Amazon's second-quarter profit contained two different AI stories. One came from selling cloud capacity and other services. The other came from marking a private investment to a higher value after new financing evidence. Both can increase shareholder wealth, but only one represents revenue from customers and neither should be used as a shortcut for the other's quality.
The distinction matters because the numbers are unusually large. Amazon's filing records a $50.5 billion upward adjustment to private-company equity, primarily its nonvoting preferred stock in Anthropic. That helped lift other income to $53.4 billion, while reported quarterly net income reached $62.65 billion. Calling the entire increase an operating windfall would misread where it entered the accounts.
Fifty billion entered earnings without entering cash
The Q2 Form 10-Q says the adjustment reflected observable price changes related to Anthropic funding rounds. Amazon classified the preferred shares as a Level 3 measurement, using the rights of different securities, expected liquidity events and a discount for lack of marketability. The gain appears in “other income,” below operating income.
That accounting treatment is not a claim that Amazon sold the stake or received $50.5 billion. It updates the carrying value of a private asset using new transaction evidence. The filing also shows why liquidity matters: an eventual initial public offering would convert instruments into common stock, but Amazon expects a customary lock-up and continuing securities-law restrictions.
The cash-flow statement tells a different story. Amazon reported $53.1 billion of cash capital expenditure in Q2, mainly for technology infrastructure supporting AWS and fulfillment capacity. It also made $24.4 billion of cash payments for acquisitions and other investments during the quarter. Those outflows are not evidence that the strategy is wrong; they are the funding burden against which future operating returns must be measured.
AWS demand is a separate proof
The operating evidence is not weak merely because the valuation gain is non-cash. Associated Press reported that AWS sales grew about 37% from a year earlier, the fastest pace in 18 quarters, while total company net sales reached $200.6 billion. Management raised expected 2026 capital spending to $220 billion and said available capacity still trails demand.
Those facts support a real cloud-demand thesis, but they do not prove every new data center will earn an adequate return. Growth is the numerator; the capital base and depreciation are part of the denominator. A useful assessment therefore separates AWS revenue growth, operating margin and incremental cash generation from the quarterly movement in private-company valuations.
This separation also improves comparisons. If Anthropic's next financing occurs at the same price, no equivalent upward adjustment may recur even if AWS performs well. Conversely, another higher funding round could produce a new accounting gain without accelerating customer usage. The two lines can move together, but accounting does not require them to.
The customer and the investment are connected
Amazon's position is strategically more complex than a passive venture holding. Its filing says it amended a commercial arrangement to provide AWS cloud services, including performance obligations tied to AWS chips, while also investing in Anthropic securities and making a financing facility available as compute milestones are reached. Amazon can benefit from Anthropic's valuation and from its demand for infrastructure.
That coupling may be an advantage: committed workloads can improve utilization, support custom-chip scale and make AWS more attractive to other AI developers. It also makes disclosure harder to interpret. A private rulemaking petition hosted by the SEC argues that cloud providers should identify revenue from companies in which they hold financial interests. This is the petitioner's proposal, not an SEC finding or adopted rule, but it identifies a valid analytical question: how much cloud growth comes from arm's-length demand and how much is linked to financed strategic partners?
Investors do not need to assume circularity to ask for the split. Commercially connected revenue can be genuine and collectible. The issue is concentration and durability. If an AI developer depends on repeated external funding to purchase compute, the quality of the cloud demand depends partly on capital markets. If its own customers fund the workload, the relationship becomes more self-sustaining.
A lower funding price would reverse the optics
Private valuations do not move in only one direction. Amazon's filing states that private-company valuations are inherently more complex because ready market data are unavailable. A lower priced round, impairment or disappointing liquidity event could reduce the carrying value and send a loss through other income. That would hurt reported earnings without necessarily damaging AWS operations, just as this quarter's gain improved earnings without creating operating cash.
The evidence that would strengthen the thesis is therefore specific: sustained AWS growth from a broad customer set; stable or improving segment margins after depreciation; cash generation that catches up with capital spending; Anthropic usage funded increasingly by external revenue; and clearer disclosure of financially connected cloud demand. Evidence that would weaken it includes delayed capacity utilization, repeated financing needed to support purchases, or a down round that exposes the sensitivity of the reported asset value.
Amazon owns an asset that became much more valuable on paper and a cloud business showing strong demand. Both statements can be true. The disciplined reading is to value the strategic option, test the operating business on operating evidence and keep the $50.5 billion revaluation out of the cash-conversion story.