Why Is Big Tech 'Acquiring' AI Startups Without Actually Buying Them?
Sign up for ARPU: Stay ahead of the curve on tech business trends.
Last week, Meta Platforms Inc. announced a stunning $14.3 billion investment for a 49% stake in Scale AI, one of the key startups providing data-labeling services for training AI models. As part of the deal, Scale’s founder and CEO, 28-year-old Alexandr Wang, will join Meta to work on its new “superintelligence” team.
In a memo to employees and customers, Scale’s new interim CEO, Jason Droege, insisted the company was not being wound down and would remain independent. Yet the move has had immediate consequences. Key customers OpenAI and Google are reportedly ending their relationships with Scale, wary of its new deep ties to a major rival.
The unusual deal structure — a massive investment and talent poach without a full takeover — is the latest and largest in a series of similar arrangements. It begs the question: in the red-hot AI sector, why are some of the biggest deals not acquisitions at all?
What are these new AI deals?
They have been described as “acquisitions in everything but name.” Rather than buying a startup outright, a tech giant makes a large investment, licenses the smaller company’s technology, and, most importantly, hires its founders and key employees. These arrangements, sometimes called “acqui-hires,” provide the buyer with the startup's most valuable assets — its human talent and intellectual property — without the formal process of a merger.
This trend has accelerated over the past year. In March 2024, Microsoft paid $650 million to license software from AI startup Inflection AI and hire most of its team, including co-founder Mustafa Suleyman, who now leads Microsoft’s consumer AI efforts. Soon after, reports emerged of Google negotiating a $2 billion licensing deal with Character.AI to use its technology and hire its co-founders, and of Amazon.com Inc. striking a similar deal with Adept AI.
In each case, the startups were seen as having been “rescued,” with the large payments primarily serving to make their venture capital investors whole while the big tech firm absorbed the core talent.
Why are companies structuring deals this way?
The primary driver appears to be the intense regulatory climate. The Biden administration had increased scrutiny of tech mergers, and the Federal Trade Commission is reportedly probing both the Microsoft/Inflection and Amazon/Adept deals to see if they were structured specifically to avoid government approval. According to a Wall Street Journal report, Google and Character.AI had considered an outright acquisition but concluded it was unlikely to get past regulators.
Beyond antitrust concerns, the structure reflects the unique nature of AI startups, where the value is often concentrated in the people. As Matt Levine of Bloomberg put it, for many of these firms, “the assets take the elevator down every night.” The companies consist mostly of their employees' knowledge, skills, and ideas. A big tech acquirer is often more interested in bringing that brain trust in-house than in owning the startup’s corporate shell, especially if it comes with fewer regulatory headaches.
How intense is the war for AI talent?
The scramble to hire top researchers has escalated to what some describe as professional-athlete levels. A small group of elite individual contributors, perhaps only a few dozen to a thousand people worldwide, are seen as having an outsized ability to advance the frontier of AI. OpenAI CEO Sam Altman alluded to this phenomenon, tweeting about “10,000x engineer/researchers” who are exponentially more effective than their peers.
This scarcity has led to staggering compensation. According to reports, top OpenAI researchers regularly receive packages worth over $10 million a year, and some who considered leaving were offered retention bonuses of $2 million on top of equity increases of $20 million or more. Google DeepMind has reportedly offered certain researchers compensation packages of $20 million per year.
This talent war is why a figure like Scale AI’s Alexandr Wang is so valuable to Meta. At just 28, Wang is known for being extremely well-connected, with close relationships to leaders like Mark Zuckerberg and Sam Altman. For Meta, bringing him and his team on board is a strategic move to bolster its own AI ambitions after its Llama 4 model launch in April was viewed as underwhelming.
What does this mean for the AI startups and their investors?
For some startups struggling to keep pace with the capital-intensive AI race, these deals can be a lifeline. For investors, they offer a lucrative exit. The Meta-Scale deal values Scale at $28 billion and provides a $14.3 billion cash payout to its existing shareholders, which include venture firms like Accel, Index Ventures, and Founders Fund.
However, the strategy is not without risk for the startup left behind. As evidenced by Scale AI, becoming so closely allied with one tech giant can alienate other major customers who are direct competitors. OpenAI, once a major Scale customer, confirmed it has been winding down its work with the company for the past year, citing a need to work with data providers that have kept pace with innovation. The loss of major clients like OpenAI and Google could challenge Scale’s stated goal of operating as a truly independent, model-agnostic company.
Is traditional M&A dead in AI?
Not entirely. The “acqui-hire” model appears to be a preferred route for absorbing foundational research talent. However, for companies with distinct products or strategic assets, traditional acquisitions are still on the table.
OpenAI’s recent $6.5 billion all-stock deal to buy Jony Ive’s hardware startup, io, is a case in point. The move gives OpenAI a dedicated team to build a new family of AI devices. Similarly, OpenAI was reportedly in talks to acquire the AI coding assistant Windsurf for around $3 billion, a deal that would give it a strong foothold in the developer tools market. Salesforce’s recent $8 billion purchase of Informatica was also a strategic move to accelerate its enterprise AI platform, AgentForce.
The key difference seems to be product versus people. When a startup offers a concrete product that fills a strategic gap — be it a hardware device or a developer tool — a full acquisition makes sense. When the prize is a small group of elite researchers, the messier, more flexible “acqui-hire” model has become the new playbook in an industry defined by a frantic race for talent and an increasingly watchful eye from regulators.