Context: In 2023, some competition watchdogs raised concerns over whether Microsoft’s partnership with OpenAI amounted to a “merger” in terms of Microsoft gaining control in exchange for the substantial resources it provides (December 23, 2023 ai fray article). Among them was the UK Competition & Markets Authority (CMA), which subsequently cleared three other BigTech-AI partnerships: Microsoft-Inflection, Microsoft-Mistral and Amazon-Anthropic (September 27, 2024 ai fray article). While the regulatory agency was contemplating the facts and weighing its procedural options for more than a year, the Microsoft-OpenAI partnership evolved considerably (January 23, 2025 ai fray article). As did the CMA, whose chairman got replaced and which will soon receive a new, even more clearly pro-growth and pro-business, “strategic steer” from His Majesty’s Government, with today being practically the last day to submit comments (March 4, 2025 ai fray article).
What’s new: Today the CMA closed the book on its Microsoft-OpenAI “maybe a merger” type of investigation (CMA web page) after “decid[ing] that Microsoft’s partnership with OpenAI does not qualify for investigation under the merger provisions of the Enterprise Act 2002.” The five-page ruling (PDF) makes reference to “recent developments in the Partnership which reduce OpenAI’s reliance on Microsoft for compute.”
Direct impact: Technically, this was a fast-track decision because the CMA decided at the end of phase 1 that there was no basis to open an in-depth review requiring a phase 2 investigation. Practically, it took the CMA very long to arrive at a rather intuitive outcome. Theoretically, any such partnership could still be investigated, at whatever stage, under non-merger competition rules, and the CMA notes in today’s decision that it reserves the right to identify competition concerns regardless of whether this partnership constitutes a merger. But no such investigation appears imminent.
Wider ramifications: The CMA looked into four Big Tech-AI partnerships and either found they did not meet the criteria for a merger review or at least that there were no issues, an exercise that may not have been the best use of agency and party resources. It failed to meet its government’s expectation of contributing to growth, investment and legal certainty, and to make its decisions reasonably swiftly. If the UK government had to give an example of why the ousted chairman, Marcus Bokkerink, was the wrong man for the job, this AI mergermania, in and of itself, would have been sufficient justification.
It was always known that Microsoft does not have shares or voting rights. There’s merely a mechanism for Microsoft to get some money back.
It’s good news for the UK economy that the CMA is now potentially cleaning up the mess created under the previous CMA chair, especially (but not only) in the merger context.
It’s even better news that the CMA effectively apologizes, in the form of a LinkedIn post by its Executive Director Joel Bramford, for how long this review of an imaginary merger took and promises to improve on four P’s: “Pace, Predictability, Proportionality and Processes.”
Mr. Bramford says “[t]he changing nature of the arrangements between Microsoft and OpenAI during the course of [the] investigation added a further layer of complexity in what is already a rapidly evolving sector.” In other words, the CMA was dealing with a moving target and generally a fast-moving market. But even if there were changes, the question for the CMA was a simple one: had Microsoft become more influential over OpenAI than in 2019 when the partnership started? The answer to that was no at any point in time during the 15-month preliminary investigation.
There are reasons for which policy makers give competition regulators greater powers over mergers than mere partnerships. But if the agencies stretch the envelope with novel and ambitious theories that put a merger label on almost any major deal, they abuse those powers. Laudably, Mr. Bramford’s LinkedIn post recognizes that with those great powers comes great responsibility:
“UK law gives us an unusually broad jurisdiction by international standards, and we have committed to clarifying and delineating our remit (as far as legally possible) by updating our guidance on how we interpret and apply these tests. We will be doing this in June and will include the learnings from this group of AI partnership investigations in our thinking to enhance certainty going forwards.”
The CMA’s “everything’s a merger to me” attitude threatened to have a chilling effect on AI partnerships. Those deals are necessary, however, to reap the benefits of AI. It takes resources, and partnerships are how innovative players like OpenAI obtain what it takes to make the wizardry work.
Prior to Brexit, major technology industry issues were not investigated by the CMA as the European Commission’s Directorate-General for Competition (DG COMP) usually handled them. That created new “opportunities” for the CMA to make headline news in the world’s media, but there is also a learning curve. Depending on how the CMA applies going forward what it learned in recent years, there is hope that it may manage to focus on issues that are of particular relevance to the UK.
Somewhat related to the CMA’s Big Tech-AI “merger” reviews that went nowhere is a cloud market investigation that was referred to the CMA by another UK government agency (Ofcom) and which the CMA now effectively wants to pass on to its new Digital Markets Unit (DMU). It is another example of a protracted regulatory inquiry into a competitive market with no particular UK interests at stake.
The cloud computing and AI ecosystem needs an environment in which it can thrive and contribute to global economic growth. Podcaster Dwarkesh Patel discussed the AI opportunity (and, in that context, also the cloud business) with Microsoft CEO Satya Nadella (YouTube link). To make AI more and more powerful, it’s key for cloud costs to continue to go down. The CMA cloud investigation, however, actually focuses on discounts (for committed spend, but also in connection with software licenses) as if price competition was a bad thing. That will be something for the agency to think about before it makes its final decision.