NEWS

Anthropic's IPO tests the circular financing that powers the AI race

Anthropic's IPO tests the circular financing that powers the AI race
Image: Redação iMasters
News
Reading
5 min
In short

Anthropic's stock market debut, on November 19, returns billions to Amazon, Google, Nvidia, and other suppliers that are also investors in the company. The case exposes a model that AI startups in Brazil replicate on a smaller scale, trading equity for cloud credits.

The day the money came back to those who gave it

On Thursday (November 19), Anthropic debuted on the stock market at US$83 per share, raising around US$141 billion in capital and pocketing US$91.7 billion in net proceeds, according to the IPO prospectus cited by TechCrunch. At first glance, it's the biggest market validation an AI startup has ever received.

But the fine print in the prospectus tells a different story. Amazon and Google, Anthropic's two biggest infrastructure partners, together hold stakes valued at US$307.9 billion at the IPO price. Amazon, which invested US$18 billion in the company, sees its 14.7% stake worth up to US$159 billion; Google, which owns 13.7%, comes in at nearly US$149 billion.

How the cycle works

The technical term is "circular financing": the investor is also a supplier, and the money it puts into the company comes back to it when the company buys computing power. Nvidia, AMD, Microsoft, Broadcom, Micron, and TSMC joined Google to commit US$26 billion to Anthropic via private placement, at the same price as the IPO.

Google alone has already invested US$12.3 billion in Anthropic since 2023, with another US$30 billion promised under performance targets. With the exception of TSMC, all the other supplier-investors (Nvidia, AMD, Broadcom, Microsoft, Micron) have announced multibillion-dollar partnerships with Anthropic, which buys their products with part of that same money.

In short: from the outside, it looks like unprecedented revenue growth for the clouds and a record fundraise for Anthropic. In practice, it's the same money changing pockets.

Why this raised suspicion

Gil Luria, head of technology research at D.A. Davidson, draws a line between the investments by Anthropic's suppliers and neoclouds like CoreWeave, which he sees as sustained in a more purely circular way:

CoreWeave wouldn't exist if Nvidia didn't prop them up, and it probably won't exist if Nvidia doesn't bail them out.

Gil Luria, head of technology research at D.A. Davidson

Joey Brookhart, who leads the Tokenomics team at SemiAnalysis, avoids describing Anthropic's financing as purely circular: to him, it's "additive to the ecosystem." According to Brookhart, the difference is that Anthropic already has revenue from real customers growing fast enough that it no longer needs supplier capital:

The revenue is all coming from other customers. Amazon doesn't need to prop them up anymore.

Joey Brookhart, head of the Tokenomics team at SemiAnalysis

What changes from today

Until the eve of the IPO, Anthropic's quarterly deficit of tens of billions of dollars was funded by suppliers and private investors with an obvious conflict of interest. From now on, the capital comes from the public market and corporate debt securities, without that conflict.

With the IPO and the US$26 billion private placement, Anthropic should close the day with around US$220 billion in cash, cash equivalents, and short-term investments. The company can also pay part of its future commitments in stock, not just in cash.

On the other side of the ledger, Anthropic has already committed to spending US$900 billion, mostly on computing and infrastructure, according to a regulatory filing cited by TechCrunch. Third-quarter 2026 revenue was US$17 billion, at an annualized pace of US$68 billion. Brookhart estimates the company needs around US$280 billion in cumulative revenue over the next four years to cover its compute bill, something he calls "achievable" given the revenue level projected for next year.

By the time the IPO is done, Anthropic's going to have more cash on hand than any company has ever had.

Gil Luria, head of technology research at D.A. Davidson

The risk the balance sheet doesn't show

For the two analysts interviewed by TechCrunch, the biggest risk for Anthropic isn't financial: it's competitive. The company is growing because it's at the frontier of generative AI, but losing that position to OpenAI, Google DeepMind, or even cheaper open-weight models would change everything quickly.

If we're all using other people's models, then Anthropic's not worth a trillion and a half dollars. It's worth nothing.

Gil Luria, head of technology research at D.A. Davidson

Curiously, infrastructure suppliers remain protected from this scenario. According to Luria, even if Anthropic loses the race, demand for computing power doesn't disappear: it just migrates to another model running in the same data center owned by Amazon, Google, or Microsoft.

What this teaches those building AI in Brazil

Circular financing isn't exclusive to the giants. Brazilian AI startups frequently strike cloud credit deals with AWS, Google Cloud, or Microsoft in exchange for future usage commitments, sometimes tied to investment rounds from those companies' own venture capital arms (such as AWS Activate or the Google for Startups Cloud Program).

The Anthropic case shows the test this model needs to pass at some point: it solves the cash problem in the short term, but pushes real validation down the road, to when the company needs to prove revenue from real customers, outside the circle of its own investors. Brazilian founders who count cloud credits as runway should ask the same question now facing Anthropic: if the supplier-investor walked away from the table tomorrow, would the company still have customers paying the bill?

It's also worth looking at the cap table with the same care Luria applied to CoreWeave. A startup whose main customer is also its main investor has a structural conflict of interest: the investor-supplier always has an incentive to keep the startup alive, which distorts any outside reading of viability. Separating revenue from independent customers from revenue generated by whoever is on the cap table is the first due diligence exercise any founder should do before raising the next round with embedded cloud credit.

Translated from the Brazilian Portuguese original · Read the original