OpenAI Just Launched a $4 Billion Consulting Firm, And It Breaks Every Promise

OpenAI, the company that spent a decade telling the world it was a research lab building AGI for humanity, just launched a $4 billion IT consulting business.

On May 11, 2026, OpenAI announced the "OpenAI Deployment Company" , a standalone entity backed by 19 institutional investors at a $14 billion valuation. Its purpose is not to build better models. It is to embed engineers inside corporate clients, identify where AI fits, and build production systems on-site. Simultaneously, the company agreed to acquire Tomoro, an AI consulting and engineering firm that brings roughly 150 Forward Deployed Engineers into OpenAI's orbit from day one. Tomoro already works inside brands like Tesco, Mattel, Red Bull, Virgin Atlantic, and the NBA, building AI travel concierges, in-game support agents, and enterprise deployment pipelines.

The move comes exactly one week after Anthropic launched a nearly identical enterprise services venture with Blackstone, Hellman & Friedman, and Goldman Sachs, backed by $1.5 billion in committed capital. That is $11.5 billion in combined AI services ventures, in a single week, from the two companies that were supposed to be competing on model intelligence, not billable hours.

The two companies that defined the AI race are converging on the same destination. And it looks nothing like what either of them promised.

What Happened

The OpenAI Deployment Company , which the company refers to internally as DeployCo , is a majority-owned subsidiary, structurally separate from OpenAI's core AI research operations. The separation is deliberate. It allows OpenAI to preserve the "research lab" identity while scaling a fundamentally different kind of business.

The $4 billion in committed capital comes from a consortium led by TPG, alongside McKinsey & Company, Bain & Company, Goldman Sachs, and 15 other institutional investors. The $14 billion standalone valuation for a company with no products of its own , just people and deployment methodology , signals how seriously the market is taking the "last mile" problem in enterprise AI.

The Tomoro acquisition provides immediate operational capacity. Founded in 2023 in partnership with OpenAI, Tomoro specialized in exactly what DeployCo needs: sending engineers into large organizations, identifying AI use cases, and building production systems that work. The firm's track record includes building an AI travel concierge for Virgin Atlantic, launching an in-game support agent for Supercell that served 110 million users in 12 weeks, and managing enterprise deployments for Fidelity International, the NBA, Tesco, and Red Bull. Tomoro also works closely with Microsoft and Nvidia as strategic technology partners , relationships that now feed directly into OpenAI's enterprise pipeline.

DeployCo does not exist in isolation. In February 2026, OpenAI announced its Frontier Alliance , multi-year partnerships with BCG, McKinsey, Accenture, and Capgemini to help enterprise clients define AI strategy, redesign workflows, and manage organizational change. DeployCo adds the technical implementation layer. Together, the Frontier Alliance handles the strategy and DeployCo handles the build. OpenAI now covers the full consulting stack , from boardroom PowerPoint to production code.

And Anthropic is building the same stack. Its May 4 joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs mirrors the DeployCo structure: a standalone entity, PE backing, embedded Anthropic engineering resources, and preferred access to Claude for portfolio companies. Each founding partner contributed roughly $300 million, with Goldman Sachs adding approximately $150 million. The combined firepower of both ventures , $5.5 billion in committed capital targeting a $375 billion industry , makes this the largest simultaneous services-market entry in tech history.

Both ventures are reportedly already in talks to acquire additional AI services firms to scale faster, according to reports. The race is no longer about who has the better model. It is about who can put more engineers inside more enterprises faster.

Why It Matters

OpenAI realized that selling AI models is a commodity business. Selling AI implementation is where the money actually lives.

The numbers make the logic clear. Enterprise now represents more than 40% of OpenAI's revenue , roughly $10 billion annualized at the company's reported $2 billion per month run rate , and is on track to reach parity with consumer revenue by the end of 2026. Consumer subscriptions and API credits are volatile. Enterprise services contracts are multi-year commitments with predictable revenue. For a company targeting a $1 trillion IPO in Q4 2026, that distinction is the difference between a successful public debut and a valuation correction.

But the deeper driver is the "last mile" problem that has plagued enterprise AI for two years. Companies are spending heavily on AI tools and model access, but they are struggling to prove return on investment. Buying API access to GPT models does not automatically translate to business results. Somebody needs to identify which workflows to change, integrate the models into existing systems, train internal teams, and manage the organizational resistance. OpenAI spent two years watching enterprises buy its models and then fumble the implementation. DeployCo is the admission that the API-alone model was not enough.

Then there is the scale of the target. Global consulting is a $375 billion industry. When both OpenAI and Anthropic launched PE-backed services ventures in the same week, industry observers described it as "a coordinated $11.5 billion strike on the $375 billion consulting industry." The prize is not just enterprise AI adoption , it is the entire consulting budget that enterprises currently spend with McKinsey, Accenture, Deloitte, and the rest.

And Anthropic forced OpenAI's hand. When Anthropic announced its $1.5 billion venture with Blackstone and Goldman on May 4, the strategic logic became inescapable. Neither company can afford to let the other own the enterprise deployment channel. If Anthropic's embedded engineers are inside Fortune 500 companies recommending Claude, those companies are not buying GPT-5.5. The AI model war is being fought in procurement meetings, not benchmark scores.

There is a deeper pattern here that goes beyond two companies racing into services. OpenAI and Anthropic are both burning cash at extraordinary rates , OpenAI projects $14 billion in losses for 2026, while Anthropic's infrastructure commitments alone now exceed $200 billion over five years. Enterprise services contracts, with their multi-year commitments and predictable billing cycles, are the most reliable way to close the gap between astronomical spending and the revenue needed to justify it. DeployCo is not just a strategy. It is a financial necessity.

The Real Problem Is What OpenAI Used to Stand For

The uncomfortable truth behind DeployCo is not that OpenAI is entering consulting. It is that consulting was always the real business model , and the research lab story was the marketing.

OpenAI was founded in 2015 as a nonprofit. Its stated mission was to ensure that artificial general intelligence "benefits all of humanity." The founding charter emphasized openness, safety, and the long-term welfare of society over short-term commercial gain. In 2026, the same organization is running a $4 billion consulting subsidiary, acquiring a for-profit services firm, and racing toward a $1 trillion IPO. The gap between founding mission and current reality has never been wider , and DeployCo makes the gap impossible to ignore.

The historical parallel is Palantir, not Google. Palantir built its entire business around Forward Deployed Engineers , engineers who embed directly with government and commercial clients, build custom solutions on top of Palantir's platforms, and capture multi-year contracts in the process. Palantir's FDEs are not a side project. They are the delivery mechanism. OpenAI's DeployCo follows the same blueprint almost exactly. The difference , and it is an awkward one , is that Palantir never claimed to be saving humanity. It was always a defense contractor with software. OpenAI spent a decade on a different story.

Then there is the margin problem. AI model companies enjoy gross margins above 70%. Consulting firms operate at 25% to 35%. DeployCo structurally dilutes OpenAI's overall margin profile. The bull case , and the one OpenAI will take to its IPO roadshow , is that services lock in enterprise customers, who then spend more on models, yielding higher combined revenue that more than compensates for the margin drag. The bear case is that services never scale like software. Every dollar of DeployCo revenue costs 65 to 75 cents to deliver, while every dollar of API revenue costs 25 cents. When you are an $852 billion company trying to justify a $1 trillion valuation, margin math matters enormously.

The conflict-of-interest problem is equally uncomfortable. OpenAI now simultaneously sells AI models and competes with the consulting firms that recommend which AI models enterprises should buy. McKinsey is a DeployCo investor. McKinsey is a Frontier Alliance partner. McKinsey also competes directly with DeployCo for the same enterprise AI implementation budgets. When a Fortune 500 CIO asks McKinsey which AI platform to adopt, and McKinsey is financially incentivized toward OpenAI, the independence that clients rely on disappears.

What happens to the word "deployment"? OpenAI has always called itself a "research and deployment company." The phrase has been in the company's self-description since the early days. But there is a structural difference between deploying models via API , which scales with software margins and a small support team , and deploying engineers via billable hours , which scales with headcount. DeployCo is the second kind of deployment. It does not scale like software. It scales like Accenture.

And for all of this criticism, Anthropic is following the exact same path. Same PE-backed services venture. Same embedded engineer model. Same blurring of the line between "AI safety company" and "enterprise IT consultancy." The two AI leaders are independently converging on a business model that looks less like the research labs that captured the world's imagination and more like the consulting firms they are now competing against.

Comparison: The AWS Playbook vs. The Palantir Playbook

There are two ways to read DeployCo , and which one you choose determines whether you see the move as a strategic masterstroke or a margin disaster.

The AWS Professional Services model is the optimistic read. When Amazon Web Services launched cloud infrastructure, enterprises needed help migrating. AWS Professional Services was born , a team of consultants who helped clients plan migrations, train teams, and deploy workloads. But AWS ProServ stayed intentionally small relative to AWS's product revenue. It was an adoption accelerator, not a standalone business. Margins were never the point. The point was getting more customers onto AWS, where the real money was made. In this reading, DeployCo is ProServ for the AI era , a necessary investment to accelerate enterprise adoption of OpenAI's core model business.

The Palantir Forward Deployed Engineer model is the more aggressive read , and the one the numbers support. Palantir built its entire business around embedded engineers. FDEs are not a sales channel. They are the product. Palantir's government and commercial contracts depend on humans-in-the-loop who configure, customize, and operate the software for clients. The $4 billion scale of DeployCo , plus the $14 billion standalone valuation, plus the Tomoro acquisition bringing 150 engineers on day one , suggests OpenAI is going the Palantir route. Services as a major business line, not an accelerator. A revenue stream with its own growth expectations and its own margin profile.

Anthropic's version, backed by $1.5 billion from Blackstone, Goldman, and Hellman & Friedman, confirms the pattern. Both AI leaders have independently reached the same conclusion: services cannot remain a side project. They must be a core business. And if that is true, it has a sobering implication for the entire AI industry.

If the two leading AI labs both decide that pure-play model companies are not viable as standalone businesses, the ceiling for every other model company just got lower. The future of AI might look less like a software company and more like a consulting firm with proprietary technology , high-touch, labor-intensive, and fundamentally constrained by headcount.

What's Next

The next 90 days will bring a rapid consolidation of the AI services market. Both OpenAI and Anthropic are reportedly in active talks to acquire additional AI consulting and engineering firms. The goal is simple: scale DeployCo-like entities as fast as possible before the other side does. Traditional IT services firms , TCS, Infosys, Wipro, Cognizant , face the most immediate threat. Their AI implementation revenue, which has been a growth narrative for the past two years, is now directly contested by the companies that build the AI models themselves.

For OpenAI's IPO roadshow, DeployCo's revenue trajectory will be one of the most scrutinized narratives. If enterprise services revenue grows faster than API and consumer subscriptions, public market investors will price OpenAI as a hybrid , part software company, part consulting firm. The multiple compression could be significant. Software companies trade at 15 to 25 times revenue. Consulting firms trade at 1 to 3 times.

The Big 4 consulting firms are hedging. McKinsey, BCG, Accenture, and Capgemini are simultaneously OpenAI's partners and its competitors. They are building their own AI capabilities while collecting fees from the Frontier Alliance. The real strategic contest of the next three years is not OpenAI versus Anthropic. It is AI labs versus the $375 billion consulting industry , and the consulting industry has been playing this game a lot longer.

In the long run, the walls between "product company" and "services company" are dissolving across the AI industry. The winners will be the organizations that can deliver outcomes, not just models. That sounds obvious. But it represents a fundamental redefinition of what an AI company actually is. And it raises the same question for every industry that thought it was different: if the crypto exchange cannot resist the AI restructuring playbook, who can?

The AI industry began with a promise of automation, software that replaces human labor at near-zero marginal cost. In May 2026, its two leading companies collectively raised $11.5 billion to hire thousands of humans to deploy their software. The irony is not subtle.

FAQ: Common Questions About OpenAI's Consulting Pivot

Is OpenAI still a research company?

OpenAI maintains its core AI research operations separately from DeployCo. The deployment subsidiary is a majority-owned but structurally separate entity. However, the sheer scale of the $4 billion services investment , and the fact that enterprise now represents over 40% of OpenAI's revenue , means the company's identity is fundamentally hybrid.

How is this different from what Anthropic is doing?

The structures are nearly identical. Both companies launched PE-backed services ventures in the same week, both have embedded engineering teams, and both are targeting enterprise deployment as a core business line. The main difference is scale: OpenAI's DeployCo is $4 billion at a $14 billion valuation, while Anthropic's venture is $1.5 billion.

Does this mean OpenAI is competing with McKinsey and Accenture?

Yes, in implementation and technical deployment. The Frontier Alliance partnerships with BCG, McKinsey, Accenture, and Capgemini handle strategy and change management, while DeployCo handles technical build. But the line is blurry, and the consulting firms are simultaneously partners and competitors.

What does this mean for OpenAI's IPO?

Enterprise services contracts provide predictable, recurring revenue , attractive to public market investors. But services margins (25-35%) are significantly lower than software margins (70%+), which could compress OpenAI's overall valuation multiple. The IPO narrative will depend on whether investors price OpenAI as a software company or a hybrid.

Should enterprises be concerned about conflict of interest?

Potentially. OpenAI now both sells AI models and competes with the consulting firms that recommend which models enterprises should buy. When a consulting firm that is also a DeployCo investor recommends OpenAI's technology, clients should ask whether the recommendation is independent.

The AI industry began with a promise of automation , software that replaces human labor at near-zero marginal cost. In May 2026, its two leading companies collectively raised $11.5 billion to hire thousands of humans to deploy their software. The irony is not subtle.

What this tells us about the state of AI is more interesting than the news itself. The technology is powerful enough to transform how enterprises operate, but not yet mature enough to do it without an army of embedded engineers holding the enterprise's hand. For knowledge workers watching this unfold, the question is whether you need a Forward Deployed Engineer from a $4 billion consulting subsidiary , or whether an AI-native knowledge base that puts AI directly in your workflow, on your terms, achieves the same result without the billable hours.