Anthropic is not just raising money for AI anymore. It is financing compute like infrastructure, and Wall Street is starting to treat chips as a serious credit asset.
Apollo Global Management and Blackstone have finalized a roughly $35 billion financing package tied to Anthropic's AI infrastructure buildout, putting private credit at the center of the next phase of the AI race. The deal matters because it moves beyond the familiar startup playbook of selling more equity at a higher valuation. This is capital being routed directly into chips, leases and computing capacity.
According to Bloomberg, the financing is one of the largest private credit transactions ever assembled and is connected to Anthropic's need for AI chips supplied through Broadcom and Google's TPU ecosystem. Several reports describe the structure as debt arranged through a vehicle that can buy or finance chips and lease that capacity back to Anthropic, rather than simply handing the company another equity check.
That distinction is important. Equity capital tells investors what a company might be worth. Chip financing tells the market what it costs to keep the product running. For frontier AI labs, those two things are now deeply connected, but they are not the same.
Training and serving frontier models used to sound like a technical challenge. It still is. But for companies such as Anthropic, OpenAI and xAI, the harder question is increasingly financial: how do you secure enough computing power without constantly diluting shareholders or depending entirely on a cloud partner's balance sheet?
Anthropic has already raised extraordinary sums. The company announced on May 28 that it had raised $65 billion in Series H funding at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital. That came after a February round that valued the company at $380 billion. In ordinary startup terms, those numbers would be the main story. In AI, they are only part of the capital stack.
The newer financing shows how quickly the market is separating model-company equity from infrastructure funding. Apollo and Blackstone are not buying the same risk as a venture investor betting on Claude's long-term upside. They are financing assets linked to real demand for compute, with chip supply, leases and counterparties forming the foundation of the transaction.
This is why private credit is interested. AI chips are expensive, scarce and essential. If usage keeps rising, the cash flows around compute capacity could look more like infrastructure than software. That does not make the risk disappear, but it makes the opportunity legible to asset managers that already finance aircraft, warehouses, fiber networks and power projects.
Why Broadcom and Google matter
The supplier angle is also revealing. Anthropic has leaned heavily into Google's TPU architecture, which gives it an alternative to Nvidia's dominant GPU ecosystem. Bloomberg reported in April that Anthropic had confirmed plans to work with Broadcom and Google, while saying its revenue run rate had topped $30 billion, up sharply from $9 billion at the end of 2025.
Broadcom's role matters because custom AI chips are not a casual procurement item. They require long planning cycles, manufacturing commitments and confidence that the buyer will keep growing. If financing is linked to TPU capacity, it suggests the capital markets are being pulled into the chip supply chain itself. That is a much deeper relationship than a startup renting more cloud servers by the month.
For Anthropic, this may also reduce dependence on one funding route. It can raise equity for research, product development and corporate flexibility, while using structured debt to support the heavy hardware layer underneath Claude. That is a cleaner match between capital and use of proceeds. Long-lived infrastructure gets financed one way. High-growth company risk gets financed another.
The risk is concentration. If demand slows, if TPU economics disappoint, or if Broadcom and Google cannot deliver capacity on the expected terms, the financing becomes more complicated. Private credit investors are comfortable with structure, but structure is not magic. The value of the chips depends on whether customers keep paying for the intelligence those chips produce.
The IPO race is changing the stakes
The timing makes the deal even more interesting. Anthropic confidentially filed for an IPO on June 1, shortly after its massive Series H round. OpenAI followed on June 8 with its own confidential IPO paperwork, according to the Associated Press and other outlets. Both companies are preparing for public markets while still needing sums of capital that look closer to national infrastructure projects than ordinary software expansion.
OpenAI's route points toward a Wall Street listing and broader public investor access. Anthropic's chip financing points toward another answer: bring Wall Street into the machinery before the IPO, through credit structures backed by compute. Neither approach is necessarily better. They show that AI companies are no longer choosing between venture capital and cloud partnerships. They are building layered financing models.
For investors, this is the practical takeaway. AI infrastructure is becoming its own asset class, separate from the companies that use it and separate from the chips that power it. The market is starting to price not only who has the best model, but who can finance the machines needed to keep improving it.
That may be the next real test for the AI boom. The winners will not simply be the labs with the strongest benchmarks or the biggest valuations. They will be the companies that can turn compute into reliable economics, and convince capital providers that the machines are worth funding long before the profits are obvious.
Also read: Britain’s AI supercomputer bet still needs American chips • Google makes AI Plus cheaper as the subscription fight turns practical • Tools for Humanity cuts staff as World chases bigger identity deals
