OpenAI launched a $4 billion enterprise deployment company on 11 May. Among the nineteen investors: McKinsey, Bain, and Capgemini — the three consultancies whose AI transformation practices the venture is designed to compress. One week earlier, Anthropic announced a $1.5 billion deal of the same architecture with Blackstone, Goldman Sachs, and Hellman & Friedman. Two enormous bets, eight days apart, both built on the same conclusion: the deployment layer is the next moat in enterprise AI, and the labs intend to own it.
Focus On: The Deal Structure That Quietly Redefined Neutrality
Read the OpenAI deal terms carefully and the more interesting story emerges. The 19 external backers are guaranteed a minimum 17.5% annual return over five years, with profits capped — a fixed-yield instrument structured more like a credit fund than a venture stake. Those same backers commit their portfolio companies to the pipeline. Collectively, they sponsor over 2,000 businesses, pre-sold into the engagement queue before a single client conversation begins.
McKinsey, Bain, and Capgemini have written cheques into a company that intends to redesign workflows inside large enterprises. That is the consulting industry’s most defensible product line. And the return is guaranteed regardless of which AI vendor a McKinsey partner recommends to a Fortune 500 board next quarter.
Here’s the thing — “consultancies betting on their own disruption” is the polite framing. The actual mechanism is that advisory neutrality on AI vendor selection has just been priced out of existence. If your strategic consultant holds equity in OpenAI’s deployment vehicle and a fiduciary interest in that investment’s guaranteed return, the recommendation engine sitting behind their advice no longer optimises for your outcome. It optimises for the cap table.
The Palantir Precedent
The strategy elucidates itself once you understand who pioneered it. Until 2016, Palantir had more forward-deployed engineers than software engineers. The model was considered too expensive, too weird, impossible to scale. Five-year stock return: 640%. Anthropic and OpenAI have copied the playbook openly. Their engineers embed inside client organisations, identify use cases, build production systems, and stay until the workflow runs. The services collapse into the platform over time — the inverse of the traditional consulting pyramid, where services compound infinitely because the platform never quite arrives.
The Market Has Already Priced It
The Nifty IT index sits more than 40% below its December 2024 peak. TCS posted its first annual dollar revenue decline since its 2004 IPO. Infosys touched its lowest level since December 2020 on the day OpenAI’s announcement landed. McKinsey itself has cut roughly 5,000 jobs over the past 18 months — the largest reduction in the firm’s nearly 100-year history. The 95% pilot-failure rate MIT identified across enterprise GenAI projects is precisely the gap the FDE model is engineered to close. The consultancies recognise that. That is why they wrote the cheques.
What This Means For Any C-Suite Buying AI
The implications are quite succinct. The “trusted advisor” model in AI transformation, as it has functioned for the past three decades, has changed shape this month. Three demands follow.
First, every transformation advisor in your rolodex should disclose, in writing, every equity position, joint venture, and revenue-share agreement they hold with any AI lab or deployment company. If the answer is opaque, the answer is no.
Second, vendor evaluation and implementation engagement should be structurally separated. The same firm should not score the candidates and then deliver the winner — that conflict was always uncomfortable; it is now economically explicit.
Third, treat the FDE relationship as a strategic dependency, not a procurement line item. Once Claude or GPT is wired into your operating model by the lab’s own engineers, your switching cost is no longer technical. It is organisational. I made this point in a different form in the CLI versus MCP issue two months ago, and the mechanics are identical here: the deployment layer is the moat, and whoever holds the moat sets the terms.
The week of 4–11 May 2026 will read, in time, as the moment enterprise AI consulting stopped being a market and became a cap table. The firms that financed their own compression are not silly — they are calculating that participation beats irrelevance. That calculation may prove correct for the firms, but it is not the calculation a CMO or CFO buying AI services needs to make.
Follow me
That’s all for this week. To keep up with the latest in generative AI and its relevance to your digital transformation programs, follow me on LinkedIn or subscribe to this newsletter.
Disclaimer: The views and opinions expressed in Chronicles of Change and on my social media accounts are my own and do not necessarily reflect the official policy or position of S&P Global.