Amazon and Alphabet posted quarterly results this week that hand Wall Street its clearest evidence yet of a pattern two years in the making: the same handful of companies are now building the AI infrastructure, funding the AI labs that use it, and booking the resulting cloud revenue as growth.

According to NYT, the profit lines at both companies show the AI boom increasingly running in a loop rather than a straight line from investment to independent demand.

That loop matters because it changes what "AI growth" actually measures. When a hyperscaler's own invested capital shows up later as cloud revenue from the company it funded, a rising top line stops being clean evidence of new, external demand for AI.

The dynamic isn't unique to Amazon and Alphabet. Nvidia has taken similar equity positions in AI labs that then spend heavily on Nvidia GPUs, and Microsoft's stake in OpenAI follows the same shape. What sets this week's numbers apart is that they show the loop closing inside two of the biggest cloud balance sheets at once, in the same earnings cycle.

Why "circular" is the right word

Amazon committed up to $8 billion to Anthropic starting in 2023, structured partly as credits redeemable on AWS. Google has put a comparable sum, roughly $3 billion across several rounds, into the same lab, with AI workloads running substantially on Google Cloud infrastructure. In both cases, capital that started on the hyperscaler's balance sheet reappears months later as usage revenue on the hyperscaler's own cloud unit.

The mechanism is reinforced by custom silicon. Amazon's Trainium chips and Google's TPUs let each company capture margin twice — once by avoiding Nvidia's markup on training hardware, and again on the compute-as-a-service layer sold to the labs they helped fund.

Where the profit actually shows up

Both companies' cloud divisions are the place to watch, not corporate topline alone:

• AWS and Google Cloud operating margins have held up even as capital expenditure on data centers and chips has climbed sharply.
• A growing share of that capex is justified internally by committed, multi-year compute contracts from AI labs the companies have equity stakes in.
• Custom chip economics (Trainium, TPU) mean AI-related revenue converts to profit at a better rate than commodity GPU rental.

None of this means the underlying demand is fake. Enterprise customers outside the Amazon-Alphabet-Anthropic triangle are also buying inference capacity. But the earnings reveal how much of the reported growth traces back to capital the hyperscalers themselves put into motion. That distinction — capital re-entering as revenue versus fresh outside spending — is exactly what separates durable AI growth from growth that depends on the hyperscalers continuing to write checks to themselves.

What it means for AI builders

For teams building on top of these platforms, the practical takeaway is about leverage, not ideology:

• Cloud credit deals tied to a hyperscaler's own investment often come with usage or exclusivity terms — read them before treating a "funding round" as free money.
• Pricing power sits with whoever controls both the chips and the customer relationship; expect fewer discounts as circularity deepens.
• Keep an inference and training stack portable (containerized, framework-agnostic) so a change in one hyperscaler-lab relationship doesn't strand your workloads.
• Track effective compute cost per token or per training run, not sticker price — custom silicon deals change the real number substantially.
• Watch whether a hyperscaler's cloud growth commentary cites named enterprise customers or leans on "AI demand" in the aggregate — the latter is harder to verify.

AiiN's takeaway

Circularity by itself doesn't prove the AI boom is a bubble; it means investors and builders should ask what fraction of reported cloud growth comes from demand outside the hyperscaler-lab relationships, not just inside them. In our estimation, the companies most exposed if that outside demand slows are the ones whose margin story leans hardest on their own invested capital coming back as revenue. For now, Amazon and Alphabet's results say the loop is still expanding — the open question is how much of the next leg of growth has to come from customers who never took a check from either company.