Google is providing the payment guarantee that underwrites Anthropic’s $35 billion chip lease across five data centers, Bloomberg reported on June 9, citing the first public confirmation of an arrangement that had not been disclosed in prior partnership announcements.
The structure works like this: Anthropic leases GPUs and TPUs to train Claude; Google backstops the debt-service payments that make the lease cover its obligations. Anthropic gains access to compute it could not independently guarantee. Google takes on AI credit risk in exchange for the strategic position of being indispensable to its most direct frontier competitor.
That last sentence is not a contradiction. It is the defining feature of AI-era capital structure. The same company investing in you, supplying your infrastructure, and competing for your customers is now also your credit guarantor.
Bloomberg’s reporting came shortly after Anthropic filed its S-1 on June 1, and the piece reads like a careful examination of what the SEC’s reviewers are now parsing. The specific detail, that Google is the backstop party on a multi-datacenter compute lease, is the kind of disclosure that lands in a risk-factors section and shapes how institutional investors read the Anthropic cap table.
The scale matters. A $35 billion chip lease is not a venture-style number. It is infrastructure debt. The fact that it required a third-party credit guarantee from a strategic partner suggests Anthropic’s standalone balance sheet, even after multiple billion-dollar raises, was not sufficient to secure this compute footprint on its own terms.
Google’s exposure here is not purely financial. By stepping in as credit support, Google has taken on a formal role in Anthropic’s compute continuity. If Anthropic were to face a liquidity crunch, Google would have standing to act. That is a degree of operational entanglement that goes well beyond a typical corporate venture investment.
This disclosure fits a pattern Bloomberg and others have been assembling across the past several months. Anthropic’s $50 billion American AI infrastructure announcement established the broad capital commitment. The SpaceX-rental disclosures gave one data point on where compute sits physically. The Bloomberg piece fills in the financing mechanic: Google is the credit party that makes the whole structure function.
The broader AI lab IPO wave is producing financial structures that do not have clean precedent. Apple pays Google roughly $1 billion annually to be the default search engine on Safari. Google backstops a $35 billion compute lease for a lab it also funds as an investor. OpenAI has reportedly explored a structure that would include a U.S. government equity stake. Each of these arrangements creates interdependencies that a traditional IPO investor would not encounter in a software company’s cap table.
For Anthropic specifically, the S-1 review will force disclosure of the full terms: the duration of the backstop, whether it is unconditional, what triggers Google’s obligations, and whether the arrangement constrains Anthropic’s ability to shift compute to other providers. None of those terms are public yet. Bloomberg’s reporting identifies the existence of the arrangement; the S-1 will have to characterize its scope.
Investors looking at Anthropic’s IPO filing should read the lease and credit-support disclosures before pricing the company as an independent AI lab. The compute structure is more constrained than the headline valuation implies.
Bloomberg (bloomberg.com), reporting by Bloomberg staff, published June 9, 2026.
