The AI Labs Are Consulting Firms Now. The IPO Will Price That.

The OpenAI Deployment Company raised $4 billion from 19 investors at a $10 billion valuation. The instrument promises a 17.5% annual return over five years. That is a fixed cash flow, not equity upside — a debt-like structure. You do not guarantee returns on a software platform with near-zero marginal costs. You guarantee them when you can predict the cash flows of a services operation. That single number is the most honest thing either AI lab has said about what its business has become. ^[1] The operational architecture confirms what the financing implies. The venture employs 150 forward-deployed engineers embedded inside client enterprises. ^[2] Anthropic's parallel effort, "Ode with Anthropic," describes its engineers "sitting down with clinicians and IT staff" to build tools that fit into existing workflows. ^[1] OpenAI has launched domain-specific models — "Spud" for high-value professional work, "GPT-Rosalind" for life sciences — and hired a Chief Revenue Officer from Slack. ^[3] This is the organizational chart of a professional services firm with practice-area specialization, not a horizontal software platform. The ventures' own documents state the goal explicitly: to "eliminate the traditional valuation discount applied to analog-heavy services businesses." ^[2] They know what they are and they know what it costs them. The pivot was not a growth strategy. It was forced by the collapse of the token model. Uber exhausted its entire 2026 AI coding budget in four months. Microsoft cancelled thousands of Claude Code licenses. ^[4] Ninety-five percent of enterprise generative AI projects failed. ^[5] Microsoft's Satya Nadella warned customers are "paying twice" — spending on tokens while surrendering proprietary data — and Palantir's Alex Karp said American businesses are "livid." ^[6] The labs did not choose to become consultants. Their customers rejected the product they were selling. Now both labs have filed confidential IPO paperwork, each seeking roughly $60 billion at target valuations exceeding $1 trillion. ^[7] OpenAI does not expect profitability until 2030. ^[8] The revenue targets require a step-change the current trajectory cannot deliver: Barclays analysts say OpenAI must add 50 million weekly active users and introduce "Agents-as-a-Service" to reach $100 billion in annual revenue by 2027. ^[9] The task-based model is explicitly a required new revenue stream, not an existing one. The IPO valuations rest on a business model that has not yet been proven. The labs are entering a space where incumbents already hold the client relationships and human-capital scale they lack. Accenture acquired Faculty AI for $1 billion, integrating 400 AI professionals and placing Faculty's CEO as CTO. ^[10] Wipro committed $500 million to a "platform-plus-services" AI model. ^[11] ServiceNow's Now Assist is tracking toward $1 billion in annual contract value — capturing task-based AI revenue through software licenses that scale. ^[12] Meanwhile, AI engineers command $900 an hour as consultants, well above the $400 to $600 traditional management consultants charge, because deploying AI is labor-intensive and failure-prone. ^[13] That is the opposite of scalable software margins. Palantir offers the template the labs are chasing. Its commercial AI Platform has become its fastest-growing revenue stream, with adopters like BP, Lowe's, and Lockheed Martin. ^[14] But Palantir's forward-deployed engineers sell software licenses, not billable hours — which is precisely the trick the labs' ventures have not yet demonstrated they can pull off. Enterprises with serious engineering capability can bypass the consulting arms entirely. BNY built its own "Eliza" AI platform with more than 100 digital employees and 120 automated tasks, integrating models from Google, OpenAI, and NVIDIA independently. ^[15] The orchestration platform n8n raised $180 million at a $2.5 billion valuation by letting organizations build AI workflows themselves. ^[16] The implementation ventures may be limited to mid-market clients who lack in-house AI talent — a narrower addressable market than the IPO valuations imply. The 17.5% guaranteed return is the labs' own assessment of what their deployment business is: a predictable-cash-flow services operation. The $1 trillion-plus IPO valuation is what they are telling investors it is worth. Those two numbers cannot both be right, and the IPO is when that contradiction gets priced.