Blackstone is pitching a roughly $36 billion debt package to help finance Anthropic's access to Google's AI chips, a mega-deal that underscores how the insatiable demand for compute has quietly become the defining battleground of the artificial intelligence arms race. The proposed financing — first reported by Bloomberg and detailed by Quartz — would bankroll Anthropic's ability to lease and run on Google's custom tensor processing units (TPUs), a story being tracked closely across latest AI business developments.
The talks, which remain fluid, arrive as private credit markets scramble to fund the eye-watering capital needs of frontier AI labs. For Blackstone, the world's largest alternative asset manager, the package represents a chance to lock down a high-yield corner of the AI buildout — lending against long-term compute contracts that the industry treats as digital oil.
A deal that could eclipse an earlier $35 billion package
The new pitch is remarkable in scale on its own, but it lands against the backdrop of an already-massive prior round. According to TradingKey, Apollo and Blackstone previously completed a roughly $35 billion private credit deal described as "chip financing" to underwrite the expansion of Anthropic's computing power. The latest Blackstone proposal could exceed even that figure, according to Silicon UK, which reported that the asset manager is discussing a package to finance the lease of Google AI chips.
In plain terms, the money does not buy Anthropic chips outright the way a consumer buys a laptop. Instead, it underwrites the multi-year agreements that give Anthropic the right to draw on large pools of Google-designed accelerators housed inside data centers — capacity that is scarce, expensive, and spoken for years in advance.
Why Anthropic needs this much money
Anthropic, the maker of the Claude family of models, has been on a compute spending spree that rivals anything in the industry. The company has previously disclosed major investments tied to data center capacity, including a multibillion-dollar loan tied to a Google-backed Texas campus. Training and serving frontier models demands enormous clusters of specialized silicon running around the clock, and the cost of that hardware — plus the electricity, cooling, and buildings to house it — runs into the tens of billions.
Google has emerged as a critical supplier of that capacity. Because Google designs its own TPUs rather than buying exclusively from Nvidia, it can offer Anthropic an alternative stream of accelerators — a strategic lifeline at a moment when Nvidia GPUs are oversubscribed and politically sensitive.
The broader financing arms race
The Blackstone-Anthropic discussions are part of a much wider pattern. Private credit firms and institutional investors are pouring capital into AI infrastructure deals that would have been considered implausible just two years ago. Debt packages, sale-leaseback arrangements, and long-dated compute contracts are proliferating as labs compete to lock in the hardware needed to train the next generation of models.
Anthropic's rivals are pursuing their own paths. Some are securing direct investments from hyperscalers, while others are signing equally enormous infrastructure commitments. The result is a market in which access to silicon — not just algorithms — increasingly determines who can compete at the frontier.
A parallel push into custom silicon
The financing news lands alongside another strategic shift at Anthropic. Reuters and TechCrunch reported that Anthropic is assembling an in-house team to design its own AI chips, with job listings already circulating for hardware engineers. Business Insider confirmed the move, framing it as an effort to gain more control over the hardware that powers Claude.
Designing custom silicon is a notoriously expensive, years-long undertaking, but it offers a potential long-term hedge against dependency on outside suppliers. Taken together — the record debt packages to rent chips today and the quiet effort to build expertise to design them tomorrow — Anthropic's strategy reveals a company trying to secure compute on every available front.
What's at stake
If the Blackstone deal comes together, it would rank among the largest private credit financings ever tied to a single technology company. It would also send a clear signal to investors and competitors: in the current AI cycle, the companies that win are the ones that can finance the most compute, the soonest.
For Anthropic's customers and partners, more financed capacity should mean more reliable access to Claude and faster model improvements. For the broader market, it is further evidence that the economics of frontier AI are now being shaped as much by Wall Street dealmakers as by researchers.
How private credit became the engine of AI
The role of firms like Blackstone and Apollo marks a shift in how AI gets funded. Traditional venture capital, which swaps equity for ownership and bets on long-term growth, is simply not large enough to finance the infrastructure the frontier labs now require. Instead, the industry has turned to private credit — loans secured against future revenue and long-term contracts — to bridge the gap.
This model has precedent in energy and real estate, where multibillion-dollar assets are routinely financed against decades of projected cash flow. What is new is applying the same logic to compute: treating clusters of accelerators, and the contracts to run them, as income-producing assets that can be borrowed against. As that market matures, access to cheap debt may matter as much as access to top engineering talent.
Political and supply-chain pressure
The financing rush is also being shaped by geopolitics. With U.S. export controls tightening around advanced semiconductors and regulators scrutinizing foreign-built hardware in data centers, labs are eager to diversify the silicon they rely on. Locking in capacity on Google's domestically designed TPUs offers one way to hedge against supply shocks and political risk. For investors backing those arrangements, the long-duration contracts provide the predictable returns that pension funds and insurers crave — a rare alignment of interests that helps explain why the numbers keep climbing.
