The Next AI Winners Won't Build Models, Says Sapphire Ventures

By Bloomberg Technology

Categories: Startup, VC, AI

Summary

The AI winners won't be model builders—they'll be vertical AI companies solving specific industries end-to-end. One portfolio company in residential real estate crossed $200M revenue by automating workflows, not just digitizing them, proving that 'grittiness' in complex industries is actually a competitive moat.

Key Takeaways

  1. Vertical AI companies tap into the labor and services market ($200B+ in residential real estate alone), not just historical software spend. This TAM expansion is what makes vertical AI fundable, versus traditional software solutions targeting the same sectors.
  2. Foundation model companies can't execute the last mile: integrating into legacy systems unchanged for decades, deploying forward-deployed engineers on-site, and covering every edge case. This creates opportunity for specialized vertical startups.
  3. Target 'gritty' industries—operationally complex with many edge cases. Most founders avoid them, but that complexity becomes your moat. Example: Elicee covers 1 in 6 US apartment buildings by solving residential real estate's entire workflow.
  4. The shift from 'race to intelligence' (2023-24) to 'true deployment' (now) fundamentally changes what startups should build. Success now requires domain expertise, measurable ROI, and workflow integration—not better models.
  5. Open weight models strengthen vertical AI companies by giving them access to commoditized intelligence. Foundation model companies supporting open models benefit from wider ecosystem adoption (more customers for them and Nvidia).

Related topics

Transcript Excerpt

That's where you focus. Vertical eye. You know, I reflect that over the last year or more. At the very early stage, we've seen insane seed rounds of Neo labs. Yes. Still intense news. Uh, headline domination of anthropic and OpenAI. You think there's a lot more out there? So we're a very interesting point in time right now where we've shifted from the first stages of enterprise AI adoption, where it was all about the race to intelligence. Who has the better models? Let's find the use cases. And now we're entering an era of true deployment, right? Right. So fitting these solutions into everyday workflows, into existing systems, figuring out precisely what the use cases are. And of course, having measurable ROI. This shift dramatically favors, I would argue, vertical AI companies. These are …

More from Bloomberg Technology