AI Is Now a Reason Buyers Walk
For a long time, an AI story was something a seller volunteered to lift the price. That has reversed inside the deal room. In Bain's 2026 survey of more than 300 M&A executives, one in five reported walking away from a deal because of the anticipated impact of AI on the target's business. The same survey found that AI adoption among dealmakers more than doubled to 45 percent, which is precisely why these risks are now being found instead of missed.
The shift is structural, not seasonal. A buyer who can run faster, deeper analysis will ask harder questions about how a company actually makes its margin, and whether that margin survives the next eighteen months of model releases. For owners and family offices on either side of a transaction, AI has moved from a talking point to a line item that diligence is built to scrutinise.
The Four Questions Behind the Price
Serious buyers now test AI exposure along four axes. Model dependency: does the target's gross margin move with a third-party provider's pricing, and could that provider ship the same feature itself. Data moat: is the proprietary data genuinely hard to rebuild, or public data with a thin layer of enrichment. Agentic substitution: can an autonomous agent already do what the target sells, which puts seat-based software most at risk. And talent concentration: if the few people who understand the model leave at close, what is left.
None of these are technical curiosities. Each one is a direct input to the price and the structure. A weak answer on any axis does not just lower the multiple, it changes the shape of the deal, pushing value into earnouts that pay only if the AI thesis is still standing in a year and a half.
What It Costs The Side That Cannot Answer
The numbers are not abstract. Valuation specialists put the multiple compression from regulatory, privacy, and technical AI risk at 15 to 30 percent. On a target carrying real exposure under the EU AI Act, where fines for prohibited practices reach EUR 35 million or 7 percent of global turnover, that discount is not a negotiating tactic, it is a priced fact. The same fine sits on the buyer's books the day after close.
This cuts both ways, and that is the point. An owner preparing to sell who cannot show clean answers on data rights, model governance, and dependency will watch a buyer use the gaps to take the price down. A buyer who skips the work inherits the liability and pays for it later. The advantage belongs to whichever side did the analysis first and can prove it.
Read next: Your AI Must Now Identify Itself · Your Hardware Is Now Regulated Software