OpenAI, Anthropic pursue AI services

Enterprise AI is turning into a services business. That is the real story here. OpenAI and Anthropic are no longer acting like model vendors that hand customers an API and walk away. They are building deal vehicles with private-equity backers and using them to buy the people who actually make AI work inside big companies. ### Why are they doing this now? Because the bottleneck is not model quality anymore. It is deployment. A company can buy access to GPT or Claude in a day, but getting that model wired into internal data, security rules, workflows, and legacy software takes months of consulting and engineering work. That gap has become big enough that both labs now want to own it. ### What exactly changed this week? OpenAI’s new venture — described as The Deployment Company in Reuters’ reporting — is raising roughly $4 billion from 19 investors including TPG, Bain Capital, and Brookfield Asset Management. Anthropic unveiled a similar venture with Blackstone, Hellman & Friedman, and Goldman Sachs, with reporting around that effort putting the backing at about $1.5 billion. ### What are these ventures supposed to buy? Services firms. Not chip companies. Not model labs. Reuters says the target list is companies that help enterprises deploy AI, and OpenAI’s venture is already in advanced talks on three deals. The point is to add hundreds of engineers and consultants fast instead of hiring one by one. ### Why not just partner with Accenture or Deloitte? They probably still will. But partnering is different from owning the capability. If OpenAI or Anthropic controls the implementation layer, each customer rollout can be tuned around its own models, its own tooling, and its own upgrade cycle. Basically, the lab stops being just the engine supplier and starts acting more like the general contractor. That is a much stickier position. ### Why is private equity involved? Because this looks a lot like a roll-up. The consulting and IT-services market below the giants is fragmented, full of smaller firms with domain specialists, customer relationships, and billable engineers. Private equity knows how to finance acquisitions and stitch those businesses together. The AI labs bring the product and the demand. The financiers bring capital and a pipeline of portfolio companies to sell into. ### What does Palantir have to do with it? A lot, conceptually. Reuters points to the Palantir playbook — embedding engineers inside customer operations so software gets adapted to the messy reality of how companies actually work. That “forward-deployed engineer” model turns out to fit enterprise AI really well, because every deployment is part software sale, part custom integration, part change management. ### Does this change the AI business model? Yes — or at least it widens it. The dream version of AI was pure high-margin software. The reality is more labor-intensive. If that holds, then the most valuable AI company may not be the one with the smartest model alone. It may be the one that can land the model, customize it, maintain it, and expand it across a customer’s business. That is a very different competitive map. ### So who gets squeezed? Smaller AI consultancies first. They are exactly the kind of firms these ventures may acquire or outcompete. Big global integrators are not going away, but the labs are clearly trying to capture more of the implementation economics themselves. Turns out the fight is moving from benchmark scores to field operations. ### Bottom line? OpenAI and Anthropic are making the same bet: enterprise AI adoption is constrained less by model intelligence than by human implementation capacity. If they are right, the next phase of the AI race will be won partly in consulting, acquisitions, and on-site engineering — not just in the lab.