Big Tech Accused of Routing Cloud Payments to Itself

Critics allege Microsoft, Amazon, Google and Oracle have structured investments in OpenAI and Anthropic so the money returns to the investor as cloud-payments, creating what opponents call a round-trip funding loop. Corporate filings show OpenAI and Anthropic account for more than half of about $2 trillion in those companies’ future cloud commitments.

Under the pattern, a technology firm provides funding to an AI startup in the form of cloud credits or similar arrangements. The startup uses those credits to buy compute and storage from the funder’s cloud service. The cloud provider records the usage as commercial revenue while the investor can mark up the value of its equity stake and record a paper gain on the income statement. Current accounting rules allow the transactions.

Microsoft’s stake in OpenAI is the largest cited example. That investment included substantial Azure cloud credits. OpenAI’s cloud consumption has been reported above $60 billion a year while the company’s reported revenue is nearer $25 billion.

Anthropic’s arrangement with Amazon follows a similar pattern. The developer spent roughly $2.66 billion on Amazon Web Services over nine months, a level close to its reported income for the same period. Critics say those payments let cloud vendors recognize additional revenue while the startups’ spending flows back to the vendor as cloud rent.

Company filings show the downstream accounting effect. Alphabet posted a record quarterly profit that included a multibillion-dollar markup tied to its Anthropic stake. Alphabet’s filings indicate about $28.7 billion of its reported profit that quarter derived from an Anthropic-related valuation change. Amazon’s filings show roughly $16.8 billion of a reported net income figure linked to transactions tied to Anthropic; the same period showed free cash flow down about 95% to roughly $1.2 billion and more than $44 billion invested in physical data centers. Microsoft attributes roughly half of its approximately $627 billion future backlog to a single customer relationship with OpenAI, and Oracle disclosed that about 54% of its roughly $553 billion pipeline depends on one buyer.

The structure can affect companies that pay cash for AI services. One large enterprise reported using its full annual AI coding budget by April after rolling out third-party developer tools, with some engineers incurring monthly API costs in the $500 to $2,000 range. Other employers have limited or blocked internal use of certain third-party tools after token consumption spiked.

Companies that supply compute report higher internal spending on cloud and chips. Nvidia’s vice president of applied deep learning, Bryan Catanzaro, has reported that his group now spends more on compute than on human salaries. Analysts note that lower per-unit prices for compute can encourage heavier workloads that push total spending higher.

Some investment frameworks highlight measures such as aggregate earnings growth, earnings quality, valuations versus history, the affordability of corporate capital expenditure and the interest-rate cycle. Recent corporate filings triggered flags on earnings quality and capex affordability for several large technology firms.

Social media commentary has framed the arrangement in blunt terms. One post read, “Bro just buy my servers with the money I invested in you,” describing the cycle where investment funds are spent on the investor’s own infrastructure.

Supporters of the funding and billing arrangements point to compliance with current accounting rules and say the investments and infrastructure spending support AI development and product rollout. Critics and supporters continue to debate how the financing and accounting practices affect corporate cash flows, reported earnings and the sustainability of large-scale cloud investment.