M&A activity in the AI sector itself was a key element of the global M&A resurgence in 2025. Talent Wars: Acquihires at Scale Many headline-making deals were acquisitions by large platforms at premium prices to secure a small number of elite researchers, engineers, and founders. Regulators in the U.S. and Europe have scrutinized these types of transactions, especially those involving big tech AI transactions. For buyers, that means careful antitrust and foreign investment analysis, even for “non-control” structures. Buying the Plumbing: Data Centers, Chips, and the AI Infrastructure Race If the front page of AI M&A in 2025 was about models and founders, the back page was all about infrastructure. Hyperscalers, miners-turnedcloud-providers, and financial sponsors raced to secure and scale the physical backbone of AI: power, land, data centers, and specialized hardware. Alongside data centers, the semiconductor ecosystem consolidated around AI. Chipmakers used acquisitions to move “up the stack” into systems, networking, and software, positioning themselves as end-to-end AI infrastructure providers rather than component vendors. For deal lawyers, these transactions raise issues more familiar from energy and infrastructure deals: longdated offtake commitments, complex landlord-tenant and power-purchase arrangements, local permitting and community opposition, and environmental scrutiny around energy and water use. The New Deal Toolkit: Hybrid Structures 2025 AI dealmaking reshaped the toolkit itself. Much of the most consequential AI activity is not structured as traditional M&A at all: ▪ Minority stakes and strategic investments paired with long-term cloud and chip commitments (for example, large platform investments into model providers or infrastructure partners tied to multi-year compute and capacity agreements). ▪ Licensing-plus-hiring deals, in which the acquirer brings over key teams and licenses core models or tooling, while leaving the target as a separately capitalized entity (as with Windsurf, Character. AI, and several other generative AI startups). ▪ Joint ventures and structured capacity arrangements in data centers and GPUs that are more like project-financed infrastructure than corporate control transactions (for example, the strategic investments in SB Energy by SoftBank Group and OpenAI, as part of The Stargate Project). These structures are often a response to high valuations, regulatory constraints, and the reality that AI assets are volatile and hard to value on traditional metrics. Earn-outs, milestonebased consideration, and performance-linked covenants (for example, around model quality, uptime, or safety) are increasingly used to bridge gaps between AI “hype” and demonstrated cash flows. 2026 Taken together, AI M&A in 2025 was less about buying operating companies in the traditional sense and more about securing talent and infrastructure. The deals that mattered most were often hybrids—part acquihire, part infrastructure financing, part strategic partnership. For buyers and sellers, that means more creativity in structuring, more complexity in diligence, and a more crowded regulatory field—but also a wider set of levers through which to get AI transactions done. These trends – and no doubt additional structuring innovations reflecting the rapid changes in the technology and business – will characterize AI M&A in 2026. 4. AI M&A: Talent, Infrastructure, and the New Deal Toolkit Earn-outs, milestonebased consideration, and performancelinked covenants … are increasingly used to bridge gaps between AI ‘hype’ and demonstrated cash flows. 46 Morrison Foerster
RkJQdWJsaXNoZXIy NTU5OTQ5