On May 4, 2026, Anthropic and OpenAI both announced joint ventures with major private equity firms to deliver enterprise AI services. Anthropic is anchoring a $1.5 billion firm with Blackstone, Hellman & Friedman, and Goldman Sachs. OpenAI finalized a $10 billion vehicle called The Deployment Company with TPG and a consortium of private equity backers. Frontier labs have moved from selling tokens to selling outcomes, and the AI consulting market just got rewired in a single day.
What Was Announced and By Whom
Two separate but parallel deals dropped on the same day.
Anthropic's joint venture, confirmed in a Blackstone press release and reported by CNBC, is a standalone enterprise services company with $1.5 billion in committed capital. Anthropic, Blackstone, and Hellman & Friedman are each putting in roughly $300 million, with Goldman Sachs investing about $150 million as a founding partner. The new firm is also backed by General Atlantic, Leonard Green, Apollo Global Management, GIC, and Sequoia Capital. Applied AI engineers from Anthropic will embed inside the firm to identify use cases, build custom solutions, and support customers long term. Initial focus sectors include healthcare, financial services, manufacturing, retail, real estate, and infrastructure.
OpenAI's parallel move, reported by Bloomberg and The Next Web, is a $10 billion vehicle called The Deployment Company. OpenAI is committing up to $1.5 billion, structured as a $500 million equity contribution at close plus an option for an additional $1 billion. A 19-investor PE consortium anchored by TPG is contributing approximately $4 billion over five years, with Brookfield Asset Management, Advent International, Bain Capital, Goanna Capital, Dragoneer, and SoftBank among the named partners. The Deployment Company will use the forward-deployed engineer model long associated with Palantir, embedding OpenAI engineers inside client organizations. Target sectors include healthcare, shipping, factories, and finance. According to the same reporting, OpenAI is guaranteeing PE backers a 17.5% annual return over the five-year period, which is unusual for an operating partner.
These announcements are not coincidental. Both labs are explicitly going after the same market: midsize and large enterprises that have struggled to translate model access into deployed outcomes.
Why Frontier Labs Want to Be Services Companies
For two years, the conventional wisdom was that frontier labs would sell models and let consultancies, systems integrators, and software vendors build the application layer. That story is now visibly over.
The economics of pure model sales have hit a ceiling. Token prices keep falling, capital expenditure for compute keeps rising, and the gap between a model that demos well and a model that produces business value sits in the deployment work, not the model itself. Anthropic and OpenAI both saw the same pattern: customers buy the model, run a successful pilot, and then stall. That is the exact dynamic we have written about in why most AI projects stall between pilot and production. Owning the deployment work captures the value the labs cannot extract through API pricing alone.
There is a second, structural reason. Private equity-owned companies are an attractive distribution channel. Sources cited by Bloomberg and the Financial Times note that the PE backers of The Deployment Company collectively have access to more than 2,000 portfolio companies. The Anthropic JV's anchor partners, Blackstone and Hellman & Friedman, similarly own hundreds of companies between them. PE firms are highly motivated to install AI in their portcos to drive multiple expansion at exit. The deal structure makes that installation easy: the PE owner is also a co-investor in the consulting firm, so internal recommendations are automatic.
Our take: This is a captive distribution play wrapped in a joint venture. The lab brings the model and the engineering talent. The PE firm brings a customer base it controls. The economics work because customer acquisition cost is essentially zero inside the portfolio.
What Changes for Businesses Buying AI
Three things shift right away for AI buyers.
The buy options just doubled. Until last week, a midsize company looking for help deploying Claude or GPT had three credible categories of partner: a Big Four consultancy with an AI practice, a boutique AI consultancy, or an in-house build. Now there is a fourth: a frontier-lab native services firm with Anthropic or OpenAI engineers embedded in the team. That changes RFP shortlists in healthcare, financial services, manufacturing, retail, real estate, infrastructure, shipping, and factories. Those are the sectors both ventures named explicitly.
The default vendor advantage tilts. Fortune characterized the Anthropic deal as a shot at the consulting industry, and that framing is accurate. A firm with first-party model engineers has unfair advantages on speed, model tuning, and access to roadmap. It also has unfair conflicts: the same firm recommending a model is the firm selling the deployment of that model. Procurement teams used to evaluating model-agnostic consultancies now have to think harder about lock-in by service contract, not just by API.
The deal terms tell you what to negotiate. OpenAI's reported 17.5% guaranteed annual return to PE backers is a signal. The lab is willing to subsidize returns in order to build deployment muscle and lock in long-term enterprise relationships. That suggests there is room to negotiate aggressive terms early in the firm's life cycle. Companies that move first on these joint venture engagements may extract better pricing and service guarantees than late entrants.
How to Decide Whether a Frontier-Lab JV Is Right for Your Business
Most companies will not need to pick a frontier-lab joint venture today. Both ventures explicitly target portfolio companies of their PE backers first, and both are still standing up engineering teams. But the JV option will become a real RFP entry within twelve months, and the buying decision is worth thinking through now.
The case for going with a frontier-lab joint venture is strongest when:
- You are a portfolio company of one of the JV partners. Internal recommendations and aligned incentives make the procurement work itself easier. Implementation timelines and pricing will likely be more favorable than for outside customers.
- Your use case depends on close model-to-product alignment. If your AI workflow needs custom fine-tuning, advanced agent orchestration, or pre-release feature access, an embedded team from the lab itself has structural advantages over a third-party integrator.
- You want a single throat to choke. Vendor management is simpler when one entity owns the model, the deployment work, and the ongoing support. The cost is reduced flexibility if a different model later becomes the right choice.
The case against is strongest when:
- You want to remain model-agnostic. Most businesses we work with end up using more than one model in production, with Claude, GPT, Gemini, and an open-source option each handling different jobs. A JV bound to a single lab will not meaningfully recommend a competitor.
- Your data and IP risk profile is high. A firm with embedded Anthropic or OpenAI engineers raises real questions about training data flows, model improvement loops, and where customer-specific learnings live. These are answerable questions, but they are questions you do not face with an independent consultancy.
- You are not in the JV partners' priority sectors. Both ventures named sectors. If you are outside healthcare, financial services, manufacturing, retail, real estate, infrastructure, shipping, or factories, you are probably not the priority customer for the early teams these firms hire.
What This Says About the Broader Vendor Stack
The joint ventures are part of a wider pattern. Frontier labs are systematically moving up the stack from raw model APIs into application tools, agent platforms, and now embedded services. The Anthropic AI design tool launch in April competed with Figma and Adobe. Anthropic's $100 billion AWS compute deal tied frontier model supply to a single hyperscaler. The new joint ventures extend that pattern into the services layer.
For CIOs and CTOs, the practical implication is that "AI vendor" is no longer one decision. It is a layered stack of decisions: model, hosting cloud, agent platform, application tools, and now services partner. Each layer has its own competitive dynamics, switching costs, and lock-in risks. Treating any of these as a commodity selection is increasingly wrong.
Key Takeaways
- On May 4, 2026, Anthropic launched a $1.5 billion enterprise AI services firm with Blackstone, Hellman & Friedman, and Goldman Sachs.
- The same day, OpenAI finalized a $10 billion joint venture called The Deployment Company with TPG and 18 other private equity backers.
- Both ventures use the forward-deployed engineer model, embedding lab engineers inside customer organizations, and target similar sectors including healthcare, financial services, and manufacturing.
- Frontier AI labs are now competitors to consultancies and systems integrators, not just suppliers to them.
- For AI buyers, the joint venture option is most attractive for PE-owned portfolio companies and use cases requiring deep model-product alignment, and least attractive for businesses that need a model-agnostic, integration-heavy partner.
Not sure where these new frontier-lab services firms fit in your AI roadmap? Book a discovery call and we will help you figure that out, no strings attached.