Both AI Labs Are Now Raising Billions From Wall Street to Deploy AI Into the Real Economy

Both AI labs decided on the same strategy, on the same day, with the same pitch: sending engineers to live inside your company.

Anthropic and OpenAI announced near-simultaneously on May 4 that they are each creating joint venture vehicles backed by private equity capital to deploy enterprise AI services at scale. Anthropic's venture is valued at $1.5 billion, with $300 million commitments each from Anthropic itself, Blackstone, and Hellman & Friedman, plus participation from Goldman Sachs, Apollo, General Atlantic, GIC, Leonard Green, and Sequoia. OpenAI's vehicle, called The Development Company, is raising $4 billion from 19 investors including TPG, Brookfield Asset Management, Advent, and Bain Capital, targeting a $10 billion valuation. Both follow the same model: forward-deployed engineers embedded inside client companies to redesign workflows and integrate AI into core processes, with preferred sales access to the PE firms' portfolio companies.

The model is not new. Forward-deployed engineers — FDEs — have been standard enterprise consulting practice since Palantir popularized the approach in government and defense contexts. The idea is simple: AI tools are only as good as their integration into existing workflows, and integration requires hands-on work that a software API cannot do. Sending engineers to sit with clinicians and IT staff to build tools that fit into workflows staff already use, as Anthropic described it, is labor-intensive and expensive. It is also the only way to actually deploy AI into regulated industries where the workflow is the compliance, not a separate consideration.

The interesting thing is that both labs chose this model simultaneously. That convergence suggests neither believes its base API product is sufficient for large enterprise deployment. The API works for developers who know how to build around it. It does not work for a hospital system that needs its AI tool to fit an EHR workflow, or a bank that needs its AI tool to pass a compliance review before it touches any customer data. The forward-deployed model is an admission that "model that can do the task" and "AI system deployed into a regulated workflow" are different problems — and that the labs cannot solve the second one with better models alone.

The PE backing is both validation and a potential distortion. Private equity firms have different incentives than enterprise software buyers. PE firms buy companies, improve their economics, and sell them — often within 3-5 years. An AI deployment tool backed by PE capital will be evaluated partly on whether it produces results the PE firm can show to justify the next acquisition or exit. That is not the same incentive as an enterprise software buyer evaluating whether AI reduces their per-unit compliance cost. The preferred sales access to PE portfolio companies — a feature of both Anthropic's and OpenAI's ventures — means the deals will be concentrated in companies the PE firm controls, not necessarily in the organizations that would benefit most from AI deployment.

For practitioners, the important distinction is between product deals and services deals. These joint ventures are services deals. The labs are becoming consulting firms with better models, at least in the near term. The forward-deployed engineer model requires human labor that does not scale like software — each engagement requires engineers on site, which means the revenue scales with headcount, not with the number of customers using a product. That is a valid strategy, and it may be what large enterprise deployment actually requires right now. But it is a different business than selling API credits, and practitioners should understand which model they are buying when they engage with these ventures.

The capital amounts are not trivial. OpenAI's $10 billion valuation for The Development Company, with $4 billion raised from 19 PE investors, is one of the largest AI services raises documented. Anthropic's $1.5 billion valuation with $300 million in initial commitments signals that the PE community views this as a durable business model, not a bridge product. Both labs are simultaneously raising large primary rounds — OpenAI at an $852 billion valuation as of March, Anthropic reportedly in final stages of a $50 billion raise at a $900 billion valuation — while also creating these separate venture vehicles. The dual-track capital strategy suggests both labs are building for an IPO while monetizing enterprise adoption through services vehicles that PE capital can accelerate.

The Dimon connection ties the Anthropic venture to the financial services story. Jamie Dimon appeared at Anthropic's May 5 finance agents announcement, the same day as the joint venture news. JPMorgan is a Glasswing consortium member — the vetted defensive network that distributes Anthropic's Mythos-class vulnerability scanning. Dimon's presence signals that the biggest bank in the world is treating Anthropic as a strategic partner, not just an API customer. That relationship, more than the joint venture itself, is the signal worth watching: when the most risk-averse institution in financial services decides an AI lab is trustworthy enough to embed in its workflows, that is a market signal that the rest of the enterprise market will eventually follow.

The forward-deployed model is a tell about where AI enterprise adoption actually is in 2026. If frontier AI models were truly ready for autonomous deployment, labs would not need to send engineers to live inside client companies to rebuild workflows by hand. The fact that both labs chose this path simultaneously is an honest admission that the last-mile problem — getting AI from "works in demo" to "works in your actual business process" — is unsolved by the model itself. The models can do the work. Getting them to do the work inside a regulated company's actual systems, with actual data, under actual compliance constraints, requires human integration work that no API can automate.

That gap is the market these joint ventures are designed to address. Whether PE-backed services vehicles are the right structure to close it — versus the traditional enterprise software motion of selling seats and letting customers figure out integration themselves — remains to be seen. The labs are making a bet that the services model unlocks enterprise revenue faster than the product model. The next 12 months of deployments inside PE portfolio companies will test whether that bet pays off.

Sources: TechCrunch, CNBC, Bloomberg