OpenAI announced on Monday that it is launching a dedicated company to embed artificial intelligence deployment engineers inside corporate clients, a move that represents both a strategic pivot and an implicit concession: building the world's most capable AI models does not, by itself, generate the enterprise revenue that OpenAI needs to justify its $852 billion valuation.
The OpenAI Deployment Company, capitalised with more than $4 billion and majority-owned by OpenAI, will place forward-deployed engineers (FDEs) directly inside organisations to identify high-value AI use cases and build production systems around OpenAI's models. The unit launches with approximately 150 engineers acquired through the purchase of Tomoro, an Edinburgh-based AI consulting firm formed in 2023 in alliance with OpenAI that counts Mattel, Red Bull, Tesco and Virgin Atlantic among its clients.
Nineteen firms have committed as partners, led by TPG with Advent, Bain Capital and Brookfield as co-lead founding partners. Goldman Sachs, SoftBank, Warburg Pincus and WCAS are among the founding investors. Consulting and systems integration firms, including McKinsey, Bain & Company and Capgemini, are also participating.
The structure borrows directly from Palantir, the data analytics company that pioneered the FDE model by placing its own engineers inside government agencies and large corporations to build bespoke systems on top of its platform. Palantir's approach solved a problem that every enterprise software company eventually encounters: the gap between what the technology can theoretically do and what a client's existing infrastructure, data and workforce can actually support. OpenAI is now acknowledging that the same gap exists for large language models.
The timing is driven by competition. At a company-wide meeting, Fidji Simo, OpenAI's chief executive of applications, told staff that Anthropic's recent gains in enterprise adoption should be treated as a "wake-up call." Anthropic's Claude has seen rapid uptake among businesses for coding, research and internal workflows, and the company has formed its own private equity-backed venture to acquire AI deployment firms. "We cannot miss this moment because we are distracted by side quests," Simo told employees, a remarkably candid admission from a company that has spent the past year launching consumer products ranging from image generation to voice assistants.
The deployment company plans to acquire additional firms beyond Tomoro, using the $4 billion war chest to consolidate the fragmented market for AI consulting services. That market is projected to grow from $11 billion in 2025 to $91 billion by 2035, and the race to control it is accelerating.
The awkward tension is with the consulting firms that have simultaneously invested in the venture. McKinsey, Bain and Capgemini are among the world's largest advisory practices, and all three are building their own AI deployment capabilities. By joining OpenAI's partnership, they gain early access to the company's models and engineering talent. But they are also funding a competitor that will embed OpenAI engineers inside the same clients they advise, performing work that overlaps directly with what their own consultants sell.
Constellation Research analyst Ray Wang identified this conflict directly, noting that the OpenAI Deployment Company is not analogous to IBM Consulting, which operates independently and integrates technology from multiple vendors. OpenAI's unit will work exclusively with OpenAI models. The chances of it recommending Anthropic's Claude, even when Claude might be the better fit, are nil. Enterprise buyers will view the arrangement through the lens of vendor lock-in, Wang argued, and that perception could limit adoption among companies that prefer to maintain optionality across AI providers.
For OpenAI, the calculation is that the lock-in risk is worth accepting because the alternative, ceding the enterprise deployment market to Anthropic, consulting firms and systems integrators, is worse. If the companies that help businesses implement AI are not aligned with OpenAI, they have no incentive to recommend OpenAI's models over competing products. By owning the deployment layer, OpenAI ensures that every client engagement reinforces its technology stack.
Related reading
- OpenAI closes $122bn funding round at $852bn valuation as IPO pressure builds
- OpenAI strengthens Sora video safety and consent controls
- OpenAI is outmanoeuvring Anthropic on cyber diplomacy, and Europe is the prize
The move also changes the revenue model. Selling API access to GPT-4o or GPT-5.5 generates usage-based revenue that scales with consumption but is vulnerable to price competition as models commoditise. Selling embedded engineering teams generates services revenue that is higher-margin, stickier and harder for competitors to replicate, because the engineers accumulate institutional knowledge about the client's operations that creates switching costs independent of the underlying model.
Whether OpenAI can execute on this strategy while simultaneously managing a consumer products division, an infrastructure buildout spanning multiple continents, a custom chip programme with Broadcom, a $134 billion lawsuit from Elon Musk and a potential initial public offering is another question entirely. Simo's warning about "side quests" may have been directed at her staff, but it applies with equal force to the company itself.
The recap
- OpenAI forms OpenAI Deployment Company with majority ownership.
- Deal brings roughly 150 engineers from Tomoro for immediate deployment.
- TPG leads investors; Advent, Bain Capital and Brookfield co-lead.
