Repeat buyers push AI-to-AI startup acquisitions 14% past last year's total

Deals up 14% against buyers up 2% means the average acquirer has already done about 12% more deals than it did across all of 2025 ^[17], with roughly a quarter of the calendar still to run ^[1]. Working back from the 14%, last year ended with about 171 such deals ^[15].

The buying is concentrated in a short list. The seven repeat buyers Crunchbase names for this year account for 32 of the 195 deals ^[16], about one in six: OpenAI with 10 ^[2], Anthropic and Legora with five each, Harvey with four, Sierra and Cursor with three each, and Cohere with two ^[3]. Over the three years Crunchbase tracked, 67 repeat buyers did about 42% of all transactions ^[4].

Legora's finance chief describes the purchases as buying time. "M&A is explicitly part of how we accelerate what we're building. The question we always ask is: does this deal get us somewhere faster than we'd get there ourselves?" Legora CFO David Eckstein wrote in a LinkedIn post earlier this year ^[9]. In legal tech, that has meant buying research, regulatory monitoring and litigation tools instead of writing them, then folding them into broader platforms ^[10].

Rama Sekhar, a partner at Menlo Ventures, which has backed both Anthropic and Legora ^[13], described the seller's side. "If you're not growing 10x, you're not interesting to growth investors, which leaves a gap in the funding market for AI startups that need a home," Sekhar said ^[11]. In his account, stock is how the buyers pay: "High valuations have also given AI startups cheap currency to use their stock to get these deals done with minimal dilution" ^[12]. A buyer that pays in shares marked at its last private round gives up a small slice of itself. The seller ends up holding stock whose value depends on the buyer's next round.

How much stock or cash has changed hands is largely unknown. Prices were disclosed for only 12 of the 195 deals, and Crunchbase said that makes total spending hard to gauge ^[8]. The two largest were Nscale's reported $1.65 billion purchase of Anyscale and Cyera's $1 billion deal for Oasis Security ^[6]. Anthropic's $400 million for Coefficient Bio, OpenAI's $300 million for Glass Imaging and Sword Health's up-to-$285 million for Kaia Health complete the top five ^[7]. Together the five come to about $3.6 billion ^[18]. That leaves 183 deals with no public price ^[19], some of them acqui-hires such as OpenAI's February deal involving OpenClaw and its creator, Peter Steinberger ^[14].

The count covers only buyers that are themselves venture-backed AI companies ^[1]. It shows more AI startups selling to other startups. It cannot show the odds that any given small company ends that way, because sales to incumbents and shutdowns sit outside it. I'd expect the same names to keep buying for as long as the funding gap Sekhar describes persists and their own stock stays expensive. That view fails in two ways. A markdown in the acquirers' private valuations would shrink the currency, and the deal count should fall with it. Or a wave of first-time buyers would pull the 2% toward the 14% ^[5]. A fourth-quarter count in which buyers grow faster than deals would prove the concentration reading wrong.