How a 2016 accounting rule supercharged Big Tech investments in AI startups
Artificial intelligence is at the heart of one of the most intriguing financial stories of recent times. And to understand what is really going on behind the triumphant headlines about record profits, we need to look at an accounting mechanism that almost nobody is talking about.
In the first quarter of 2026, Amazon announced net income of $30.3 billion, representing a 77% increase over the same period last year. Revenue from Amazon Web Services, its cloud computing division, grew 28%. Advertising surged. Analysts celebrated what appeared to be the definitive proof that investments in artificial intelligence were finally paying off.
But there is a detail that got somewhat buried in all the celebration: $16.8 billion of that profit did not come from sales, services, or actual operational growth. It came from an accounting revaluation of the company’s stake in Anthropic, done through the mark-to-market method, which records the fair value of an asset based on its current market price rather than its historical cost. This pre-tax gain exceeded all of Amazon’s operating income growth for the quarter, based solely on Anthropic’s most recent funding round. Amazon shares rose more than 4% after the results were released.
Amazon was not alone in this. Alphabet reported roughly $28.7 billion in similar unrealized gains in the first quarter, driven primarily by its portfolio of stakes in private companies, dominated by its share in Anthropic. Nearly half of Alphabet’s record $62.6 billion profit came from updating the value of stakes in companies that are not even publicly traded. Two of the three largest cloud providers in the world reported their best quarters in years, and in both cases, the biggest contributor to the bottom line was a paper gain on a private company that has never paid them a single dividend.
Which raises a question worth asking out loud: is all that profit real, or is it, at least in part, an accounting illusion fueled by a self-reinforcing financial loop? 😬
What the ASU 2016-01 rule actually does
The answer runs through a rule created in 2016 called ASU 2016-01, developed by the FASB (Financial Accounting Standards Board), a nonprofit organization authorized by the SEC (Securities and Exchange Commission) to set accounting and reporting standards for publicly traded companies in the United States.
Before this rule took effect in 2018, companies that held minority stakes in other companies could record those positions at historical acquisition cost, essentially the price they paid at the time of the investment. Unrealized gains on those stakes were recorded on the balance sheet as accumulated other comprehensive income, a category excluded from net income. This created a serious transparency problem because the market had no way of knowing when those assets had appreciated or depreciated significantly.
The rule had a solid rationale. Under the prior regime, financial institutions could hold assets valued at historical cost indefinitely, obscuring their true economic position from investors and regulators. This treatment delayed the ability of accountants and investors to recognize how much assets had deteriorated before and during the 2008 financial crisis. By requiring fair value measurement, the new rule aimed to give markets a more accurate and timely picture of a company’s financial exposure.
In practice, here is what changed: every time a startup receives a new funding round and its valuation goes up, any company that already held a stake in that startup must update the book value of that stake. That update flows directly as a gain into net income, even if not a single penny has changed hands. No cash came in the door, no product was sold, no service was delivered. What happened was simply a new estimate of value.
The current AI landscape has made this dynamic especially explosive. Funding rounds for AI startups have been reaching astronomical valuations in very short periods of time, which means that the Big Tech companies that participated in earlier rounds see their stakes being revalued upward with unprecedented frequency and magnitude.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The cloud credits loop
Before getting into the accounting loop itself, it helps to understand a parallel mechanism that the Federal Trade Commission (FTC) mapped in its January 2025 report on AI partnerships.
The FTC report detailed how major cloud providers invest billions in AI startups: Microsoft has investments in OpenAI, Amazon and Google in Anthropic. A substantial portion of these investments takes the form of credits for the investor’s own cloud platform. The startup spends those credits training models on the investor’s infrastructure. Anthropic, for example, spent $1.35 billion on AWS in 2024 and $2.66 billion from January through September 2025 alone. The investor records that usage as revenue. That revenue growth supports the stock price, which justifies the next round of investment.
This loop, where Big Tech companies inject money into AI startups and the startups send that money right back as payment for cloud services, has four clear stages: invest, spend, record, justify. The first quarter of 2026 revealed a second parallel loop running on top of the same investment.
The mark-to-market loop that inflates the numbers
When Amazon invests $8 billion in Anthropic, and part of that investment is structured as cloud credits for AWS, the money comes back to Amazon in the form of revenue. But the investment also buys an equity stake. When Anthropic raises its next funding round at a higher valuation, as it did in February with its Series G, Amazon’s stake gets revalued upward. The gain flows directly into Amazon’s net income. Amazon’s $8 billion investment in Anthropic is now worth more than $70 billion, according to Amazon itself.
The central question is how independent that valuation change really is. Amazon is both an investor in Anthropic and one of its key commercial partners. When Amazon commits additional capital or deepens its cloud partnership in ways that make Anthropic’s business more viable, it contributes to the conditions that push the valuation, and therefore its own reported net income, higher.
As Robert Willens, a tax and accounting consultant who served as an adjunct professor at Columbia Business School, pointed out: companies are able to control or influence the value of one of their own assets through commercial transactions with that entity.
To be precise about the causal question: Anthropic’s and other AI startups’ new funding rounds are priced by incoming investors, such as venture capital and private equity firms, who conduct their own due diligence. The valuation is not set unilaterally by Amazon or Alphabet. But those incoming investors price the round in a context shaped by the commercial commitments of existing strategic investors. Anthropic’s revenue trajectory, its access to computing infrastructure, and its competitive position depend substantially on the cloud partnerships that Amazon and Google provide. The circularity is partial, but it is real, and it is exactly the kind of entanglement that ASU 2016-01 was not designed to address.
The mark-to-market revaluation thus reveals a second loop running in parallel with the first. The cloud credits loop converts investment into revenue. The mark-to-market loop converts the same investment directly into profit, skipping the revenue step entirely. A single dollar invested in Anthropic pulls double duty: it comes back as cloud revenue and simultaneously appreciates as an equity stake. The more Amazon invests, the more revenue it records and the more profit it reports, without Anthropic ever returning a dollar on the equity itself. In a sense, Amazon, Google, and Microsoft have built twin engines from a single investment.
The risk that few people are seeing
Amazon and Alphabet followed generally accepted accounting principles (GAAP), overseen by the same FASB, in their SEC filings for the first quarter. They clearly disclosed the gains, as required by the 2016 FASB rule. No fraud was committed. The cloud credits loop, along with the profits recorded from investments that raise the value of Anthropic and OpenAI, are, as Bloomberg pointed out, perfectly legal.
However, the question is not whether these deals violate existing rules. It is whether existing rules have become infrastructure for a financial loop that inflates reported results without a corresponding generation of cash.
The cash flow data backs up this concern. In the same quarter that Amazon reported $30.3 billion in net income, its trailing twelve-month free cash flow dropped 95%, to $1.2 billion. Capital expenditures hit $44.2 billion. The company is spending unprecedented amounts on AI infrastructure while reporting record profits, and those profits are substantially driven by the revaluation of an asset whose value depends on the continuation of that spending.
When people talk about an AI bubble, most think of overvalued startups and valuations detached from reality. But the risk this dynamic introduces goes beyond that, because it is embedded in the balance sheets of the Big Tech companies themselves, which are broadly held in pension funds, ETFs, and retirement plans of millions of people. 📉
What happens if the bubble bursts
The mark-to-market gain does not just sit on Amazon’s income statement. Its inflation flows into every investment vehicle it touches, including the household savings of many Americans.
When Amazon’s earnings per share come in at $2.78 against a consensus estimate of $1.64, the stock responds. Market capitalization rises. Because the S&P 500 is weighted by market cap, Amazon’s share of the index increases. Every target-date fund, every passively managed 401(k), every index-tracking ETF adjusts its allocation accordingly. More of the money that flows automatically into these funds with every paycheck gets directed toward Amazon.
Since the Pension Protection Act of 2006, U.S. employers have been incentivized to auto-enroll workers in 401(k) plans and designate target-date funds as the default option. By 2025, 69% of participants in Vanguard-administered plans were invested in a professionally managed allocation, the vast majority in a single target-date fund. The overwhelming majority of retirement savers passively channel their contributions into whatever the cap-weighted index holds.
The critical point is not that passive indexing transmits inflated valuations — that happens with any overvalued stock. The critical point is that the cloud credits loop uniquely generates the inputs that passive investors use to assess value. Earnings per share, the denominator of the price-to-earnings ratio, is inflated by mark-to-market gains that originate within the loop itself. This makes the standard valuation heuristics that retail investors and automated rebalancing systems rely on — headline P/E, earnings surprises, and consensus beats — less reliable as signals of actual economic performance.
The Magnificent Seven trade at roughly 28 to 29 times forward estimated earnings, an elevated level but not in obvious bubble territory by headline metrics. Strip out unrealized gains from investments in entities with which the reporting company has commercial relationships, and the multiple looks quite different. But the headline number is what drives index composition and what shows up on the statement that a retirement plan participant glances at once a quarter.
What regulators are not seeing
Current regulatory attention is focused on the competitive structure of AI markets. The FTC has documented the cloud-AI partnerships. Senators Elizabeth Warren and Ron Wyden have argued that these partnerships function as de facto mergers. The Department of Justice and the FTC, in their joint investigation into competitor collaboration guidelines, are examining non-traditional deal structures, including investments made through cloud credits.
These are important interventions. But they share a common analytical horizon: market power and competitive structure. They do not address the accounting mechanism that converts the self-referential dynamics of the loop into reported profits.
Three measures would help address this gap:
- First: the SEC should require cloud providers to separately disclose the portion of their reported earnings derived from mark-to-market revaluations of entities with which they maintain cloud business relationships. The SEC already has authority under Regulation S-K to require segment-level disclosures. When the entity whose revaluation generates the gain is also a major customer whose spending constitutes revenue for the investor, the standard disclosure framework is inadequate.
- Second: the FASB should revisit whether ASU 2016-01’s requirement to flow unrealized gains through net income is appropriate for stakes in entities with which the holder has a commercial relationship. The goal should not be to return to historical cost, but to ensure that the transparency gains of fair value accounting are not undermined by blending commercially entangled revaluations with operating results. At a minimum, requiring a separate presentation — below the operating income line and clearly labeled as gains from commercially related entities — would preserve the informational value of fair value measurement.
- Third: the Financial Stability Oversight Council (FSOC) should examine whether the interaction between cloud credit investments, mark-to-market accounting, and passive indexing flows constitutes a systemic risk. The FSOC should specifically assess what happens if the loop reverses: a flat or declining valuation round for a major AI startup would trigger mark-to-market losses that would flow through the same income statements, depress the same earnings per share, reduce the same market capitalizations, and cascade through the same index funds that currently transmit the gains.
The crucial difference from the Berkshire Hathaway case
Some readers may recall that this is not the first time ASU 2016-01 has distorted results. Berkshire Hathaway’s quarterly net income has swung wildly since the rule took effect, driven by mark-to-market movements in its public stock portfolio, prompting Warren Buffett to warn investors to focus on operating earnings instead.
But the Berkshire case differs in one fundamental way. Berkshire holds shares in publicly traded companies like Apple, Coca-Cola, and Bank of America, whose valuations are set by liquid markets and whose business results are independent of Berkshire’s own commercial relationships with them. Berkshire does not sell cloud services to Apple, record Apple’s spending as revenue, and then book a gain when Apple’s stock goes up.
The cloud credits loop is different precisely because the investor has a material commercial relationship with the entity whose revaluation generates the gain, making the investor both a beneficiary of and a contributor to the conditions that produce the revaluation. It is this entanglement, not fair value accounting itself, that creates the self-referential loop.
Accounting, transparency, and the future of AI investments
To be clear, ASU 2016-01 is not a malicious rule, nor was it created with the intent to distort results. The FASB had good reasons to require more transparency about the real value of stakes in private companies, and in many contexts, this rule does exactly that. The problem is that no accounting standard is written with every possible future scenario in mind, and the AI investment boom has created a context that amplifies the side effects of this rule in ways its creators probably never anticipated.
The cloud credits loop converts investment into revenue. The mark-to-market gain converts investment directly into profit. Household retirement savings absorb the inflated valuations through passive indexing. Every step is legal. Every step is individually rational. The aggregate result is a self-referential system in which investment, revenue, profit, valuation, and household wealth mutually reinforce one another without requiring the underlying technology to generate proportional returns.
What seems clear, looking at the numbers from the first quarter of 2026, is that the narrative of runaway success in Big Tech AI investments deserves to be read more carefully than the headlines suggest. Impressive profits are still impressive profits, and companies like Amazon and Alphabet remain giants with solid, diversified operations. But when such a significant slice of the bottom line comes from accounting revaluations of assets that have not yet been sold, the question of how much of that growth is structural and how much is a reflection of a valuation cycle that could reverse remains one of the most relevant for anyone following the sector closely.
The question that regulators, accountants, and investors need to grapple with is whether the financial architecture supporting the AI industry has become sophisticated enough to generate its own evidence of success, and whether the accounting standards, securities rules, and retirement structures we rely on can distinguish manufactured performance from real economic value. 🤔
