Under this interpretation, the 2026 gains were unusual in size but not meaningless. They represented the market’s recognition that AI labs and space infrastructure had become extremely valuable. Investors who stripped the gains out and ignored the holdings entirely would understate shareholder wealth.

Strongest Skeptical Interpretation

The skeptical case begins with cash. The companies were committing nearly $600 billion of annual capital expenditure while Amazon and Alphabet reported negative free cash flow on important measurement periods. Some of the customers supporting infrastructure demand were also recipients of strategic investment and financing from the same hyperscalers.

Private valuations can be fragile. Financing rounds may include investor protections that make headline values difficult to apply to older securities. Level 3 models require assumptions. Public-market prices can fall quickly after lockups expire or expectations change. The Rivian precedent demonstrates that large gains can become large losses.

The ecosystem may also contain more concentration than headline cloud growth suggests. A small number of frontier labs account for enormous commitments. If one lab loses market share, faces regulation, cannot raise capital, or develops more efficient models requiring less compute, projected demand could fall.

Finally, rising valuations can create a feedback loop in reported earnings. Investment gains make parent companies appear more profitable, but they do not directly fund the capital expenditure unless assets are sold. Investors may award higher market values based partly on gains generated by the same ecosystem that depends on continued financing.

Under this interpretation, the operating businesses were strong, but the market should not capitalize the investment gains as recurring earnings or assume every announced commitment will produce attractive returns.

What Would Change the Assessment

Several developments would strengthen the bullish case:

• Continued cloud growth from customers outside the frontier-model companies.
• Improving free cash flow despite high capital expenditure.
• Evidence that AI products generate durable enterprise and consumer revenue.
• High utilization of new data centers and stable or improving cloud margins.
• Successful public offerings or secondary sales that convert paper gains into cash near reported values.
• Greater disclosure showing that strategic customer commitments are diversified and enforceable.

Other developments would strengthen the skeptical case:

• Down rounds, impairments, or sharp declines in newly public investees.
• Repeated free-cash-flow deficits without corresponding acceleration in operating profit.
• Contract renegotiations, cancellations, or slower usage by frontier labs.
• Falling cloud prices caused by excess capacity.
• Longer depreciation lives or other accounting changes that materially boost profit without stronger cash generation.
• Regulatory restrictions that weaken exclusive or preferential cloud relationships.

The next several quarters will matter more than the single record-setting period. Once the initial revaluations enter the comparative base, year-over-year earnings growth can slow sharply even if the businesses remain healthy. Conversely, another rise in the investments could create further gains.

Taxes, Lockups, and the Difference Between Reported Value and Spendable Cash

A large unrealized gain increases accounting net income, but the parent company cannot necessarily spend the reported amount. Three deductions often stand between a carrying value and cash available to shareholders: taxes, liquidity restrictions, and strategic constraints.

Alphabet’s disclosure made the tax effect visible. The $99.0 billion equity-security gain increased its tax provision by $21.9 billion, leaving a $77.1 billion benefit to net income. The tax expense did not necessarily mean Alphabet paid the entire amount in cash during the quarter. Deferred-tax accounting recognizes obligations associated with gains before a sale occurs. Timing depends on the security, jurisdiction, tax basis, and eventual disposition.

Amazon’s first-half filing similarly showed a large tax consequence. Its income-tax provision included $15.9 billion of net discrete expense primarily attributable to upward adjustments in Anthropic investments. The figure demonstrates why a pre-tax valuation gain should not be added directly to net income or market value. A portion belongs economically to tax authorities if and when the gain is realized, subject to future changes in law and transaction structure.

Liquidity restrictions create a second discount. Newly public shares may be subject to lockups, contractual sale limits, securities-law restrictions, or negotiated standstill provisions. Even when a market price exists, a strategic holder may be unable to sell at the quoted price. A block worth tens of billions of dollars can also be too large to liquidate quickly without affecting the market.

Alphabet’s SpaceX position illustrates the issue. Public trading created a transparent reference price, but substantial portions of the holding were restricted. The accounting value reflected the applicable fair-value framework, while the practical proceeds from an immediate sale could have been lower after restrictions, market impact, and tax.

Strategic constraints are less formal but equally important. Amazon may not want to sell Anthropic securities if ownership supports its cloud partnership. Microsoft may view its OpenAI position as central to product access and competitive strategy. Alphabet may retain SpaceX shares because the relationship has long-term commercial or technological value. An asset can be valuable without functioning like excess cash.

For valuation purposes, analysts often apply a holding-company discount to strategic stakes. The discount can reflect tax leakage, illiquidity, governance, uncertainty, and the possibility that management reinvests proceeds rather than returning them to shareholders. There is no universally correct rate. The important point is that $100 billion of carrying value is not automatically equivalent to $100 billion of immediately distributable cash.

Why Analyst Consensus Missed the Gains

The enormous earnings surprises did not necessarily mean analysts had misunderstood the operating businesses. Consensus models are designed to forecast recurring revenue, expenses, margins, interest, taxes, and share counts. They are less suited to predicting private-company financing rounds, IPO pricing, or quarter-end movements in strategic holdings.

Amazon’s consensus EPS estimate of $1.82 was compared with GAAP EPS of $5.75. The estimate was broadly consistent with an operating model that did not anticipate a $53.4 billion non-operating gain. When the actual figure arrived, data systems recorded a conventional beat even though most of the difference came from an item outside the normal forecasting process.

Alphabet’s $9.11 EPS versus a $2.88 estimate created the same mismatch. The analyst figure roughly represented expected operating earnings plus ordinary non-operating items. The actual result included $6.26 per share from the equity gain. Comparing the two without adjustment overstated the degree to which Alphabet’s business exceeded expectations.

This is not merely a semantic concern. Earnings-surprise data influence quantitative strategies, factor models, media headlines, and performance evaluation. A system that treats an unforeseen securities revaluation as equivalent to a gross-margin improvement can send a noisy signal.

Analysts can reduce the problem by publishing two estimates: GAAP EPS including an explicit assumption for investment gains, and normalized EPS excluding them. The first preserves comparability with reported results. The second measures operating execution. When the gain cannot be forecast reliably, analysts can label the GAAP estimate as subject to material portfolio volatility rather than pretending to offer false precision.

Data providers can also separate surprise categories. A company may beat operating-income expectations, miss free-cash-flow expectations, and report an unforecast investment gain. Presenting those outcomes independently would be more informative than a single EPS surprise percentage.

The quarter also highlighted a base-effect problem. When the current period contains a massive gain, next year’s comparison becomes unusually difficult. If Alphabet’s operating EPS rises in the second quarter of 2027 but no comparable equity gain occurs, reported EPS could fall sharply. An automated growth screen might classify that as deterioration even though the core business improved.

Analyst models therefore need a normalized historical series. Removing unusual gains and losses from both the current and comparison periods allows a cleaner estimate of recurring growth. The investment portfolio should remain in valuation models, but not in the operating growth rate.

A Scenario Framework for Future Earnings Volatility

The size of the holdings makes scenario analysis more useful than a single-point forecast. Consider three broad outcomes for the next several quarters.

Scenario One: Stable Portfolio Values, Strong Operating Growth

In this case, Anthropic, OpenAI, and SpaceX valuations remain broadly stable while AWS, Google Cloud, and Azure continue expanding. Reported net income would fall back toward operating earnings because the quarter would lack another large revaluation. Year-over-year GAAP growth could look weak or negative, but cash flow and operating income would improve.

This would be a healthy outcome despite an unfavorable headline comparison. It would confirm that the businesses can grow without repeated paper gains. Investors would need to ignore the optical decline in EPS and focus on normalized measures.

Scenario Two: Further Valuation Gains and Continued Cloud Demand

A new Anthropic or OpenAI financing at a higher valuation, or a rise in SpaceX shares, could produce additional gains. If operating businesses also remain strong, reported earnings could stay elevated. The companies would benefit from both sides of the AI cycle.

The risk in this scenario is valuation complacency. Repeated gains may lead investors to treat them as a normal earnings stream even though they depend on market prices and financing events. Capitalizing them at a high multiple would assume continued appreciation indefinitely.

Scenario Three: Portfolio Losses During an Infrastructure Slowdown

The most adverse scenario combines declining investee valuations with slower cloud demand. A public-stock decline or private down round would reduce other income. Lower utilization of newly built capacity would pressure operating margins and free cash flow. The same ecosystem that amplified earnings on the way up would amplify weakness.

This is the scenario in which concentration matters most. The parent companies have diversified businesses and substantial liquidity, so a valuation loss alone would not threaten solvency. It could nevertheless produce large headline losses, weaker investor confidence, and pressure to slow spending.

Scenario Four: Lower Compute Costs Expand Demand

A fourth outcome is more nuanced. Models may become much more efficient, reducing the computing cost of each task. At first glance, that could threaten infrastructure demand. Lower costs may instead expand usage enough to increase total consumption, similar to the way cheaper computing and bandwidth supported new applications.

Under this scenario, frontier-lab valuations could remain high even as unit prices fall. Hyperscalers would compete on efficiency, custom chips, software, and distribution. Returns would depend on whether volume growth exceeds price declines.

No scenario can be assigned certainty. The framework shows why investment gains and capital expenditure should be analyzed together. Both are claims about the future scale and profitability of AI, but they enter financial statements through different channels and on different timelines.

What the Quarter Means for Corporate Strategy

The results may encourage more public companies to make strategic investments in private technology firms. A successful stake can deliver financial gains, secure commercial relationships, and signal access to important innovation. The model is attractive to companies with large balance sheets and infrastructure platforms.

It also creates governance responsibilities. Boards need to evaluate whether the investment serves shareholders or primarily protects management’s strategic ambitions. They should consider concentration limits, conflicts between the parent and investee, information barriers, valuation controls, and exit planning.

A strategic investment should not be justified solely by the possibility of an accounting gain. The gain depends on future investors assigning a higher value. The stronger rationale is a measurable improvement in the parent’s competitive position, customer economics, technology access, or return on capital.

Management compensation can complicate incentives. If executive bonuses depend on GAAP net income or EPS, an unrealized gain may increase compensation even though no cash was realized. Compensation committees can address this by using operating and cash-flow measures or by excluding symmetric portfolio changes from performance targets.

Boards should also examine downside liquidity. A parent may be committed to additional financing precisely when the investee’s prospects weaken. Commercial agreements can create obligations beyond the carrying value of the equity. Maximum exposure therefore includes future funding, guarantees, capacity commitments, and reputational costs—not just the amount on the balance sheet.

A Practical Earnings Checklist for Readers

When a company reports a large earnings surprise, the following questions can reveal whether the change reflects operations, accounting, or both:

1. Did revenue grow at a similar rate to net income? A wide gap may indicate margin change, taxes, asset sales, or investment gains.
2. How did operating income change? This removes most financing and portfolio effects.
3. What appears in other income and expense? Read the table and footnotes rather than relying on the press-release headline.
4. Was the gain realized? A completed sale has different cash characteristics from a mark-to-market adjustment.
5. What was the tax effect? Pre-tax gains should not be compared directly with after-tax profit.
6. How did operating cash flow and free cash flow compare with net income? Large divergences require explanation.
7. Did capital expenditure rise? Growth may require more cash than the income statement suggests.
8. Are adjusted measures consistent? Check whether gains and losses are removed symmetrically and whether similar investments receive similar treatment.
9. What percentage of demand comes from strategic investees? Customer concentration can affect the durability of growth.
10. Can the investment be sold? Consider lockups, liquidity, taxes, and strategic restrictions.
11. What happens to next year’s comparison? A one-time gain can create a difficult base and misleading future growth rates.
12. Does the valuation model separate operations from investments? Avoid multiplying a volatile gain by a recurring-earnings multiple.

Applied to the 2026 quarter, the checklist produces a balanced answer. Revenue and operating income were strong. Other income was unusually large. Much of the gain was unrealized. Free cash flow was weaker than net income at Amazon and Alphabet. Capital expenditure was rising. Strategic customer relationships were material. The investments were valuable but not equivalent to recurring operating profit.

What Happens Next

The immediate focus will shift from the size of the second-quarter gains to their persistence and reversal risk. Public holdings such as SpaceX will be remeasured each quarter. Private holdings such as Anthropic and OpenAI may change when new financing transactions, conversions, impairments, or liquidity events occur.

Amazon’s future results will show whether AWS can sustain growth near the second quarter’s pace while capital spending rises. Investors will also watch whether the company’s Anthropic and OpenAI commitments translate into recognized revenue, cash collections, and acceptable returns on infrastructure.

Alphabet will face scrutiny over free cash flow and the effect of SpaceX price movements. Because the second-quarter equity gain was so large, quarterly net income may remain volatile even if Google’s operations are stable. Search growth, cloud profitability, and capital efficiency will provide a more useful operating scorecard.

Microsoft will continue reporting the impact of OpenAI investments and may face pressure to broaden its adjustments if Anthropic gains or losses become material. Azure capacity, commercial obligations outside frontier labs, and free cash flow will determine whether the infrastructure buildout is producing attractive economics.

At the index level, analysts will need to normalize the second-quarter base when forecasting 2027. A year-over-year comparison against a quarter containing more than $150 billion of unusual gains will be difficult. Headline growth may turn negative even if operating earnings continue to rise. That would be the mirror image of the current distortion.

Frequently Asked Questions

Why did S&P 500 earnings growth jump to 47.4%?

The blended growth rate rose sharply after Alphabet and Amazon reported unusually large GAAP earnings that included investment-related gains. FactSet said excluding the two companies would reduce the rate to 28.8%.

Were Amazon’s Anthropic gains cash profits?

Mostly no. Amazon’s $53.4 billion of non-operating pre-tax other income primarily reflected valuation changes and accounting recognition related to Anthropic investments. The company did not receive an equivalent amount of cash from selling the stake during the quarter.

How much did Alphabet’s investment gain add to earnings?

Alphabet said the $99.0 billion equity-security gain increased net income by $77.1 billion after taxes and diluted EPS by $6.26.

What was Alphabet’s EPS without the disclosed investment effect?

Subtracting the disclosed $6.26 effect from reported EPS of $9.11 produces approximately $2.85. This is a calculated reconciliation, not a company-reported non-GAAP figure.

Did Microsoft’s adjusted EPS remove all AI investment gains?

No. Microsoft’s formal adjustment removed the impact of OpenAI investments. The company separately disclosed a $3.2 billion Anthropic gain that remained an important component of other income.

Are unrealized investment gains allowed under GAAP?

Yes. U.S. accounting standards generally require changes in the value of equity securities to be recognized in net income. Private securities without readily determinable fair values may be adjusted when observable transactions provide evidence of a new value.

Why not ignore the gains completely?

The investments belong to shareholders and can create real economic value. Ignoring them would understate the company’s assets. The better approach is to separate portfolio value from recurring operating earnings.

Which financial measure best shows the core business?

No single measure is sufficient. Revenue, operating income, segment profit, operating cash flow, free cash flow, and capital expenditure together provide a clearer view than headline net income alone.

Could the gains reverse?

Yes. Public holdings can fall in market value, and private holdings can be written down after down rounds, impairments, or weaker liquidity expectations. Amazon’s earlier Rivian gains reversed into large losses in 2022.

Does this mean the S&P 500 earnings season was weak?

No. FactSet estimated 28.8% earnings growth even after excluding Alphabet and Amazon. The season was strong; the investment gains made it look substantially stronger.

Why did Amazon and Microsoft stocks rise while Alphabet initially fell?

Investors focused on operating outlooks. Amazon and Microsoft reported accelerating cloud growth and strong guidance, while Alphabet’s higher capital-spending forecast and negative free cash flow raised concerns despite strong revenue growth.

What should investors watch in future quarters?

Key indicators include cloud growth excluding frontier-model customers, capital expenditure, free cash flow, data-center utilization, changes in investment carrying values, and the conversion of paper gains into cash.

Final Assessment

The second-quarter earnings season revealed both the strength and the complexity of the AI economy. Amazon, Alphabet, and Microsoft produced genuine operating growth. Their cloud platforms expanded, advertising remained resilient, and operating income increased. The quarter would have been strong even without the investment gains.

Yet the 47.4% S&P 500 earnings-growth headline was not a clean measure of that operating strength. Alphabet and Amazon alone lifted the index rate by 18.6 percentage points. Their gains were valid under GAAP and economically relevant, but they were largely non-operating, unrealized, and difficult to repeat.

The strongest evidence supporting the bullish interpretation is the operating data: rapid cloud growth, higher segment profit, large customer commitments, and continued demand beyond available capacity. The strongest concern is the mismatch between record net income and heavy cash spending, combined with strategic financing relationships that blur the line between independent demand and ecosystem-supported growth.

The right conclusion is neither that Big Tech’s earnings were fake nor that the headline numbers should be accepted without adjustment. The companies created value in two distinct ways: by running profitable operating businesses and by owning stakes that appreciated. Investors should recognize both, value them separately, and avoid treating a portfolio revaluation as though it were recurring profit from customers.

The next test will be cash realization. If AI products generate enough external revenue to support the infrastructure commitments, the current investment gains may prove to be an early accounting reflection of durable economic value. If valuations fall or demand depends too heavily on continued strategic financing, future quarters will expose the downside with equal force.

Sources

• FactSet: Amazon Drives S&P 500 Earnings Growth for Q2 Above 45% on Valuation Gains
• FactSet: S&P 500 Earnings Season Update, July 31, 2026
• Amazon: Second-Quarter 2026 Earnings Release
• Amazon Form 10-Q for the Quarter Ended June 30, 2026
• Alphabet: Second-Quarter 2026 Earnings Release
• Alphabet Form 10-Q for the Quarter Ended June 30, 2026
• Microsoft Fiscal 2026 Fourth-Quarter Earnings Release
• Microsoft Fiscal 2026 Fourth-Quarter Earnings Conference Call
• FASB Accounting Standards Update 2016-01
• FASB Accounting Standards Update 2018-03
• SEC Compliance and Disclosure Interpretations: Non-GAAP Financial Measures
• Amazon: 2022 Full-Year Results and Rivian Valuation Loss
• Associated Press: U.S. Market Close, August 3, 2026

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

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.