Anthropic is on the verge of finalizing a $1.5 billion joint venture with private equity, according to a new report from The Information. The deal would mark one of the more unusual capital structures we’ve seen from a frontier AI lab, and it signals just how aggressively Anthropic is hunting for compute and runway as the model arms race intensifies.
The Information reports the Claude maker is close to wrapping up the arrangement, though specifics on the PE partner and exact terms remain under wraps. What stands out here is the structure itself. Most AI labs raise through traditional equity rounds from venture capital or strategic investors like cloud providers. A joint venture with private equity is a different animal.
Why a JV instead of a regular round
Joint ventures let companies raise large amounts of capital without diluting existing shareholders the way a straight equity round would. They’re often used to finance specific assets, infrastructure, or business units rather than the whole company. For Anthropic, that likely means one thing: compute.
Frontier model training is brutally capital-intensive. GPU clusters, data center capacity, and inference infrastructure now run into the tens of billions for top labs. Structuring a JV around those physical assets, instead of issuing more equity at a sky-high valuation, gives Anthropic a way to fund the buildout while keeping the cap table cleaner.
This is significant because it mirrors a pattern we’re starting to see across the AI sector. OpenAI has explored similar off-balance-sheet structures for its data center ambitions. Microsoft, Oracle, and others have committed multi-year compute deals that look more like project finance than software investing.
Context: Anthropic’s funding trajectory
A few data points worth holding in mind:
- Anthropic has raised more than $13 billion to date from investors including Amazon, Google, and a long list of VCs.
- Amazon alone has committed up to $8 billion, with Anthropic agreeing to use AWS Trainium chips for training.
- The company’s most recent valuation discussions have placed it well above $60 billion.
- Annualized revenue has reportedly crossed several billion dollars, driven by enterprise Claude API usage and Claude Code.
Adding $1.5 billion through a PE joint venture on top of that stack tells you the spend curve isn’t slowing. If anything, it’s steepening. The gap between what a frontier lab earns and what it needs to spend on next-generation models keeps widening, and capital markets are responding with creative structures to bridge it.
What this means for the industry
A few takeaways for practitioners and operators:
- Compute is the new real estate. AI labs are increasingly financing GPUs and data centers the way developers finance buildings. Expect more JV, project finance, and infrastructure-style deals.
- PE is entering AI in earnest. Private equity has mostly watched from the sidelines while VCs led the early rounds. A $1.5 billion commitment from a PE partner suggests the asset class now sees AI infrastructure as fundable on PE terms.
- Anthropic is doubling down on independence. Despite deep ties to Amazon and Google, this structure keeps Anthropic from leaning even harder on a single hyperscaler. Diversifying capital sources reduces strategic dependency.
- More capital means more model cadence. Funding like this typically gets earmarked for the next training run. Read it as a signal that Claude’s next major versions are already in the pipeline.
What to watch next
The identity of the PE partner will matter. Whether it’s a generalist firm, an infrastructure-focused fund, or a sovereign-linked vehicle will tell you something about who’s underwriting AI at this scale and on what terms. Also worth tracking: whether Anthropic discloses how the capital gets deployed, especially the split between training compute and inference capacity.
The report from The Information lands at a moment when every frontier lab is rewriting its capital playbook. If this deal closes as described, expect competitors to follow with their own variations. The era of “just raise another equity round” is giving way to something more structured, more leveraged, and more intertwined with the physical infrastructure of AI.
Full details available at the original source.