The vast majority of the US economy is services-based, with sectors like business/professional services, real estate, and healthcare dominating spend and employment. Small businesses, in particular, employ half of the American workforce and contribute to ~44% of US GDP. In these industries and businesses – what I call the “real economy” – adoption of software, let alone advanced AI, lags significantly. However, these markets represent the biggest opportunity for digital transformation over the coming decades. I believe AI rollups will be a key part of how this transformation takes place.
What is an AI rollup/buyout?
The idea of a rollup or buyout is not new. Typically, they have been orchestrated by private equity firms or enterprising recent business school graduates. The basic idea is to buy up and combine small businesses (often in boring, “real economy” industries like HVAC or laundromats) with a combination of equity and debt. These businesses can typically be bought for 1-3x EBITDA and the combined entities as a result of multiple expansion can be worth 5-10x EBITDA. The base case is all about multiple expansion with upside for those who can make operational improvements that increase margins or growth.
However, with the advent of AI models that can substantially automate human knowledge work, the past 6 months has seen the space flooded with ambitious technologists. The basic idea is that AI – either built in-house or bought from 3rd party vendors – can increase gross margins by orders of magnitude by augmenting and eventually replacing human labor. This is particularly true in services industries driven by knowledge workers (eg. accounting, staffing, managed IT providers). The venture opportunity is to build massive services businesses with software-like gross margins.
Growth buyouts are a related idea, where a scaled software company buys a large potential customer and implements their tools in-house. Metropolis is the canonical example of a successful company in the growth buyout/rollup space; they built incredible computer vision technology to modernize parking lot payments and then went about acquiring parking lots to implement their technology. Today, they own 280+ parking lots across the country and are worth nearly $5B.
Why a rollup versus a vertical software play?
The natural question in this space is: isn’t it better to own just the software vendor? The answer is not obvious. To me, it depends on whether the software solution developed has enough of a moat. In certain markets, where this moat exists, software vendors provide their customers a lot of value and can’t be easily replaced. They can therefore command a high enough ACV to justify the CaC of selling into a fragmented, often resource-constrained buyer base. Over the past decade, there have been a number of vertical software companies that have executed on this exceptionally well, including Mindbody, Toast, and ServiceTitan (which most recently had a successful IPO).
However, in the age of AI, software is increasingly becoming commoditized. The barriers to build high-quality software are rapidly approaching zero. If we believe that the vertical software layer becomes a commodity and pricing is a race to the bottom, then ultimately we’d want to own the customer of the software, not the software provider itself.
What are the characteristics of markets ripe for AI rollups?
There are a few key characteristics of the markets/sectors where AI rollups will be particularly effective. In no particular order, these markets should be:
- Highly fragmented
- Large and growing
- Lacking technology expertise and reluctant to buy software
- High revenue, low margin
- Driven by human knowledge work (human knowledge labor is a majority of OpEx)
- Relationship-based; businesses and their customers have sticky, long-term, trust-based contracts
Additionally, the very best rollups will buy businesses that own unique and valuable datasets that can be used to train models. The very best sectors will be ones where AI-powered companies not only cut costs, but can increase revenue by offering a product or service that is substantially better than their competition. Finally, the ideal sectors to play in are the ones where rollups are not already ubiquitous and therefore acquisitions can still happen at reasonable prices (ideally 1-3x EBITDA).
A great example of a market ripe for rollups, in my opinion, is staffing. US staffing alone is a ~$200B market, estimated to grow 5% YoY. It is highly fragmented with over 23k agencies in the US and only ~200 doing north of $100m in annualized revenue. The market has been slow to adopt software, has high revenues and low margins, and the majority of OpEx is spent on humans in call centers who screen and interview candidates. What makes staffing particularly exciting is how much of an advantage an AI-native staffing agency has over its competitors. The agency can interview candidates 24/7, in any language, with unlimited throughput and zero human error or bias; it’s like lighting up a call center with an unlimited supply of objective, multi-lingual agents working around the clock. This sort of capability would allow the AI-enabled agency to offer a substantially superior service than competitors, at a lower price. I’m certain there are many more markets like this today, and as AI models develop, even more markets will be unlocked.
Why are AI rollups so compelling?
AI rollups have a number of interesting characteristics that I believe make them compelling investment opportunities. First, they offer an opportunity for the most ambitious technologists to be big fish in smaller ponds. By owning and operating “real world” businesses often run by older, less ambitious individuals, entrepreneurs face competitors who are unfamiliar with technology and less ambitious about building generational businesses. Second, these businesses have the potential to be far more capital efficient than traditional venture-backed companies. The businesses being bought are already cash flowing, and perhaps significantly more cash flowing assuming AI works. These cash flows can be used to help fund product development, growth, and further acquisitions. It is very possible that the best entrepreneurs can build $1b+ companies on under $10m in equity capital. Finally, private equity firms are very acquisitive when it comes to cash-flowing businesses; this provides a nice landing ground for the companies who don’t make it to the billion-dollar scale. Unlike with vertical software, where there exists a “make it big or go bust” dynamic, it feels like there is a wider range of outcomes possible.
What do exceptional teams look like in the space?
I believe the founders who build generational businesses through rollup/buyout strategies won’t look too different from the folks who have built large software companies in the past. I believe having a technical background will be very important in understanding how AI can fundamentally transform businesses, how to build the right tooling, and whether to build or buy these tools. I also believe advantage will accrue to young people, who are “AI native,” think differently, and have the ambition and stamina to persevere through the ups and downs of entrepreneurship. Finally, I think domain expertise will be super helpful in picking the right sectors and spaces and ultimately operating these businesses effectively.
There are, however, a few notable differences to how rollup teams will be composed. Great companies will require a Chief Investment Officer – somebody who comes from a private equity background – to help execute these acquisitions. It will be important that this person is an early employee, if not founder, of the company. Additionally, it will be very important that teams are augmented by individuals who come from industry and ideally have deep leadership experience. As an example, I would be much more comfortable backing an accounting rollup composed of a teammate who has actually run an accounting firm before. This combination of a non-technical, experienced, industry expert with a young, AI-native founder, will be critical in building the best companies.
At the end of the day, investing in rollups will be fundamentally about picking great entrepreneurs and technologists, which is why traditional private equity firms will not be well equipped to win in the space.
What are the outstanding questions that remain?
There are a number of outstanding questions that remain, the answers to which will determine whether this will be a lasting asset class or just a fad:
- Can AI actually improve the gross margins of services business by a significant amount?
- How will these software-enabled services businesses be valued on the public markets? Will they have multiples that are closer to SaaS companies or closer to services companies?
- Does there exist a robust enough private credit market to fund these buyouts, especially in a relatively high interest rate environment?
- What is the long-term future of AI rollups? Is this a durable asset class or does AGI suddenly render this whole discussion pointless?
These are all initial thoughts and are subject to evolve over time. Nonetheless, I think we are in a special and exciting moment where AI can be applied to the “real economy” and I can’t wait to see how exceptional entrepreneurs seize this opportunity.