Context: Major regulatory agencies in different jurisdictions have recently taken an interest in strategic partnerships between cloud computing providers and AI companies (December 23, 2023 ai fray article).
What’s new: Yesterday the U.S. Federal Trade Commission announced discovery letters sent to Microsoft, Google and Amazon about the former’s partnership with OpenAI and the two other companies’ relationship with Anthropic (January 25 press release by the FTC). On the same day, the agency held a 4.5-hour webinar named FTC Tech Summit (agency website) to discuss consumer protection and antitrust enforcement in connection with AI.
Direct impact: The discovery letters to the five companies were sent out under FTC Act § 6(b), which means that theoretically the recipients could oppose or seek to narrow them like any discovery request. At this point, what the FTC is doing is more of a reconnaissance mission than actual enforcement, which may or may not follow.
Wider ramifications: The FTC demonstrates through actions and words that it views AI as a high priority during the final part of President Biden’s first term. Other regulatory agencies have also taken, or are contemplating, AI-related initiatives.
It’s key to distinguish between whether the FTC is doing the right thing and whether it is, or will be, doing things right:
- It would be difficult to make a principled argument that the FTC and other regulators should turn a blind eye to major developments in AI. In that regard, the FTC is doing the right thing.
- The agency is also doing the right thing in terms of a 360° angle, looking into a major AI partnership involving each of the three major U.S. cloud hyperscalers.
- One could argue that Google’s acquisition of UK AI company DeepMind (Wikipedia article) should also be on the radar, but quite some time has passed and the deal was cleared at the time. The FTC can’t be blamed for not sending out discovery requests about it, but should bear Google-DeepMind in mind when analyzing other (non-acquisition) deal structures.
Given that the FTC has the authority to perform certain inquiries even without there being any indication of wrongdoing, it’s not acting ultra vires (beyond its powers).
It’s simply too early to take a position on those Sec. 6(b) inquiries. It firstly has to become clearer what the FTC is trying to achieve and on what grounds. It’s ai fray‘s position that partnerships that don’t involve voting rights (or veto rights that have the same effect, such as a veto on executive appointments) can give rise to “merger”-like concerns only if there are some highly unusual circumstances. The reinstatement of OpenAI’s CEO was plausibly due to the fact that virtually the entire team supported him, and without that team, OpenAI couldn’t easily pursue its vision.
It was obviously no coincidence that the announcement and the FTC Tech Summit on AI took place the same day.
The Tech Summit was a well-organized, highly informative event with many competent speakers. The agenda (AI & Chips and Cloud, AI & Data and Models, AI & Consumer Applications) had significant breadth. What was striking, however, is that the speakers were mostly handpicked: people of whom the FTC knew they would advocate and support aggressive and ambitious enforcement. In that regard, the angle was only 180°.
There were vairous situations in which the event came across as highly political and even populist. For instance, they said that AI should benefit the people and not just a few Big Tech companies. There actually isn’t an inherent conflict. It’s not a zero-sum game. The goal from the U.S. perspective should actually be to maximize the benefits of AI for society, for the economy at large, and also for some of America’s largest corporations. An inclusive statement would have been better than a mutually exclusive one.
The argument was made that cloud resources are key to making AI work and that this puts cloud hyperscalers in a pole position. It was also explained that there are only a few companies that make chips, and in some respects Nvidia may constitute a monopoly. The conclusion from all of that was that there is a competition problem because startups can’t possibly compete with cloud hyperscalers and they particularly can’t compete in the chip business when hyperscalers even make their own chipsets and/or buy large quantities of scarce chipsets from Nvidia.
The fact that a certain type of business is resource-intensive limits competition, but that is not the same as a restraint of competition from an antitrust point of view. It’s more a law of nature. A newcomer can’t compete in the Manhattan skyscraper business without billions of funding either.
The FTC is right that (as was stated yesterday) tech markets gravitate toward concentration. It’s also true that chipset scarcity is an issue, and the contestability of the market is complicated by companies making their own chipsets, though the FTC may underestimate to what extent that fact actually serves as a competitive constraint on merchant chipset makers (those who sell their chipsets to others as opposed to making them only for their own use).
Yes, there must be competition, and if necessary, there must be competition enforcement. But from a geopolitical and national sovereignty point of view, it must also be considered that the United States shouldn’t inflict unnecessary harm on some of its most critical companies in the area. A balance must be struck. Sometimes intense competition between three companies can actually be more or less an optimal state of competition, though the FTC’s counterargument here is that high switching costs can result in lock-in (that point was in fact made yesterday).
To be fair, a 4.5-hour conference with a significant breadth and a large number of speakers is not the setting in which every statement can be perfectly nuanced. But the FTC could have opted for a more balanced roster of speakers.
It was impressive to see how committed the FTC is to this effort, which includes the fact they’ve brought a number of technology-savvy officials on board.
There is more to come from the FTC with respect to AI, and ai fray will watch it all with great interest.