AI roll-ups are emerging as a new M&A play, and Thrive Holdings is the clearest example

On Aug. 12, Thrive Holdings, led by Joshua Kushner, said it had raised more than $2 billion at a $12 billion valuation. That brought total funding for the company, which is just over a year old, to more than $3 billion. Thrive is not a general-purpose AI model builder, nor is it an AI software company like Cursor or Lovable. Its core business is something far more familiar to private equity: acquisitions. Over the past year, Thrive has been buying accounting firms and IT services companies, and it now owns and operates more than 70 businesses. The play is not simply to fold those businesses into a standard holding company. After each acquisition, Thrive pushes AI into the operating workflows themselves, redesigning tasks such as tax preparation and technical support. OpenAI took an equity stake in Thrive Holdings late last year and sent research, product and engineering staff to work on the portfolio companies’ AI transformation. That approach is increasingly being described as an AI-enabled roll-up. The logic is simple: rather than build a product and then persuade accounting firms, law firms or property managers to buy it, investors acquire the firms first and use AI to change how the work gets done. Traditional roll-ups rely on scale, cost control and financial engineering. The AI version keeps the acquisition structure, but changes what happens after the deal closes. Backers argue that AI can turn professional services businesses into higher-margin, more efficient platforms by cutting the link between revenue growth and headcount growth. Thrive’s accounting platform, Current, shows how that can work in practice. As of June, Current included nearly 30 independent U.S. accounting firms, employed more than 2,000 people and generated more than $500 million in annual revenue. The firms kept their local brands and management teams, while technology, capital and part of the back office were centralized. During this year’s tax season, Current’s Tax AI handled about 7,000 tax returns. According to Current, firms in the pilot program cut tax-preparation time by 31% on average, and some data-entry tasks reached accuracy rates as high as 98%. Dwelly, a UK real estate company, has pushed the model in a more direct way. Instead of selling property-management software to local agents, it has been buying independent rental agencies in the UK and moving their work onto its AI platform. In July, Dwelly raised $170 million in Series B financing, including $95 million in equity and $75 million in debt, and said one of the main uses of the money would be more acquisitions. At the time, it managed more than 15,000 properties. The company says a property manager typically oversees about 100 homes under the old model, but after AI took over large parts of tenant support, maintenance scheduling, contract issues and day-to-day communication, that figure rose to more than 300. Each acquisition also brings data that is hard to buy any other way: landlord contacts, tenant message histories, repair-vendor information, property records and the operating habits embedded in the business. General Catalyst has framed this as acquisition as the fastest form of customer acquisition. It has backed companies including Dwelly, Titan MSP, Crescendo, Eudia and Long Lake, and in 2026 it plans to expand its related Creation Fund to $1.5 billion from $800 million. The model is attracting capital because it tries to convert AI-driven productivity gains into a new growth formula for service businesses that have long depended on headcount expansion. A software company can usually add customers without adding many employees. An accounting firm or IT services provider typically cannot. If AI allows the same team to serve more clients, margins can rise without a proportional increase in labor costs. Even a modest change can matter: if a $100 million revenue company with a 15% EBITDA margin lifts that margin to 20%, EBITDA rises from $15 million to $20 million. With the same multiple, enterprise value would be up by one-third. That optimism is already showing up in deal activity. In May, Long Lake, one of General Catalyst’s earliest AI roll-up bets, agreed to buy American Express Global Business Travel for $6.3 billion in cash. The plan is to embed its Nexus AI system into the business. Beacon Software raised $225 million in June after acquiring more than 30 vertical software companies, while Bending Spoons has spent years buying mature digital products and turning long-term M&A into a growth model of its own. The problem is that productivity gains are easier to describe than to fully monetize. Current says tax preparation time fell 31%, but that does not automatically mean labor costs can fall at the same pace. Dwelly says one property manager can oversee more homes, but that does not mean every extra dollar of revenue becomes profit. Professional services still involve review, client communication, legal liability and work that is difficult to standardize. That is where valuation risk enters. If investors begin to assume that AI can materially improve margins after an acquisition, they may pay up for those businesses today, effectively pricing in future operating improvements before they are realized. If the AI transformation underperforms, or if more capital drives acquisition multiples even higher, returns could be squeezed rather than expanded. Thrive Holdings is the clearest example of both the promise and the risk. It has raised more than $3 billion, is valued at $12 billion and operates more than 70 businesses. Yet it has not disclosed group revenue, EBITDA or free cash flow, and it has not said how much profit AI has added across the portfolio. The market can see that Current has cut tax-prep time and that more of Thrive’s IT platforms are using AI. What it cannot yet see is whether those operating gains are enough to justify a $12 billion valuation. That gap between visible efficiency gains and unproven financial outcomes is what makes AI roll-ups one of the most important, and most fragile, new structures in private markets.