On March 25, 2026, Granola raised a $125 million Series C led by Index Ventures, with Kleiner Perkins joining existing investors Lightspeed Venture Partners and Spark Capital, valuing the company at $1.5 billion (TechCrunch). The round arrived with a rebrand: Granola dropped “personal meeting notetaker” from its own description and now positions itself as an enterprise AI context product. That word swap, from notetaker to context product, is the real story in this raise, more than the size of the check.
Ten months earlier, in May 2025, Granola raised a $43 million Series B at a $250 million valuation and launched collaborative features (TechCrunch). A 6x valuation increase in under a year happens when investors decide the product being priced was the corpus building up behind every transcript, not the transcript itself.
The asset was never the transcript
Every meeting a company runs produces a record of decisions, objections, and commitments that used to disappear the moment the call ended. Note-taking tools started by capturing that record for one person's convenience. What has changed is what happens to the record afterward: it becomes searchable, linkable across meetings, and increasingly the material other AI agents draw on to draft a follow-up, brief a new hire, or reconstruct why a decision got made months earlier. The product moved from writing a meeting down to remembering it, so something else in the company's stack can use it later.
Otter.ai's numbers tell the same story from a different angle. The company crossed $100 million in annual recurring revenue with a team of fewer than 200 employees and more than 35 million users worldwide (Otter.ai). A nine-figure revenue business run by a couple hundred people only works if the product compounds as a data asset instead of scaling as a service. Otter co-founder and CEO Sam Liang put it directly:
"Our $100M ARR milestone validates that businesses are ready to embrace AI agents that augment human intelligence in meaningful ways." (Otter.ai)
Fireflies.ai pushes the same pattern further. The company crossed a $1 billion valuation through a tender offer that let early employees sell shares, without raising new primary capital since its 2021 Series A (Fireflies.ai). It has raised $19.1 million in total funding across three rounds and says it has been profitable since 2023 (Getlatka). A company does not reach unicorn status on $19.1 million of outside capital, serving more than 20 million users and 500,000 plus organizations including 75% of Fortune 500 companies (Fireflies.ai), unless what it sells compounds without proportional headcount or continuous reinvestment, the profile of infrastructure rather than a utility app.
The market is pricing this as infrastructure
The category is expanding to match. The global AI note-taking market was sized at $623.5 million in 2025 and is projected to reach roughly $3.48 billion by 2035, an 18.75% compound annual growth rate (Precedence Research). A market growing that fast is being priced on what gets built downstream of the record, search, retrieval, agent context, and institutional memory that survives employee turnover, far more than on how well any single tool transcribes a call.
The adoption gap exposes the real friction
Adoption data shows where the tension actually sits. Small businesses use AI meeting notetakers at a 78 to 81% rate, versus 43% at enterprises with 5,000 or more employees, and 73% of businesses cite privacy as the primary barrier to wider adoption (Fellow.ai). That gap is the direct consequence of the shift described above, not a rough edge a better interface can smooth over. Once a vendor captures every negotiation, every performance conversation, and every strategy call as a durable, queryable asset, the question stops being whether the tool saves time and becomes who can see the record and what happens to it later. Enterprises adopt slowest because they have the most exposure if that question goes unanswered.
For any founder whose team already runs meetings through Granola, Otter, Fireflies, or a competitor, the adoption decision was likely made as a productivity purchase. Treat it as an infrastructure decision instead, starting this week:
- Pull the vendor's data retention and usage terms and confirm in writing whether transcripts or derived summaries can be used to train models beyond your account.
- Decide explicitly which meetings (board, legal, compensation) should never touch the tool at all.
- Assign an owner for the corpus itself: who can query it, export it, or delete it when someone leaves.
The valuations above were built on the premise that conversations compound into an asset. Make sure your company decided who owns that asset before the vendor's cap table did.