VCs go vertical in backing specialized AI
Generative AI was the dominant theme of venture capital in 2023 and will likely continue to be this year. However, investment in the sector is likely to become more specialized.
A record $29.1 billion was invested across 691 generative AI deals last year, according to PitchBook data. This represents a 268.4% increase in deal value over 2022’s total.
Much of the focus among investors has been on providers of large language models, including OpenAI and Anthropic. But as the market matures, attention is shifting toward applications within specific industry verticals, also known as vertical AI.
“We’re seeing a lot of activity in (vertical AI) right now,” Northzone principal Molly Alter said. “Because of the increase in sophistication of (large language models), more applications are being built that leverage those advancements in a way that is tailored to specific industries.”
Vertical AI is concerned with more targeted applications of AI that are tailored to the unique needs and challenges of a sector. This is unlike so-called horizontal AI, which provides more general-purpose solutions.
According to Pender Ventures partner Isaac Souweine, it is becoming increasingly hard for AI companies to build to compete at the foundational modal level, especially at the foundational model layer, given the competition from the likes of Google or OpenAI. Building a large language model or generalized product requires significant amounts of capital and time; this isn’t feasible for many companies and investors.
“On the horizontal side, it’s an arms race,” Souweine said. “It’s insanely expensive, and a lot of investors don’t have the capital to place these huge bets. Vertical markets are easier to play in. You can still invest in the AI revolution, but you don’t have to come with a massive mega-fund to do it.”
In contrast to foundational models, vertical AI projects typically involve smaller datasets and therefore need less computation, cutting down costs and time. By focusing on one sector, they can produce more accurate solutions and a faster return on investment for businesses.
Customer acquisition tends to be more straightforward for vertical AI companies given that their products are built with a specific segment in mind, Souweine said. Combined with lower costs, vertical AI has a higher potential for profitability, which can be more appealing to investors as they have shifted away from hypergrowth in the VC downturn.
Although generative AI can have widespread applications, it has the potential to be more disruptive in some sectors than others. Financial services is a prime target for vertical AI, according to Lux Capital partner Grace Isford, due to the vast amount of data involved.
But adoption in the sector has been slow thus far, Isford said, “even though it has one of the largest pools of interesting data and a history of implementing and adopting machine learning. I would expect to see far more [large language models] live in production crawling and classifying financial transaction and merchant data.”
Healthcare is another area of interest for VCs looking to break into vertical AI, Alter believes. The space has already seen several substantial rounds including Hippocratic AI’s $50 million seed investment and Corti’s $60 million Series B.
Analyzing large amounts of medical data, personalizing treatments and identifying potential drug candidates are key parts of the sector that can be streamlined through the use of new technology.
Knowledge is power
While vertical AI has clear advantages for VCs, rushing into the space could be unwise.
“Even if a vertical is ripe for disruption, just as we’ve seen with vertical SaaS it takes time for adoption of new technologies,” Isford said. “Certain vertical AI sales cycles may take longer to grow, and grow at a slower rate than sufficient to justify a venture-backed business.”
A natural consequence of being more specialized is that the total addressable market is smaller. The number of potential customers vertical AI startups operating in a niche sector can acquire is much lower than for a horizontal AI startup. Vertical AI companies are thus less likely to reach the outsized valuations seen for foundational models.
Alter said that knowledge of generative AI can be less important for VC than sector expertise when it comes to vertical AI. Without an understanding of how a specific market operates, investors lack the capacity to distinguish between vertical AI startups that are solving real problems and those that are offering more generic solutions.
“One thing to remember is that [vertical AI] requires a lot of grunt work,” Alter said. “It’s not enough to just understand how foundational models work if you’re looking at investing in different markets. It requires a lot of time spent within those industries and understanding what their needs are.”
Nevertheless, VCs are likely to see a lot more competition for deals in the vertical AI space as more investors seek to capitalize on the AI boom. But with typically smaller growth prospects and a requirement for specialized knowledge, the rise of vertical AI will likely come with less hype.