Microsoft dodges UK antitrust scrutiny over its Mistral AI stake

Microsoft won’t be facing antitrust scrutiny in the U.K. over its recent investment in French AI startup, Mistral AI, with the country’s Competition and Markets Authority (CMA) on Friday concluding that the partnership “does not qualify for investigation under the merger provisions of the Enterprise Act 2002.”

The decision comes three weeks after the CMA revealed a trio of early-stage probes into Amazon and Microsoft’s various AI investments and partnerships, including the Redmond-based company’s $16 million investment in Mistral AI, an OpenAI rival working on large language models. Shortly after, Microsoft hired the team behind Inflection AI, another OpenAI rival, essentially gutting the startup.

Elsewhere, the CMA said it was also poking at Amazon’s $4 billion investment in Anthropic, a U.S.-based AI company working on large language models.

Big Tech and the quasi-merger

There has been growing scrutiny of Big Tech’s latest tactic to dodge regulatory oversight by pursuing “quasi-mergers,” through which they seek to secure control over new technologies without buying startups outright. This might be through making investments, procuring seats on boards, hiring founding teams, and so on.

Early in 2024, the Federal Trade Commission (FTC) launched investigations into Alphabet, Amazon and Microsoft’s investments in emerging AI firms to establish whether the “partnerships pursued by dominant companies risk distorting innovation and undermining fair competition.”

The CMA’s efforts are part of that same regulatory push. Two of its recently-announced “invitations to comment” are still ongoing, and may lead to formal in-depth probes. Still, it’s telling that the CMA is throwing out the Mistral AI case on the grounds that it doesn’t “qualify” for investigation under existing rules.

Alex Haffner, competition partner at U.K. law firm Fladgate, says this finding suggests that the structure of Microsoft’s partnership with Mistral AI doesn’t grant the bigger company sufficient rights or influence, at least as it relates to M&A regulation. Ultimately, it was a minority investment into a double-unicorn that had closed a $415 million round just a few months earlier.

“In so doing, the decision vindicates Microsoft’s stated position on the tie-up,” Haffner said.

This “stated position” was that making a small investment isn’t enough to procure meaningful clout in the future direction of an up-and-coming AI startup. Microsoft would effectively own less than 1% of Mistral AI when its investment converts to equity at the French startup’s next funding round.

A Microsoft spokesperson said at the time of the CMA’s initial probe announcement:

“We remain confident that common business practices such as the hiring of talent or making a fractional investment in an AI startup promote competition and are not the same as a merger.”

Microsoft spokesperson, April 2024

While the CMA maintains that Big Tech could be adopting new methods to protect themselves from antitrust scrutiny, it has now confirmed that Microsoft hadn’t acquired any “material influence on Mistral AI’s commercial policy.”

“The CMA has considered information submitted by Microsoft and Mistral AI, together with feedback received in response to its invitation to comment,” a CMA spokesperson said. “Based on the evidence, the CMA does not believe that Microsoft has acquired material influence over Mistral AI as a result of the partnership and therefore does not qualify for investigation.”

Pollination works

Just last month, the CMA sounded an alarm over Big Tech’s waxing influence on the advanced AI market, expressing concerns over the growing connection and concentration between developers in the snowballing generative AI space. But the CMA has now said that at least one of the deals on its radar doesn’t qualify for investigation, suggesting that Big Tech’s tactics to pollinate the AI ecosystem far and wide might be working to a degree.

But that still leaves two more outstanding cases: Amazon’s gargantuan investment in Anthropic, and Microsoft’s hiring of key Inflection personnel. Could we expect a similar outcome there?

“The CMA has concluded that the arrangements between Microsoft and Mistral are not sufficient to give Microsoft ‘material influence’ over Mistral, which is the relevant jurisdictional test,” Haffner said. “Time will tell, but the assumption is therefore that the application of the test is more clear-cut here than with the other AI partnerships under investigation by the CMA.”

It’s certainly not as cut-and-dry. Anthropic got Amazon’s biggest venture investment to date, constituting more than half of the $7.6 billion the AI company has raised since its inception three years ago. And while Inflection technically still exists, Microsoft scooped up its founders and various key colleagues — in many ways, that was as good as an acquisition.

And let’s not forget about the CMA’s other separate, but related, ongoing case looking at Microsoft’s close ties with OpenAI. The regulator launched a formal “invitation to comment” aimed at relevant stakeholders in the AI and business spheres last year, and the European Commission (EC) followed suit in January.

So we probably shouldn’t make too many conclusions about the other pending cases based on today’s news.

“That the CMA has only confirmed the conclusions of the Mistral investigation is interesting, as it leaves open the position on the other two deals, as well as the CMA’s ongoing investigation into Microsoft’s role in the Open AI project,” Haffner said. “Overall, therefore, it is clear that the competition authorities are continuing to engage very closely with developments in the AI sector, and we can expect several more announcements by the CMA in the near future as to the outcome of their ongoing workstreams in this space.”