Context: Amazon’s $4B investment in AI provider Anthropic, known for its Claude LLM that has most recently made news because of its new ability to perform mouse movements on a computer (October 22, 2024 Ars Technica article), was the last of three AI partnerships to be cleared by the UK Competition & Markets Authority (CMA) after some sort of merger inquiry (September 27, 2024 ai fray article).
What’s new: Today the CMA announced (PDF), somewhat surprisingly, the launch of a Phase 1 merger inquiry into last year’s $2B deal between Google parent Alphabet and Anthropic. The CMA claims that, after a consultation launched in late July, it has enough information for today’s procedural step. The Phase 1 deadline is December 19, 2024. At the end of the Phase 1 investigation launched today, the deal will either be deemed cleared or referred to an in-depth probe (Phase 2).
Direct impact: Given that Amazon invested twice as much in Anthropic, a competition law issue arising from Google’s deal with the Claude maker is unlikely. There would have to be something unusually anticompetitive (and unknown in the industry) in that agreement to raise competition concerns, and even then the deal would still not be a merger, but (at worst) an anticompetitive agreement.
Wider ramifications: It’s possible that the CMA’s primary objectives here are to (i) reinforce its expansive (if not simply borderless) definition of a “merger” and (ii) demonstrate that even relatively old AI partnerships can still give rise to merger investigations. They may be doing this (again, unless there is something outrageous that is not publicly known about the Google-Anthropic deal) in preparation of an announcement concerning Microsoft’s partnership with OpenAI.
În late 2024, the CMA is now launching a merger review of an AI partnership announced in early 2023. And even after clearing Amazon’s twice-as-large investment in the same company. That is puzzling.
Of course, it is theoretically possible that some whistleblower provided information to the CMA that suggests Google’s deal with Anthropic has an anticompetitive dimension. It’s also possible they found out about it as part of their inquiry into Amazon-Anthropic. But how likely is it that an AI provider strong enough to partner with two Big Tech companies (Google and Amazon) at the same time is really coming under the control of one of them? It’s extremely hard to imagine.
The CMA hasn’t said anything substantive about the merger. They made a call for input, and today they provided a commencement notice. No specifics are known.
There may be something that the CMA knows and almost everyone else doesn’t. It could also be that this is performative and just an interim step toward an ultra-high-profile inquiry into Microsoft’s partnership with OpenAI, which is also quite old and far from a merger.
Anthropic is drawing ever more interest by regulators, but also has to defend itself against copyright infringement accusations (August 20, 2024 ai fray article). It’s become too successful, and is talked about too much, to fly under the radar.