There is a real debate happening right now in the investment world about whether to invest in vertical or horizontal software companies. You hear it at conferences, in partner meetings, in pitch debriefs, and in LP conversations. Both sides have smart people making confident arguments. And honestly, both sides have a point. Many of my friends will have a different perspective on this than me. And that’s ok. That’s why different funds and strategies exist. Otherwise, this whole world should be a single firm making algorithmic investing decisions. But because I’ve gotten this question hundreds of times in the last few months, I started writing my own thoughts down and lo and behold: a blog post…
The Case for Vertical Software and AI
The pitch for vertical software is intuitive. Pick an industry that’s underserved, build something deeply specific to how that industry actually works, and become indispensable. Contractors don’t want generic project management software. They want something that handles lien waivers, subcontractor compliance, and certified payroll. Specialty pharmacies don’t want Salesforce. They want a CRM that knows what a prior authorization is.
The argument is that depth creates defensibility. Deep workflow integration is sticky. Domain-specific data compounds over time, and best of all, since you know exactly who your customer is, what they care about and how to reach them, the sales motion can be much more targeted. This lends itself to focus, and marketing becomes simple: one message, one product, one ICP.
There’s something real here. Vertical SaaS businesses like Veeva, ServiceTitan, and Toast have proven the model. They clearly defined their ICP, they built a product for them, they had non-competitive marketing advantages, and they quickly won customers. In an AI world, the pitch gets stronger: if you have proprietary data from a specific industry, you can train better models, build better automations, and deliver outcomes that a generic tool simply can’t match. Additionally, the unit of work is more defined, so pricing becomes more relatable and topical.
The cons are just as real, though.
Vertical markets are, by definition, smaller. You’re not building for everyone. And in most verticals, the market is fragmented, buyers are slow, and the sales cycles are long. The ceiling on revenue is lower, which means the ceiling on valuation is lower too. That matters a lot when you’re trying to return a fund (especially as fund sizes have scaled).
The defensibility argument also has a soft underbelly. You only have a data moat if you have all the data, or at least enough of it that no one else can replicate your model quality. Vertical startups rarely achieve that. There are always competitors. Definitionally, startups don’t have 100% market share, so another company with another subset of customers will have competitive data and knowledge. And customers almost always use more than one tool. Your “proprietary” data is seldom as exclusive as the pitch deck implies.
The Case for Horizontal Software and AI
Horizontal companies go wide instead of deep. They build capabilities that many industries can use: infrastructure, developer tools, AI models, data platforms, workflow automation, and security. The customer could be a bank, a hospital, a retailer, or a logistics company. The product doesn’t care.
The upside is obvious. The addressable market is enormous. A company that solves a problem for every enterprise isn’t constrained by the size of any single industry. Network effects, if you get them, compound faster with more customers across more use cases. And the brand becomes something more durable: you become a category, not a niche.
Horizontal companies also tend to attract the best engineering talent. The problems are harder. The scale is higher. And the mission feels bigger.
But horizontal has real risks too.
Without a clear initial wedge, you can easily spread too thin too early. “We’re building AI for everyone” is not a go-to-market strategy. Horizontal companies often have longer paths to product-market fit because the customer definition is blurry. They also face competition from much larger players from day one, since big platform companies have incentive to eat general-purpose categories.
Winning horizontally often requires technical moats, more capital, and/or more patience. It is not the right model for every team.
Where I Come Out
I’ve spent a lot of time with both camps, and here is what I believe.
Vertical approaches generally produce smaller outcomes. That’s not always true, but it is true often enough to matter when building an investment portfolio. The math of fund investing requires some percentage of companies to return multiples on the entire fund. Smaller addressable markets make that harder, not easier. And in a world of increased competition - where more startups are built to tackle these same problems - just getting off the ground becomes more challenging. Building the product might be cheaper and easier, but competition is deeper, marketing becomes more expensive as others bid for the same keywords, and your differentiation fades.
I don’t buy the data moat argument for vertical startups. As I said above, the idea that owning workflows or domain-specific data creates defensibility only works if you have dominant market share. Vertical startups, by definition, don’t have that. Someone else in the market has similar data, even if they only get a sliver of the market to start. Someone with more resources can acquire similar data. A startup in a vertical doesn’t have the leverage to maintain exclusivity over information that exists in that industry. It’s a compelling story. It rarely holds up to scrutiny.
Competition is coming faster in the AI era, not slower. One of the defining features of AI-native development is how dramatically it has reduced the cost and time to build a product and acquire early customers. That sounds good for founders. It also means more competitors show up earlier. A vertical startup that spent two years building a niche product now faces a new competitor who built a comparable MVP in a few months - or 20-30 competitors. The first-mover advantage has a shorter window than it used to. And in a world where everyone can become a software developer, so does unique industry insight.
And it’s not just other startups. Large software companies are reorienting their engineering resources toward vertical-specific functionality. Salesforce, ServiceNow, SAP, and dozens of others are building industry clouds with genuine depth. They have distribution, relationships, and infrastructure that a startup can’t easily replicate. Let’s say a 1,000 person engineering org used to focus on a core product. If AI makes these people even 2x as productive, 500 engineers now have nothing to do. Incumbents could fire them all and be more profitable. But will they really risk these 500 people leaving and starting or joining a competitor? In my conversations, many of these people are being reoriented to what was previously perceived as lower priority projects, including point solutions, all built off of a core platform with years of a head start (technically and with customer lock in). Competing against other startups is hard enough. Competing against a large enterprise vendor who has decided your niche matters to them is a different problem entirely.
Horizontal investing at the early stage gives you a different kind of option value. A company with genuinely strong technology and no fixed vertical constraint can find the best market opportunity as it learns. It can double down on the industries where it wins. It can expand into adjacent use cases. The vision doesn’t change, but the path to scale is more flexible. That flexibility is worth something, and I think it’s systematically undervalued.
The only thing that’s really defensible today is technical differentiation. Not product differentiation. Not feature differentiation. Not brand or marketing differentiation. Those can all be copied. What’s genuinely hard to replicate is a core technical capability: a better model, a faster inference architecture, a novel training approach, a proprietary data pipeline that produces something others can’t, a distribution edge. That is what creates a durable competitive advantage. Everything else is a head start, not a moat.
The Bottom Line
I might be wrong about some of this. Venture is probabilistic and the best investors I know have been wrong about big things and still produced great returns. I’m not claiming certainty.
But I invest where I see genuine upside potential and where I believe the company can reach that potential without needing infinite capital to get there. A smaller market with easier competition is not where I find that combination. A company attacking a large and growing problem with a real technical edge is.
Will we invest in vertical B2B companies that can explain their technical edge and why it’s not just a head start but an enduring moat? Of course, but being vertical as a moat is no longer enough.
Founders, there are great investors for you out there if you want them. They come in all shapes and sizes. Nobody is necessarily wrong. They’re all coming at investing from a different angle, with different experiences, and trying to connect the dots through their lens.
That’s my viewpoint. It’s not complicated. There are people much smarter than I who will disagree. But I think it’s the right approach for this moment and welcome the counterpoints.
we are seeing that vertical wins for margins, horizontal wins for velocity. and it's seen in Polsia (single problem, solo), Hyperliquid (one clearing engine, then layers), vs companies trying to be everything.
Good read. I think as Claude Code etc create more for solo founders, there’s another argument for going vertical. Distribution is everything so from a marketing and network building point of view, it helps to close some doors