OpenAI, Anthropic, Google push enterprise

Enterprise AI is starting to look less like software sales and more like a services business with very expensive backers. That’s the real story here. OpenAI and Anthropic both moved on May 4 to create new vehicles for getting their models inside actual companies, not just sold to them, and Google had already made a similar move in April through a partnership with Thoma Bravo. The gap they’re trying to close is simple — plenty of firms want AI, but very few can wire it cleanly into real workflows. ### What changed this week? OpenAI finalized a new venture focused on helping businesses deploy its tools, with more than $4 billion raised from investors including TPG, Brookfield, Advent, and Bain Capital, at a reported $10 billion valuation. Within hours, Anthropic announced its own enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs. These were not ordinary reseller deals — they were new structures built to push adoption from inside customer operations. (anthropic.com) ### What is Google doing in the same lane? Google’s move came a few weeks earlier, on April 15. Google Cloud and Thoma Bravo launched a strategic partnership that gives Thoma Bravo portfolio companies access to Gemini, Gemini Enterprise, Google engineers, and Google’s marketplace and co-sell machinery. So the pattern is bigger than two rivals copying each other in one news cycle — major model providers are all trying to secure distribution through private-equity ecosystems. (bloomberg.com) ### Why bring in private equity at all? Because private-equity firms already control giant clusters of enterprise customers. They own or influence hundreds of software companies, healthcare groups, manufacturers, banks, and service businesses. If an AI lab partners with those firms, it gets a shortcut to buyers, budgets, and executive attention. The investors also get a way to lift the value of their portfolio companies by pushing AI deeper into products and operations. (googlecloudpresscorner.com) ### Why isn’t selling the model enough? Because the hard part is not the demo. It’s the plumbing. Anthropic says deployments need hands-on engineering and familiarity with how each business actually runs, and its new firm will embed Applied AI engineers alongside customer-facing teams. Google is promising forward-deployed engineers too. Basically, the model is the engine, but companies still need someone to bolt it into the car, connect the dashboard, and make sure the brakes work. (googlecloudpresscorner.com) ### What kind of work are these teams doing? Anthropic’s own example is a healthcare services group dealing with documentation, coding, prior authorizations, and compliance reviews. The point is not “add a chatbot.” The point is to build systems that fit the workflow staff already use. That means APIs, identity and access controls, audit trails, reliability, human review, and support over time. In regulated sectors, that integration layer is where most of the pain lives. (anthropic.com) ### Who does this threaten? Commodity IT work looks exposed first. If more of the value shifts to model providers plus tightly coupled deployment teams, generic implementation work gets squeezed. But the opposite is also true for higher-skill services — firms that can do workflow redesign, security, data integration, and long-tail maintenance may become more important, not less. The winners are likely the people who can make AI survive contact with enterprise reality. (anthropic.com) ### Why does this matter beyond one news cycle? Because it shows where the market thinks the bottleneck is. Not training bigger models. Not launching another assistant. The bottleneck is enterprise adoption — getting AI into finance, healthcare, procurement, support, and security systems without breaking everything around them. These new ventures are a bet that the next layer of value sits in deployment, not just invention. (anthropic.com) ### Bottom line The AI labs are moving downstream. They don’t just want to sell intelligence anymore — they want to own the messy last mile that turns a model into business software. And that last mile, turns out, is where a lot of the money and leverage now sit. (anthropic.com)