Context: Two weeks ago, it became known that the U.S. Federal Trade Commission (FTC), the UK Competition & Markets Authority (CMA) and the European Commission (EC) are trying to find out whether Microsoft’s partnership with OpenAI amounts to an acquisition of control that would be reviewable like a conventional merger.
What’s new: While there have been no further developments since those reports, something will have to be decided in the months ahead. ai fray would now like to comment on the timing of those preliminary investigations.
Direct impact: At this point it appears unlikely that there is a reasonable legal basis for any or all of those antitrust authorities to launch formal merger investigations. If anything, the recent turmoil at OpenAI, which ended with CEO Sam Altman staying at the helm, actually showed that Microsoft passed on an opportunity to effectively acquire invaluable expertise.
Wider ramifications: Competition watchdogs are demonstrating that they will look into partnership agreements that are not formal M&A transactions in order to prevent an end run around merger control procedures. However, should any formal investigations be launched without a shred of evidence that Microsoft acquired control over OpenAI, regulators may appear to be driven by publicity more so than by actual merits.
It’s been almost a year since Microsoft announced (on January 23, 2023) the third phase of, and what observers are sure constitutes its largest investment in, its partnership with OpenAI that was originally formed in 2019.
OpenAI is neither a run-of-the-mill non-profit nor a startup. It has raised funds from investors, but unlike traditional equity deals, where investors can generate theoretically unlimited profits by selling their shares to a strategic acquirer or on the stock market, OpenAI has a “capped-profit” model, which means that there is a ceiling for each investor’s return. The potential return on investment (ROI) is not the same: apparently early-stage investors have a far greater ROI opportunity than later-stage ones.
All that the public has learned about the deal suggests that Microsoft does not have any control over OpenAI’s decisions. It has a contract and can seek to benefit from the partnership, and it may get its money back with a capped return. That is fundamentally different from the rights that traditional venture investors get, such as formal approval requirements that give investors de facto control even if they mathematically hold only a minority of a company’s shares.
If competition regulators had decided in early 2023 that a strategic alliance in the most important field of technology of our times and reportedly involving a payment of well over $10B deserved a closer look regardless of not being a traditional share deal, that would have been understandable. If watchdogs simply contented themselves with a broad claim by the relevant entities that no voting rights are involved, companies could try to circumvent merger control procedures through creatively-structured agreements.
It is an undeniable fact that Company A could effectively control Company B without owning a single share, much less a single voting share. For example, it could grant a loan that comes with a variety of approval rights, which in the aggregate might mean that Company A runs Company B. A deal structure like that wouldn’t generally enable an end run around merger reviews: in the end, an acquisition target’s shareholders want to get paid. But there could be outlier scenarios in which an acquisition of control occurs without a traditional purchase of shares. Besides a loan, what could also think of Company A controlling a critical and unreplaceable input that Company B depends on.
Regulatory inquiries would have been even more understandable is if something had happened that suggests Microsoft effectively gained control over OpenAI without a straightforward M&A transaction. For instance, imagine a world in which more and more Microsoft executives would have joined OpenAI’s management team, possibly even in a dual role where they’d still have retained their positions at Microsoft. One could also think of a situation in which OpenAI would have agreed to amendments to an existing partnership agreement that were clearly unfavorable to OpenAI and not justified by any objective need (such as a deteriorating market environment). In other words, OpenAI would have accepted sub-market terms, despite better options having been legally and commercially available. Only a de facto subsidiary would do that at the behest of its parent company, but not a rational independent entity.
Counterintuitively, what got antitrust authorities interested at this point should have given them comfort rather than pause.
The sequence of events
- OpenAI’s (meanwhile restructured) board fired CEO Sam Altman.
- Almost the entire team stood by Mr. Altman. They were going to work for him at OpenAI or else.
- There was the possibility of Mr. Altman and those loyal to him ending up somewhere else (creating a startup of their own or being hired by the likes of Google), so Microsoft gave him a safe new home to pursue his technological vision.
- Once OpenAI came to reason, Microsoft did not stand in the way and Mr. Altman was able to return immediately.
- The restructuring of the board included a non-voting (i.e., observer) seat for someone to be picked by Microsoft.
Neither the outcome nor what led to it suggests in the slightest that Microsoft has OpenAI in its pocket.
If Microsoft had had the power to get OpenAI’s board to fire Mr. Altman, the last company he’d have been prepared to join would have been… Microsoft.
If Microsoft had had the power to make OpenAI’s board backtrack, they’d have done so directly instead of first hiring Mr. Altman and then letting him go.
If Microsoft had wanted to take advantage of the recent situation surrounding Mr. Altman, they’d have persuaded him to stay, and it was clear that the team was going to work for him no matter where, no matter what. Antitrust regulators can hardly prevent a company from hiring talent that is on the market, much less if they go against non-poach agreements and non-compete clauses that limit employees’ choices.
Throughout that situation, Microsoft was not in the driver’s seat. Even the hiring decision was purely reactive. It was not a poaching scheme. Otherwise they’d never have enabled Mr. Altman to return to OpenAI.
The powerful player here was Mr. Altman, and even his power mostly depended on the loyalty of his team. He had apparently earned his team’s trust; the previous board of directors had not. Microsoft trusted him, too. He trusted Microsoft. That is not the kind of trust that “trustbusters” (as those antitrust agencies are sometimes referred to) need to worry about.
In the CMA’s case, one is left to wonder why the outright acquisition of the UK’s crown jewel in AI by Google was approved at the time, but a partnership that does not involve voting rights (just an opportunity to be repaid) should now draw scrutiny:
Oddly, it’s conceivable that Google, either directly or through lobbying fronts, sought to instigate these investigations just to harm a rival. Google also opposed, behind the scenes, Microsoft’s acquisition of video game maker Activision Blizzard (a deal that the FTC is still challenging in a process followed by ai fray‘s sibling site games fray).
ai fray will comment on future regulatory developments concerning Microsoft’s partnership with OpenAI as well as any major (and actual) M&A activity in this field, though other AI regulation topics will likely be far more relevant.