Early-Stage VC in the Age of Vertical AI

Why founders & investors must reimagine software success

In an essay published in October last year, we discussed the reckoning underway in the seed stage market and how the assembly-line approach to early-stage venture capital was becoming challenged as graduation rates hit an all-time low.

Optimizing fund strategy for graduation rate, however, is a risky proposition that is getting riskier. An unintentional consequence of such an approach can be the “assembly line” approach to early-stage venture capital. A high volume of initial investments requires intense systemization–and often simplification–of the picking process. Many firms rely heavily on thematic strategies that focus dollars into “hot” or fast-growing sectors. So, while conveyor-belt investing may produce impressive short-term performance on paper, it will inevitably underperform without a consistent way to consistently foresee cycles.

An increasing number of Vertical AI success stories, both inside and outside our portfolio, has led us to rethink implications for early-stage investors once again. Historically, vertical platforms have been able to achieve higher penetration rates than their horizontal peers, in large part because the solutions are built to spec for an industry. As a result, switching costs are high with lower competitive intensity. As discussed in our essay last week, AI has also significantly eroded barriers to software adoption across even the most recalcitrant industries.

The early adoption curves of top-tier Vertical AI—some with growth rates that look more like best-in-class consumer or marketplace businesses than B2B software—make it clear that there is something unique to the formula. And if this phenomenon proves replicable throughout the economy, the already massive opportunity in Vertical is set to experience a generational acceleration.

Large language models (LLMs) enable vertical software solutions to engage customers—and often their customers—based on how, where, and when they work. Instead of needing customers to adjust to a new app and user experience (UX), vertical AI solutions can connect directly through voice, text, or industry-standard documentation. These wedge products require significantly less, if any, market education or training. These systems can also integrate and operate within existing systems of record, eliminating the need to overhaul workflows and avoiding high-friction integration processes. The main traditional barrier to adoption that vertical software has encountered—persuading customers to change how they do business—is diminishing rapidly.

In addition to reimagining the architecture and feature prioritization of new platforms, Vertical AI is changing how buyers evaluate new tools. In a world where the burden of implementation and education is decimated and new solutions are seamless to try, the adoption curves (and resulting growth rates) for new vertical B2B applications are looking almost consumer-like in some cases. What this ultimately means, in our view, is an order of magnitude increase and acceleration of the size and scope of the Vertical software opportunity. At the same time, it raises interesting questions for founders and the venture ecosystem.

In sum, the profiles of vertical AI company-building and go-to-market must force investors to reimagine what early-stage success looks like. And we believe this transformation may require more than a few new roadmaps. Much of the AUM and professional experience in the asset class revolves around the cloud software business model. Vertical AI is showing early signs that its top performers have completely different capital needs and cadences.

VCs clutching to check sizes and strategies predicated on historical software capital patterns—inviolate for at least two cycles and now at fundamental risk—may soon be in for a rude surprise.

Vertical AI Startups at Early-Stage

Properly addressing an investment opportunity as monumental as vertical AI requires more from investors than simply looking for deals in the space—as with cloud software in prior generations, holistic investment strategies likely need to evolve. The first step on that journey requires an understanding of the core characteristics of a vertical AI platform and how they might differ. We believe most successful startups within the theme of B2B, industry-specific AI will have most, if not all, of the following differentiated characteristics:

• Lower upfront costs. Fewer employees and less R&D spend required to get a wedge product to market, given their low-UI form factor and greater ability to work around onerous upfront integrations.

• Founder-market fit is even more essential. With lower (initial) barriers to entry comes a heightened need to focus on pain points that win the right to build defensible platforms over time. Teams that have a deep understanding of stakeholder workflows have a strong competitive advantage here.

• Product-led growth focus. Low integration and adoption hurdles enable startups to leverage products as an acquisition channel. Referral buying effects are amplified. Alternative pricing models (including usage or output-based) are a significant accelerant.

• Higher initial growth. Analyzing its Vertical AI portfolio companies, Bessemer Venture Partners found far faster growth than traditional peers (~400% YoY), at similar ACVs (~80%) to SaaS. They predict we’ll see five $100M ARR+ Vertical AI companies within the next 2-3 years, which we believe is conservative.¹ We have seen several Vertical AI businesses such as MagicSchool and Abridge go from zero to $10M+ ARR in <2 years, and anecdotally, many more run to $2M ARR in <1 year (and on very little capital)

• Potential for mid-stage growth plateau. A critical mid-stage juncture and failure point when the company expands from wedge to multi-product. Expansion requires increasing integration, data ingestion, and multi-workflow. Startups without an angle on workflow and data will likely plateau as “feature” companies and struggle without long-term competitive differentiation.

  

• Different competitive dynamics. In some cases, more realistic competition from legacy market leaders. We have seen this effect with some traditional vertical SaaS startups that only achieved hyper-growth after incorporating an LLM aspect, such as CaseText or EliseAI. Some even more established players (think public SaaS) will likely be able to do the same and capture vertical AI opportunities before they begin. However, we believe this most mature bucket will be innovation-constrained—not only because incumbents rely on legacy architecture but also because their multi-product nature makes it harder to reap a core vertical AI advantage: system-of-record agnosticism. The classic Innovator's Dilemma rears its head again: startups can out-specialize and out-execute incumbents with mandates to protect old products and strong cash flows.

• Optimizing growth against capital efficiency & product investments. The potential to leverage early growth into successful fundraising and a strong balance sheet will allow some teams and startups to skip the mid-stage plateau and move to multi-product earlier than competitors. Startups and their VC backers must balance the desire to invest heavily in GTM and growth against the critical R&D investments required to build long-term differentiation. Abridge is an excellent example here. Between anecdotal information and public reporting, it seems safe to say the business will have hit $100M ARR in its first three years of full commercialization.²

• Profitability. With enough initial success and careful management, however, some vertical AI platforms may be able to achieve and maintain cash-flow positivity early. Thus far, the most prominent examples of Vertical AI seem to have taken capital in relatively traditional amounts (if not standard cadences). There is a bit of selection bias here, however—those that raise large rounds are more highly reported and their founders likely less dilution-sensitive. Amongst Vertical AI companies that have raised <$3M, we have seen several that have crossed Series A milestones rapidly (e.g., a few million ARR), are profitable, and do not plan on raising further capital until the growth stage.

• “J-Curve” capital consumption for Vertical AI. Vertical AI businesses still need inception-stage capital. Mega-seeds >$3M, however, may not be necessary. We have seen many examples of Vertical AI startups—including several in the Euclid portfolio—achieving a “2-2-2,” surpassing $2M ARR, on <$2M paid-in, in <2 years. If startups can achieve wedge product hyper-growth, however, their negative cash burn will likely decrease quickly, making the need for a traditional seed obsolete. Capital needs, however, are likely to rebound as the product goes multi-product. We will discuss this J-Curve effect at length below.

Implications for the Venture Ecosystem

What does this all mean for the venture capital investment and allocation ecosystem? Certainly, investors will need to understand and adapt to the new dynamics of Vertical AI success. The changes required in VC underwriting will benefit some strategies and incite growing pains for others. A premium on founder-market fit, for example, has been a hallmark of Euclid’s since founding. We have shared in past essays our focus on backing founders with demonstrated, deep connectivity to their vertical, whether through past vertical software startups or industry roles.

Just as Vertical AI will serve as a fund-maker for many VCs, we believe it may also be destructive to others. Traditional SaaS startups often face increasing CAC as they grow. To some extent, the fundamental nature of competition may make this unavoidable. Combined with the increase in costs, regulatory burden, expectations to IPO, plus a proliferation of late-stage capital over the last two decades, even the most capital efficient (e.g. vertical) software businesses often raised $100M+ in venture capital to reach their “venture-scale exit.”

Breakout Vertical AI businesses, however, are demonstrating the ability to build a different relationship with capital. As introduced above, we believe that the capital requirements for many Vertical AI winners will resemble a “J-Curve,” most acutely affecting Seed and Series A stage capital allocators. If Vertical AI succeeds in demanding less capital between inception and scale, growth-style investing, check sizing, and valuations may begin to creep into phases of startup maturity historically dominated by Series A funds.

Mapping capital requirements to our “Vertical AI” playbook, we believe the resilient phase of investment—but perhaps hardest to capitalize on for most funds—will be the inception stage. While it’s cheaper to get a Vertical AI product to commercial viability, it’s not free—nor straightforward enough to obviate the value of expertise and support in the earliest days. Post-commercialization, the local minimum of the J-Curve will reflect a startup’s ability to grow rapidly without needing heavy investments in customer acquisition. The step-change in user experience, revenue increase, and ROI novel AI wedges that can facilitate are so pronounced that the best of Vertical AI startups can achieve hyper-growth without hyper-capital.

Herein lies the danger for the early-stage venture ecosystem—particularly, the assembly-line approach to investing that has been so prevalent over the past few years. Investors applying the last cycle’s lessons may be left behind altogether. As of Q4’24, the median time from founding to seed stage funding is 22 months, yet the median time from seed to Series A is 25 months.³ We don’t believe these numbers are sustainable or rational for most startups, especially not in Vertical AI. Extrapolating these figures, the “median” founder would be raising three times within the first two years of company life: formation (angel), pre-seed (6+ months), and seed (18+ months)… and they would still be two years away from a Series A.

So what do the earliest stages of VC fundraising look like in the era of Vertical AI? First, we believe the best founding teams will opt to avoid the programming, dilution, and time investment associated with accelerators. Their domain experience and strong networks enable them to sell well in advance of the product. With low upfront costs to launch and scale an initial wedge, moreover, these teams can achieve product-market fit quickly and affordably—potentially skipping the seed stage altogether. Angel syndicates and F&F rounds will always have some value but bring additional complications and lack the resources to support company growth. Referring back to our piece on the seed-stage reckoning, these observations inform our focus on the inception stage at Euclid:

Our conviction in this accelerated start reflects in our strategy—all but one of our investments occurred within ~3 months of company founding. The “n” is small but our conviction is even higher that providing those teams with access to institutional capital and networks from formation enables them to accelerate their path to market, capitalize on established demand, and scale enterprise value creation.

It is possible that early-stage investors—relying on obvious signals regarding traction and market perception—could miss out on a generational opportunity in Vertical AI. Vertical market development is forward-looking, and that trend will only accelerate with AI. What matters is TAM at exit, not at entry. Retrospective approaches will fall short. Logic and history dictate that early-stage funds aiming to invest in Vertical AI successfully must do so from first principles. At the risk of over-quoting ourselves, we articulated this in another recent essay:

Deducing early-stage investment priorities from downstream brand-fund preferences, however, is like stock picking based on last year’s S&P 500 performance….EMs, whose first checks are usually their most significant driver of alpha, should be concerned with what will succeed—and hence attract downstream capital—tomorrow. And a lot changes in one year, let alone three, four, or five years down the line.

In an industry so intensely focused on the future as venture capital, historical perspective is too often lost. Learning from the past, however, is all the more critical in moments of massive potential disruption. We believe this is one of those times. Private capital AUM has grown by an order of magnitude over the last 20 years. On the VC side, the marginal-cost growth potential of cloud software was no small part of that explosion. As underlying technology and business models shift, astute companies and funds should continually re-examine traditional assumptions. The best founders, in Vertical AI and otherwise, have no loyalty to the status quo—investors would be foolish to act differently.

Thanks for reading Euclid Insights! If you know a Vertical AI founder thinking through their next idea, market, or wedge product, we would love to be helpful. Please reach out via LinkedIn, email, or Substack—we look forward to hearing from you.

1. Bennett, Deeter, Droesch, Feinstein, and team (2024). Part I: The Future of AI is Vertical. Bessemer Venture Partners.

2. Clark, Palazzolo (2024). Elad Gil’s Latest AI Bet Is in Health. The Information.

3. Dowd, Neville (2024). State of Private Markets Q4 2024. Carta.