In-depth reporting and analytical commentary on artificial intelligence regulation. No legal advice.

FTC releases Staff Report on AI Partnerships & Investments: collection of information and innuendo with unclear contours

Context: Today is the last day in office for the Federal Trade Commission (FTC) of the United States prior to the transition of power on Monday. Approximately a year ago, FTC announced it was going to take a closer look at partnerships between three major cloud service providers (CSPs) and two AI technology makers: Amazon, Google, Microsoft, OpenAI and Anthropic (January 26, 2024 ai fray article).

What’s new: Today the FTC released the Staff Report on AI Partnerships & Investments that had been a year in the making (January 17, 2025 news release by the FTC). Its publication was approved unanimously, though the two Republican commissioners, one of whom will become President Donald Trump’s FTC chair, filed statements by which they distanced themselves from any insinuations that there were reasons at this point for regulators to be concerned about anticompetitive conduct. The Republicans did, however, support the idea of putting out information that parts of the general public as well as certain government agencies may find interesting.

Direct impact: There are no bombshell revelations in the report, nor does it raise any concrete issues. It describes a wide range of business terms found in commercial agreements, with some terms apparently having reported with respect to only one of the three partnerships at issue (Amazon-Anthropic, Google-Anthropic, Microsoft-OpenAI). In order to identify a problem, one would have to know at least the entirety of the terms agreed upon between two companies. There will soon be a third Republican commissioner, at which point it is unlikely that a majority of the commissioners would vote in favor of “merger” investigations or similar measures based on the report’s findings.

Wider ramifications: It cannot be ruled out that the FTC’s Democratic majority (soon minority) primarily wanted to release this report now in order to support its allies in other regulatory agencies around the world in political and PR terms.

The Staff Report released today is one of those documents that are best understood by starting with what is not contained in it:

  • There is no single business term that the report reveals and that would amount to illegal conduct (such as price-fixing).
  • There is no indication of a particular partnership agreement containing a combination of business terms that would raise such concerns.
  • There is no evidence of any of the agreements having stifled competition or innovation. Much to the contrary, the report itself notes what vast resources are required to operate AI services, and it does not propose any realistically available option that would allow AI technology developers to secure those resources on more favorable terms. Put differently, without those partnerships there would be less competition among technology makers, and as a result, less innovation, than in a but-for world where such partnerships would not be formed and only a few companies would have the ability to develop and operate AI services at all.

Here are some examples of business terms discussed in the report:

  • There is a suggestion that “circular spending” could raise concerns. What is meant by this is a deal structure under which one party (here, a large cloud service provider) invests money in the other, but the investment target has to spend some of the money on the investor’s products or services. The staff report describes this as “one avenue through which CSP partners may potentially aim to reduce the magnitude of potential loss invested directly into their partners and cloud infrastructure to serve their partners.” At any rate, such deals were very popular in the late 1990s during the “dotcom boom” where strategic investments were used in part to increase the revenues of a larger player, increasing its company value by whatever multiple (price-earnings ratio) the market applied. Smart investors were always able to factor that in. If anything, it would be a financial regulation topic, but it has nothing to do with chilling competition.
  • The FTC discusses not only “exclusivity rights” but also “parity” clauses under which a technology maker has to offer its latest and greatest technology on the partner’s platform. But parity is not exclusivity as there can still be any number of others offering the same. In connection with Amazon’s relationships with third-party resellers, parity may restrict competition, but that is because Amazon bars those partners from offering lower prices elsewhere.
  • The report discusses “significant equity and certain revenue-sharing rights.” In absolute terms, the billion-dollar amounts that have been invested by large technology companies in AI technology providers may appear very “significant,” but the report cannot show examples of Big Tech actually controlling entities like OpenAI or Anthropic. As for revenue-sharing rights, it is unclear why a competition watchdog would be worried about them at all. What else could Microsoft have agreed upon with OpenAI? The acquisition of a large ownership position could have triggered merger reviews. Revenue-sharing means that the investor is enabled to provide resources without acquiring a major ownership position. Like in so many contexts, the report does not outline a viable alternative that would still enable certain AI services to be developed and operated. In a fictitious world, it would be greated if Big Tech could selflessly “donate” to the likes of OpenAI, but in the world in which we live, there must be some way for investors to recuperate their investments, and revenue-sharing is actually the structure most favorable to the target.

There are hardly any specific recommendations in the report. Where the FTC is correct is that making AI software available on open source terms (meaning that software can be distributed free of charge, and its source code can be inspected and modified) may not be enough: the data resulting from training may also be needed, which is why the FTC mentions the idea of “open-weights foundation models” that it already discussed before (July 10, 2024 FTC article).

The staff report makes it sound like AI chips developed by large technology companies for their own use are inherently problematic because they will not be sold to others. But at this point there is a severe shortage of AI chips, and the greater the extent to which certain large companies can use their own chips, the more chips made by a company like Nvidia will become to others.

All in all, the FTC Staff Report is more of a collection of bits and pieces than a coherent description of clear and present dangers to competition and innovation. In some antitrust contexts the problems can be downplayed by discussing each element of a greater scheme in isolation. Here, what the staff report tries to do is to portray presumably unharmful deal structures as potentially anticompetitive by taking individual contract terms out of context and suggesting between the lines that they could give rise to concern.

The partial dissent statements filed by the two Republican commissioners, Andrew Ferguson and Melissa Holyoak, give sound advice: one may find the staff report as a whole informative, but should ignore its Section 5, which suggests that competition and innovation may be in jeopardy. And if the concern is about startups, it may be best to defer to America’s leading startup investors Marc Andreessen and Ben Horowitz (November 4, 2024 ai fray article).