Acquirers are increasingly buying what they cannot build. KPMG’s 2026 outlook, drawn from seven hundred senior dealmakers across twenty jurisdictions, puts the acquisition of technological capabilities and talent alongside market expansion as the motivations most cited this year, and reads the cycle as capital concentrating on durable advantage rather than on scale for its own sake. The premium is the price of not waiting three years, and paying it can be the soundest decision on the table.
It is also a measure of the difference between what the target can do and what you can. Which makes it a measure of what you stand to destroy.
The deal closes and the business is handed to an integration function whose purpose, competently discharged, is to make it resemble the parent. Systems consolidated, processes aligned, approval thresholds applied, pay bands harmonised, planning cycles synchronised. Every one of those removes difference, and difference was the purchase. The organisation applying them is the same one that could not produce the capability, which is why it went shopping. So the premium is paid for what the acquirer lacks and then converted, step by diligent step, into more of what it already had.
This is hard to anticipate because executives think of what they bought as an asset. Market share is an asset: it transfers with the entity, and once the contracts novate it is yours. A capability is a group of people, the way they work, the decisions they take without asking, the pace they are used to, and the pricing freedom that let them compete. None of that transfers, because none of it was in the entity. Change the conditions and you do not own a capability. You own the company that used to have one.
Four things then go wrong, usually reported as separate failures when they are one failure in different clothes. The parent’s defaults win. They win through process, when the playbook is applied because applying it is the job and nobody asked which parts of this business it should leave alone. They win through politics, when the acquired firm turns out to do something better and the incumbent method prevails anyway, because people’s standing is bound up in the current way and defending it is rational for them even when it costs the deal. They win through omission, the least discussed and among the most expensive: the business is left to make its own way while inheriting the parent’s economics, and a cost base that sits above the market it sells into will price its proposition out of that market by an internal decision nobody experienced as strategic. And they win while claiming the opposite, when an acquisition is announced internally as a merger. Staff establish what happened within weeks, so the leadership narrative is spent early, and the first to notice are the senior, marketable people whose judgement was the reason for buying.
Underneath all four is a measurement problem, and it is the reason careful organisations watch this happen without seeing it. Everything pointed at a deal after close measures cost, over one year or two. Synergy capture, savings run-rate, systems retired, headcount rationalised, all visible quickly and all reported upward. Capability value arrives later than that and never appears in a cost line. The erosion runs on a longer clock still, because people leave in ones and twos across three or four years and each departure reads as an ordinary resignation rather than as the deal failing. The instruments therefore show a successful integration for as long as they are looking. You will receive a clean integration report and own a dead capability, and the two facts will not meet for years.
The earnout is where this becomes actively harmful, because it is treated as the thing that protects value. It measures the acquired unit standalone for two or three years, which sets its leadership to defending their own numbers rather than building anything combined, makes every conversation about shared clients or shared cost adversarial, and then expires at the precise moment retention risk peaks. Worse, it crowds out the plan. Everyone points at the earnout as the value mechanism, so nobody writes one, and when it ends there is no instrument left aimed at the original ambition and often nobody who remembers what it was. I have watched a capable business handed an earnout, the parent’s rate card and no guide to how the parent actually worked, which is a combination that would defeat almost anyone.
The clearest version of all this is currently happening in public. Technology firms have spent two years buying teams outright: Microsoft paid Inflection $650 million in licensing fees in March 2024 and hired much of the team including Mustafa Suleyman, Amazon took the founders and senior people from Adept and then from Covariant, and Google struck a $2.7 billion licensing deal with Character.AI and hired its founders. Here the capability and the people are the same thing, with no product to hide behind, and the pattern has been visible enough for CNBC to call the results zombie startups, organisations hollowed out once the talent has gone. Whatever one makes of the structures, they settle the general point. When what you bought is a group of people, every integration decision is a retention decision, and retention becomes the investment thesis rather than a workstream running beside it.
So the practical answer. The plan for a scale deal is a list of what will be combined. The plan for a capability deal is mostly a list of what you will not touch, and it has to be written before close, because after close you will not win that argument against a function whose job is to combine things. The list names the specific conditions that made this business able to do what yours could not: its decision rights, its pace, its pricing freedom in its own market, its tools, its hiring bar, whichever the diligence identified as the source of the difference. If the diligence never identified them, that is the finding, and it is better to know before you pay.
Then a second list, shorter, of what you will actively lend them. Access to the parent’s clients, capital and reach is the reason the combination is worth more than the parts, and access does not happen because it was announced. Someone who knows how the parent genuinely works, as distinct from how the org chart says it works, has to be accountable for teaching them to use it. That is a different job from tracking their integration milestones, and it is almost never assigned to anyone.
And then a measure that runs on the right clock. If the capability was worth a premium, name what it should produce in year three and year five, and report it alongside the synergy numbers rather than after they have stopped being collected. None of this means leaving an acquisition alone, which fails for its own reasons. It means deciding, in advance and in writing, which differences are the asset.
There is a reason to settle this now. KPMG has titled its 2026 outlook the year of the carve-out, with half its respondents expecting moderate to significant growth in carve-out activity over the next twelve to twenty-four months and six percent expecting a decline. Some of that is portfolio housekeeping and some is companies selling businesses they absorbed and diminished, which is the market correcting the last cycle in public. So a share of what comes to market this year has already had its conditions changed once, and what looks like a capability at a sensible price may be an organisation that used to have one. Two questions cover both sides of the trade. What exactly makes this able to do the thing we cannot? And which of our own habits will remove it?
References
KPMG International (2026). 2026 Global M&A Outlook: The year of the carve-out. Based on a survey of 700 senior M&A decision-makers across 20 countries and jurisdictions, spanning corporate and private equity organisations across ten sectors, fielded January 2026. https://kpmg.com/xx/en/our-insights/value-creation/global-m-and-a-outlook.html
Computerworld (2025). Meta officially ‘acqui-hires’ Scale AI, will it draw regulator scrutiny? Reports Microsoft’s March 2024 payment of $650 million in licensing fees to Inflection AI and the hiring of much of its team, including co-founders Mustafa Suleyman and Karén Simonyan. https://www.computerworld.com/article/4006994/meta-officially-acqui-hires-scale-ai-will-it-draw-regulator-scrutiny.html
CNBC (2025). Silicon Valley’s AI deals are creating zombie startups: ‘You hollowed out the organization’. 19 August 2025. Reports Amazon’s acquisition of founders and senior talent from Adept in June 2024 and from Covariant two months later, and Google’s $2.7 billion licensing deal with Character.AI alongside the hiring of its founders. https://www.cnbc.com/2025/08/19/how-ai-zombie-deals-work-meta-google.html