In a 72-hour stretch this week, the three largest AI companies effectively signed up some of Wall Street’s biggest buyout firms as distribution partners. Anthropic, OpenAI, and Google all moved in the same direction in the same window, according to reports compiled by Chinaventure, a Chinese investment-news outlet. The coincidence is not accidental: private equity firms, which own thousands of companies that need AI and cannot build it themselves, have become the most efficient channel for enterprise AI adoption.
A busy 72 hours
Anthropic struck first, forming a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs. The structure is simple in concept: the firms bring their portfolio companies, Anthropic brings Claude, and the joint venture deploys the model across the portfolio in exchange for long-term contracts.
OpenAI’s move was bigger and more structured. It closed terms with TPG, Brookfield, and Bain on a $4 billion vehicle, reported by Chinaventure as “DeployCo,” designed to roll out OpenAI’s models across the three firms’ holdings. The arrangement reportedly includes a guaranteed annualized return of 17.5 percent for the capital partners – an unusual feature that shows how far the AI lab is willing to go to lock in enterprise distribution, and how much certainty the buyout firms demanded before committing.
Google, for its part, is reported to be in talks with Blackstone, KKR, and EQT over an all-inclusive Gemini licensing program – essentially a fleet deal that would put Gemini into every company the firms control, from industrial manufacturers to healthcare groups to financial services businesses. If it closes, Google would match both rivals with its own PE distribution arm, completing a clean sweep of the industry’s three biggest labs.
Why PE firms became the channel
The logic is about access. The traditional enterprise software sales motion – a vendor knocking on CIO doors, running pilots, and negotiating procurement – is slow, and AI vendors cannot afford slow. Private equity firms sit at the top of thousands of portfolio companies where the general partner can effectively direct purchasing decisions. A single conversation with a PE firm can put a model in front of more enterprises than a year of field sales.
There is also a capital angle. The guaranteed return in OpenAI’s DeployCo shows that the buyout firms are being compensated for taking deployment risk, and that the AI labs are willing to share economics to get scale. For the PE firms, the deals are a way to add value to portfolio companies – access to frontier models at favorable terms – while earning a return on capital that sits outside their traditional fund structures.
The scale of the opportunity is why the labs are moving so fast. A single large buyout firm can hold several hundred companies across manufacturing, healthcare, logistics, and financial services, and each one is a potential deployment site for coding agents, customer service automation, and internal knowledge tools. PE firms have also been quietly building their own AI enablement teams, which means they can absorb a frontier model into their portfolios faster than a traditional enterprise can evaluate one.
From product to service
The deeper shift in these deals is about what AI companies are actually selling. The FDE – forward-deployed engineer – model, in which AI vendors embed engineers inside customer organizations to build custom solutions on top of their models, has been spreading through the industry for over a year. These PE distribution deals take that logic to industrial scale: instead of one engineer at one customer, a single agreement blankets an entire portfolio.
That has consequences for the software industry. If AI is sold as a service attached to private equity portfolios rather than as a product with a license, the traditional SaaS economics – high gross margins, self-serve onboarding, standardized pricing – come under pressure. The AI war, as Chinaventure put it, is moving from being product-centered to service-centered, and the firms that own the customers are collecting a toll on every deployment.
The model carries risk on both sides. For the buyout firms, a failed deployment is a public event: models that underperform in a portfolio company’s operations, integrations that stall, or returns that fall short of a promised 17.5 percent will be measured against the guarantee. For the AI labs, service-led distribution brings lower margins and heavier operational loads than selling software licenses, and the guaranteed-return structures mean the labs are effectively underwriting part of the risk themselves.
Three labs, three deals, one strategy: rent the distribution networks that Wall Street spent decades building. The winners in the next phase of the AI race will be decided less by model quality alone than by who can put their models in front of the most enterprises. Right now, the buyout firms hold the keys.
