Why Startups Are Rolling Up Old Industries, and Why Two-Thirds Still Fail: PMI in Legacy Sectors, Succession, Debt vs. Equity, Acquisition Criteria, Post-Deal Trouble, and a Pre-Deal Checklist

Hello, this is Ryuta Hamamoto from TIMEWELL.

Since the start of 2026, I have heard the word "roll-up" from founders and investors far more often. In the United States, an acquisition company built by General Catalyst announced a $6.3 billion take-private of a listed company, and OpenAI-backed Thrive Holdings raised $2 billion at a $12 billion valuation. In Japan, Hinoki, which acquires successor-less small companies one after another, raised 3.1 billion yen, and FUNDiT has executed more than 100 acquisitions in five years since founding. It is fair to say that startups are entering an era of buying and combining existing businesses rather than building from zero.

Let me also put an old number on the table first. A study cited by Harvard Business Review found that more than two-thirds of roll-ups created no value for investors.¹ Japan's SME Agency found that 24 percent of small companies that did an acquisition said it "fell short of expectations."² This article gives equal weight to the reasons for the boom and the reasons for failure, then covers how to see PMI through in old industries, what it means as business succession, how to split debt and equity, how to raise money and set acquisition criteria, the trouble that tends to follow closing, and a pre-deal checklist, all in one piece. If you want to check your own team's AI footing before putting AI into an acquired company, use the AI literacy check first.

Why roll-ups are in fashion right now

There are three reasons: more sellers, a hypothesis that AI can raise margins, and a change in how capital is stacked.

The first is supply. Teikoku Databank's 2025 survey of about 270,000 companies nationwide put the rate of companies with no successor at 50.1 percent. It has improved for seven straight years, but half still have no one to take over. In the breakdown of successions, internal promotion at 36.1 percent overtook family succession at 32.3 percent for the first time, and M&A and other routes accounted for 20.6 percent.³ According to Tokyo Shoko Research, 67,210 companies suspended operations or dissolved in 2025, a record for the third straight year; 52.8 percent of them were profitable and 34.0 percent had a president aged 80 or older.⁴ More than 30,000 companies a year disappear while still in the black, for no reason other than having no successor. Recof counted 5,115 M&A deals by Japanese companies in 2025, the first time above 5,000, and 80 percent of them were between domestic companies.⁵

The second is the demand-side hypothesis. In an August 2025 essay, General Catalyst argued that U.S. services are a $6 trillion-plus market left behind by software, took the position that "customers care about outcomes, not products," and laid out a strategy of giving founding teams that have built vertical AI the capital to acquire and operate service companies in that vertical. The target is a "Rule of 60," meaning 10 to 20 percent growth combined with 30 to 40 percent margins.⁶ Long Lake, which it backs, reported a 25 to 30 percent productivity gain in HOA management, acquired more than 30 companies in three years, and in May 2026 announced a $6.3 billion take-private of Amex GBT, the corporate travel manager.⁷ Thrive Holdings has combined more than 50 accounting firms into "Current" and about 20 IT support firms into "Shield," and says its tax AI processed more than 7,000 returns at 98 percent accuracy while cutting preparation time by 30 percent.⁸

The third is how capital is stacked. Older roll-ups often used their own stock as currency on the assumption of a listing, and stopped when the share price fell. Now the dominant form is to raise a block of equity first, then layer borrowing on top to build acquisition capacity. Hinoki has said it will combine 3.1 billion yen of equity with borrowing to create 50 to 100 billion yen of capacity and acquire 20 companies with EBITDA of 100 million to 1 billion yen over ten years. It also positions itself as a permanent holder, with no exit assumed.⁹ FUNDiT's cumulative funding, including bank debt, is 6.2 billion yen, and it has acquired more than 100 small IT businesses in advertising and promotion.¹⁰ Prossell Holdings, founded out of a technical college, acquires small factories and aims to cut design and quoting from a week to about an hour with AI; it made a Niigata ironworks a subsidiary in 2024.¹¹

Half of all sellers without a successor, a hypothesis that AI lifts margins, equity stacked with debt. With all three in place, the roll-up became a fashion.

Why roll-ups still don't succeed that often

Having the reasons for a boom in place is not the same as succeeding.

HBR's 2008 article "Seven Ways to Fail Big" analyzed 750 of the largest business failures of the previous 25 years and counted roll-ups among seven failure strategies. The study it cites found more than two-thirds of roll-ups created no value for investors.¹ The survey in the SME Agency's PMI Guidelines found that 24 percent of companies said their acquisition fell short, for these reasons: no synergies, 44.7 percent; the target's management and organization were weak, 36.8 percent; the target's employees were dissatisfied, 28.9 percent; the price was too high, 23.7 percent; the cultures were hard to merge, 22.8 percent.² U.S. search funds show the same shape. Stanford's 2026 study found an aggregate IRR of 33.9 percent and ROI of 4.75x across 862 funds, but only 58 percent went on to acquire a company, and the returns are concentrated in the top funds.¹² The average looks good; the median is a different picture.

Why does this happen? I think there are four structures.

First, relying on multiple arbitrage alone. Buy a small company at three times EBITDA, have the combined group valued at eight times, and value appears without doing anything. This is the original appeal of the roll-up, but you do not control the exit multiple. Rates rise, the listing market cools, buyers vanish, and the spread disappears. Multiple arbitrage is a design in which only the debt remains the moment the spread is gone.

Second, integration costs rise with the number of deals. HBR's point is simple arithmetic: integrating five companies takes close to five times the work of integrating one company five times the size.¹ Accounting systems, pay structures, customer contracts all differ by company. By the time you buy the tenth, the first is not yet integrated.

Third, AI-driven margin improvement is still a hypothesis. The numbers from General Catalyst and Thrive Holdings are numbers they publish as their own success stories. Nobody has yet proven they reproduce across industries. If you price AI gains into what you pay, the buyer eats the entire loss when the gains do not show up.

Fourth, acquisitions become the goal. From the moment the money is raised, "the next deal" becomes the top priority. More deals look like a track record and are easy to report to investors. Integration is unglamorous and hard to report. That asymmetry pushes integration to later. The SME Agency's PMI Guidelines open by saying that closing an acquisition is "merely the starting line," because the government knows this asymmetry too.²

The roll-ups that lose are not the ones bad at buying. They are the ones so good at buying that integration cannot keep up.

How to see PMI through in an old industry

So what do you do, in what order, after buying a company in an old industry? The SME Agency's PMI Guidelines for small and medium-sized companies are the most systematic answer Japan has. Here are the key points in my words.²

The timeline has three stages: "pre-PMI," gathering information with integration in mind from the deal stage; the "intensive period" of roughly one year after closing; and "post-PMI" thereafter. In particular, establishing the integration team, building trust with stakeholders, and understanding the current state are to be done intensively within the first 100 days after closing.

The most important thing in the first 100 days is people. The guidelines call for communicating the deal "without delay, to all employees of the acquired company, simultaneously, equally, and accurately," and, before that, for informing key people, those with outsized influence on operations and other employees, ahead of everyone else and consulting them throughout the process. They also say that even where change is needed, the buyer should "not deny the acquired company's employees' existing work or ways of working, and respect them." The failure case they present is a buyer that rolled out its own "common sense" one item after another right after closing, lost the cooperation of employees, and found even running the existing business difficult.

What I want to add is the order in which AI goes in. Most startups that call themselves roll-ups assume they will put AI into acquired companies to raise margins. But the profit of a company in an old industry usually depends on one or two craftsmen and a handful of long-standing customers. The judgment inside that craftsman's head is written nowhere. Drop your AI tool in there first and you become the guidelines' failure case. The order is: first write down the tacit knowledge on the shop floor and turn it into data. How drawings are read, how the basis of a quote is set, how exceptions are judged. Only after that is in a form AI can read can automation be discussed. Prossell Holdings choosing "design and quoting" as its first target is, I think, the right order.¹¹ This "making shop-floor knowledge AI-ready" is the same story as Japan's path to winning that I wrote about in the physical AI article.

That requires the person running integration to sit next to the shop floor. Not someone who writes the integration plan in a conference room, but someone who goes to the acquired company's office several days a week, watches the craftsman build a quote, feeds it to the AI on the spot, and reads the reaction. This is the way of working we call the FDE style, described in What is an FDE. The PMI of a roll-up is decided not by the quality of the integration plan but by the number of people sitting on the shop floor.

What it means to do this as business succession

Reduce the roll-up to "buy cheap, sell dear" and you fall straight into the four failure structures. I think the reason to do this in Japan lies on the succession side.

The numbers again: of the 67,210 companies that closed or dissolved in 2025, 52.8 percent were profitable and 34.0 percent had a president aged 80 or older.⁴ Profitable, skilled, with customers, and still gone because nobody would take over. What disappears is not only the company. The supply chain of the customers who relied on its parts, the jobs in that region, and the craftsmen's skills disappear with it. The roll-up is, for now, one of the few mechanisms that can stop this "profitable closure."

It also means something different to the seller. A fund typically holds for five to seven years and then sells again. Hinoki's explicit stance as a permanent holder that does not assume a fund-style exit is an answer to what sellers care about most: what happens to my company and my people afterward.⁹ I think this is exactly where Japanese succession-driven roll-ups should differ from American AI roll-ups. Design the business to work even when multiple arbitrage disappears, that is, so that the acquired company's cash flow and integration gains alone service the debt and fund growth, and you no longer depend on an exit.

Policy is pushing in the same direction. The SME Agency continues to run its business succession and M&A subsidy, with the fifteenth round's awards and the sixteenth round's guidelines published in September 2026.¹³ The SME M&A Guidelines reached their third edition in August 2024 and the registration system for M&A advisers is taking shape.¹⁴ Design the roll-up in the context of succession, and both money and institutions become available.

Splitting debt and equity, and how to raise the money

Now the money. A roll-up is funded by two kinds of capital, equity and debt, and the rule for splitting them is a single one. The part that the target's stable cash flow can repay is debt; investments that do not yet produce cash are equity.

Debt covers the part of the purchase price that the target's operating cash flow over the past several years can service even in a worst case. In Japan, the Japan Finance Corporation's business succession, consolidation, and revitalization loan is available. Its SME unit lends directly up to 1.44 billion yen, with terms of up to 20 years for equipment and 10 years for working capital, each with a grace period of up to five years. When the funds are used to buy shares, including goodwill, without collateral, a rate reduction applies, and the program can be combined with the exemption from personal guarantees.¹⁵ The consumer unit offers a separate 72 million yen. On top of this, the usual structure combines co-lending from private banks, a seller note in which the seller receives part of the price later, and an earn-out paid according to performance.

Equity covers integration, AI, and people: turning the acquired company's shop-floor knowledge into data, a shared accounting and ordering platform, hiring the people who run integration. None of these produce cash for one to two years. Fund them with debt and repayment comes first and integration stops. Hinoki raising 3.1 billion yen of equity first and then layering borrowing is, in my reading, a way to keep this order.⁹

Raising money splits into two approaches: deal by deal, or as a platform up front. Deal by deal is easier for the first company, but every subsequent deal means persuading investors again, and deals slip away in the meantime. Raising up front means explaining to investors what criteria, what kinds of companies, and how many, before any deal exists. Whether you have criteria you can explain at that stage is, in fact, what largely decides a roll-up's outcome.

Here are my rules of thumb. Cap debt at the level where repayment still holds if the target's operating cash flow falls 30 percent. Set aside 10 to 20 percent of the purchase price, separately, in equity for integration costs. And do not sign the second company's final agreement until the first company's integration has stabilized past day 100. They may look hard to keep, but most roll-ups that failed did so after crossing one of these three lines.

Acquisition criteria, and the trouble that follows closing

I think five criteria are enough to explain to investors.

Criterion	Question	Fails when
Purpose	Can you say in one sentence what changes in the group with this deal	"It's cheap," "the deal came to us"
Dependence	Which customers and whose skills does the profit depend on	Top three customers are 70 percent of sales; one craftsman writes every quote
Reproducibility	Has the margin gain from your AI or system been reproduced elsewhere	Applying the number from deal one to a deal two in a different industry
Debt service	Does repayment hold on worst-case cash flow	Modeled on profit at acquisition continuing unchanged
Integration person	Is the person who will sit on this company's floor several days a week named	"The integration team will handle it"

A deal that fails even one is not approved by lowering the price. The SME Agency's finding that "the price was too high" was cited by 23.7 percent is, I think, the result of deciding on price rather than criteria.²

The trouble after closing follows set patterns. The third edition of the SME M&A Guidelines states that disputes over non-performance of final agreements are occurring and specifically points to buyers that were supposed to take over the seller's personal guarantee and then did not, asking advisers to help exclude such buyers from the market.¹⁴ In other words, a startup buyer can unintentionally look like one of these "inappropriate buyers." The handling of personal guarantees has to be agreed with the banks and put in writing before the final agreement.

The other typical patterns are these. Skipping due diligence and inheriting off-balance-sheet or contingent liabilities; a share purchase is simpler procedurally, but because the legal entity transfers as is, this risk is higher.² Key people hearing the news late and resigning the month after closing. Paying the seller's retirement allowance up front while intending to keep them on as an adviser, and having the retirement disallowed for tax purposes. Bringing every accounting and approval rule of the buyer in from day one and stopping the shop floor. Being late to greet major customers and losing them to competitors. Every one of these is a failure pattern in the guidelines, and every one can be prevented by the design of the first 100 days.

Pre-deal checklist

Finally, a checklist to fill in yourselves before signing the final agreement. If three or more of the 20 items are blank, delaying the signing by a month is cheaper.

Area	Item
Purpose	1. Wrote in one sentence what changes in the group with this deal
Purpose	2. Defined "success one year from now" in numbers
Target	3. Confirmed the top five customers and the conditions under which their contracts continue
Target	4. Named the craftsmen and staff who carry the profit and discussed retention terms with them directly
Target	5. Ran financial and legal due diligence and confirmed off-balance-sheet and contingent liabilities
Target	6. Agreed the handling of the seller's personal guarantee with the banks and put it in writing
Target	7. Wrote down why a share purchase or an asset purchase was chosen, in terms of risk isolation and licenses
Price	8. Built the price on a base case with no AI gains included
Price	9. Confirmed debt service holds with operating cash flow down 30 percent
Price	10. Estimated integration costs separately from the price and secured them in equity
Funding	11. Can explain the debt-to-equity ratio and the reason for it to investors
Funding	12. Seller note and earn-out terms do not contradict the seller's retention
Integration	13. Named the integration lead who will sit on the floor several days a week
Integration	14. Set the dates for advance disclosure to key people and simultaneous disclosure to all employees
Integration	15. Wrote what will and will not be done in the first 100 days
Integration	16. Separated the buyer's rules to bring in on day one from those not to bring in for a year
Integration	17. Decided the order and owner for greeting major customers
AI	18. Chose one business process to put AI into first and assigned someone to write down its tacit knowledge
AI	19. Can explain to the seller and employees where shop-floor data is stored and whose training it is used for
Next	20. Decided not to sign the second company's final agreement until the first has stabilized past day 100

Summary

Roll-ups are in fashion because three things came together: the supply of a 50.1 percent no-successor rate and profitable closures, the hypothesis that AI raises margins, and equity stacked with debt. More than two-thirds still create no value because four structures grow stronger, not weaker, in a boom: dependence on multiple arbitrage, integration costs that rise with each deal, unproven AI gains, and the asymmetry that makes acquisitions the goal.

The answer for seeing it through is unglamorous. Define closing as the starting line, spend the first 100 days on people and trust, do not bring in the buyer's common sense, put AI in only after shop-floor tacit knowledge is data. Cap debt at what worst-case cash flow can repay, fund integration and people with equity, and do not sign the second deal until the first is stable. Five acquisition criteria: purpose, dependence, reproducibility, debt service, and a named integration person.

And the reason to do this in Japan is to keep the more than 30,000 profitable companies a year that would otherwise vanish, along with their skills and supply chains. Design it to work even when multiple arbitrage disappears, and you do not need an exit, and you can promise the seller what comes after.

If you are a founder planning a roll-up, or a corporate new-business team considering M&A in a legacy industry, let us start with the integration design of your first company. Turning an acquired company's shop-floor knowledge into a form AI can read, and developing the people who sit on the floor, are what the WARP programs cover. Whether you have a deal in hand or are still building your criteria, let's talk.

Footnotes

1. Seven Ways to Fail Big (Paul B. Carroll and Chunka Mui, Harvard Business Review, September 2008). Analysis of 750 of the largest business failures of the previous 25 years; cites a study finding more than two-thirds of roll-ups created no value for investors and notes that five integrations mean roughly five times the integration difficulty ↩ ↩² ↩³

2. SME PMI Guidelines (Japan SME Agency, March 2022, Japanese). The 24 percent below-expectation figure (n=475), its reasons (n=114), the pre-PMI / 100-day / intensive-period framework, advance disclosure to key people, and the failure case are from the guidelines. A summary is also available ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

3. National Survey on Successor Absence Rates 2025 (Teikoku Databank, November 21, 2025, Japanese). About 270,000 companies; 50.1 percent without a successor; internal promotion 36.1 percent, family succession 32.3 percent, M&A and other 20.6 percent ↩

4. Business suspensions and dissolutions in 2025 hit a record 67,200 (Tokyo Shoko Research, January 9, 2026, Japanese). 67,210 cases, up 7.2 percent; 52.8 percent profitable; presidents aged 60 or older 90.6 percent, 80 or older 34.0 percent ↩ ↩²

5. Japanese M&A in 2025 in review (Recof Data, MARR Online, January 5, 2026, Japanese). 5,115 deals, up 8.8 percent; 35.7 trillion yen; 4,086 domestic-to-domestic deals ↩

6. The Future of Services (General Catalyst, Marc Bhargava and Kate Bender, August 28, 2025). The "Rule of 60" and Long Lake's 25 to 30 percent productivity gain are from this essay ↩

7. General Catalyst's $6.3B AMEX deal puts its AI roll-up strategy on display (PitchBook, May 7, 2026; syndicated on Yahoo Finance). Long Lake founded in 2023, 30 acquisitions; $1.5 billion Creation fund raised in 2024 ↩

8. OpenAI-backed Thrive Holdings raises $2B to bring AI to the enterprise (TechCrunch, August 12, 2026). $12 billion valuation, led by SoftBank, D1, and Altimeter; the firm counts and TaxAI figures are the company's own ↩

9. Hinoki raises 3.1 billion yen for serial business succession (Hinoki Inc., August 21, 2026, Japanese). Targets with EBITDA of 100 million to 1 billion yen; 50 to 100 billion yen of capacity with borrowing; 20 companies in ten years; permanent holding ↩ ↩² ↩³

10. FUNDiT closes Series D, cumulative funding including debt tops 6 billion yen (FUNDiT Inc., May 11, 2026, Japanese). 6.2 billion yen cumulative; more than 100 acquisitions since founding in November 2021 ↩

11. Prossell Holdings, a technical-college-born AI roll-up of small factories, closes a seed round led by Chiba Dojo Fund (Prossell Holdings, August 25, 2026, Japanese). 250 million yen cumulative including debt; Echigo Tekkosho acquired in 2024 ↩ ↩²

12. 2026 Search Fund Study: Selected Observations (Stanford GSB, Peter Kelly, Stefanos Zenios, Dom Ng, 2026). 862 funds, aggregate IRR 33.9 percent, ROI 4.75x, acquisition rate 58 percent, median time to acquisition about 20 months. Summarized in Stanford GSB Insights (July 13, 2026) ↩

13. Business succession (Japan SME Agency, Japanese). Fifteenth-round awards (September 11, 2026) and sixteenth-round guidelines (September 4, 2026) of the business succession and M&A subsidy ↩

14. SME M&A Guidelines, third edition (Japan SME Agency, August 2024, Japanese). Non-transfer of personal guarantees, exclusion of inappropriate buyers, and the risk of skipping due diligence are from the guidelines ↩ ↩²

15. Business succession, consolidation, and revitalization loan (Japan Finance Corporation, Japanese). Direct lending up to 1.44 billion yen and the rate reduction for uncollateralized share purchases including goodwill are from the SME unit page ↩