On May 11, 2026, OpenAI announced a $4 billion subsidiary built to do the enterprise integration work its customers had been hiring Accenture, Cognizant, and Infosys to do. The press release thanked, among others, McKinsey and Bain — Accenture's two closest competitors — for writing cheques into the round. By the closing bell, Accenture was down about 3%, Cognizant about 5%, and Infosys about 4%, per the Seeking Alpha desk note. The OpenAI Deployment Company — internally called DeployCo — is the structural announcement of the year and the one that explains why so many enterprise contracts have been quietly slipping.

The company's own announcement is short on numbers. The wire reporting and the limited-partner press releases fill them in. PYMNTS confirmed a $4 billion initial round at a $10 billion valuation. Brookfield disclosed a $500 million investment of its own. Bain & Company put out its own release with named partners. Stacked together, the structure is unusual enough to deserve walking through twice.

The 19 investors who paid OpenAI to compete with them

The capital stack, per The Decoder's breakdown and the Bain release, runs as follows. TPG leads. Advent International, Bain Capital, and Brookfield Asset Management co-lead. Goldman Sachs, SoftBank, and Warburg Pincus are limited partners. BBVA — the Spanish bank with 120,000 employees across 25 countries — is both an investor and the marquee customer reference. The 19th investor list also includes Bain & Company, Capgemini, and McKinsey & Company — three of the consulting houses whose existing AI-integration practices DeployCo is most directly built to compete with. OpenAI retains majority control of the new entity.

Rebecca Burack, head of Bain's global private equity practice, framed the deal in the firm's own release: "Creating value in portfolio companies today takes both strategic insight, real technical capability, and change management." Chuck Whitten, Bain's global head of digital, added the line that matters: "The companies that deploy AI fastest and most effectively will pull away from their competitors." Bain is paying to make sure that when the pull-away happens, Bain is selling both ends of the trade — the partnership in DeployCo and the change-management work alongside it. McKinsey and Capgemini bought the same option.

It is the kind of capital stack that only makes sense if every limited partner has decided that the alternative — staying outside, watching DeployCo eat the integration market — is worse than funding the venture that may eat your practice. The defensive-LP logic that Microsoft once used to lock in OpenAI is now being run by everyone else, against OpenAI.

The Tomoro acquisition: where the bodies come from

DeployCo on May 11 was a balance sheet and a press release. The actual work gets done by a 150-person team that OpenAI is acquiring in parallel: Tomoro AI, an Edinburgh-based consultancy founded in 2023. The customer book they're bringing is not toy: Tesco, Virgin Atlantic, and Supercell, per Decoder. The headcount transferring is ~150 forward-deployed engineers, in the language of the announcement. The Tomoro team confirmed the move in their own LinkedIn post: "Our belief hasn't changed, but the scale of the mission has. At The Deployment Company, we're going to build that future and amplify the change we're already creating for our customers."

The choice of language is doing real work. "Forward-deployed engineer" is not a term consultants use. It is a term Palantir uses, borrowed from the military's forward-deployed — meaning embedded in the customer's environment rather than billing time from a remote office. Palantir built a $300+ billion company on the premise that the FDE — the engineer who lives at the client and ships against their workflow — is the unit of competitive advantage that pure software cannot replicate. The Decoder's framing of DeployCo as "adopting Palantir's playbook" is the right read. The difference is that Palantir built its FDE bench one company at a time. OpenAI bought 150 of them on day one and put a $4 billion balance sheet behind hiring the next 1,500.

What gets sold, and to whom

The customer side is where the cheek of the announcement is loudest. BBVA — 120,000 employees, 25 countries — is the first named anchor. According to Decoder's reporting, roughly 2,000 portfolio companies from the LP investors are already in queue as the addressable inbound. That number is the part that should make every services-integrator analyst flinch: the 19 investors did not write cheques in exchange for equity alone. They wrote them in exchange for priority access to joint Bain–Deployment Company work — Bain's own phrasing — for their PE-backed portfolios. The deal locks in a deployment pipeline that the consulting incumbents have spent twenty years building inch by inch.

OpenAI's Chief Revenue Officer, Denise Dresser, put the official line on it: "AI is becoming capable of doing increasingly meaningful work inside organizations. The challenge now is helping companies integrate these systems into the infrastructure and workflows that power their businesses." That is a sentence designed to be quoted in an S-1. It is also the cleanest statement yet that OpenAI considers integration revenue — the revenue Accenture and Capgemini have been collecting on its products — to be revenue it should be collecting itself.

How the consulting market priced it in

The market did the rest of the work. Within the day, Accenture closed down roughly 3%, Cognizant about 5%, and Infosys about 4%, on volume that signalled this was not a rotational trade. UBS held its Buy rating and $320 price target on Accenture, with the analyst note arguing that "recent M&A activity shows how new entrants NEED incumbent skills + deployment capability to implement." That is the bull case in compressed form: DeployCo cannot scale to thousands of enterprise deployments off 150 Tomoro engineers, and the Big Three integrators will get pulled into the work as subcontractors regardless.

The bear case is that DeployCo did not need to be self-sufficient on day one. It needed to capture the deployment relationship — the contract, the architect-of-record status, the platform-of-record decision — and then subcontract execution at margin to the incumbents whose stock just fell. The same incumbents whose parent firms wrote cheques into the round. The capture is the asset; the execution is fungible.

Worth flagging: OpenAI already has a TCS/Infosys/Cognizant alliance for Codex enterprise rollout, signed in April. The trio that lost market cap on Monday are the same three firms OpenAI was, three weeks ago, calling its global rollout partners. They still are. They are also, now, subcontractors-in-waiting.

What this looks like from OpenAI's seat

DeployCo solves three problems OpenAI was visibly carrying in Q1. First, the Q1 80x growth number Dario was hitting at Anthropic's developer day is matched on the OpenAI side by enterprise demand growing faster than enterprise can absorb it. The bottleneck is not GPT capability; the bottleneck is the customer's ability to wire it in. Second, the integration revenue OpenAI was leaving on the table — by Decoder's estimate, multiples of license revenue — was flowing to the incumbents. Third, the Tomoro/Palantir model gives OpenAI a moat the labs explicitly cannot match: workflow knowledge that lives in the engineers, not the weights. A frontier-lab competitor can match GPT-5.5. It cannot match six months of an FDE sitting inside BBVA's risk team.

The Decoder's line on this — "a moat from workflows no lab can simulate" — is the part that will read best at the eventual IPO. It is also the line that explains why the consultancies wrote the cheques. If the moat is workflow proximity, then sitting outside DeployCo is sitting outside the moat. McKinsey and Bain did the math. Cognizant and Accenture's parent companies were not given the option to.

What to watch

1. Whether the 19th investor list closes or expands. Capgemini, Bain, and McKinsey are in. Accenture, EY, Deloitte, and PwC are not. The next 90 days will tell whether the holdouts buy in at a higher valuation, lobby Brussels into an antitrust review, or build a counter-alliance. The Big Four of consulting have never been on the wrong side of an integration platform of this scale before. None of the three plausible responses is cheap.
2. The Tomoro retention curve. OpenAI is paying for a 150-person bench, but the bench's value depends on the engineers staying. FDE talent at this seniority is the most poachable population in tech. If the public LinkedIn announcements turn from "joining the Deployment Company" in May to "excited to share my next chapter" in August, the moat thesis weakens before the first big customer ships.
3. The S-1 reference. OpenAI is preparing for a public listing in the same window as SpaceX and Anthropic. DeployCo gives the prospectus a second revenue line that does not depend on token-per-second margins. The number that matters in the S-1 will not be DeployCo's headcount; it will be the contracted enterprise integration revenue, against which the IPO will be priced. Wall Street will know within two quarters whether $10 billion was the right valuation or a placeholder. The number was, in any case, chosen the way OpenAI chooses numbers.

DeployCo is the cleanest signal yet that the next phase of the AI build-out will not be won on model benchmarks. It will be won on whose engineers live inside the customer. OpenAI just bought the first hundred and fifty, took $4 billion from the firms that thought they owned that relationship, and let the market price the rest.