This year's "most extraordinary" unicorn: ten months, valuation increased sixfold

Ask AI · How does Granola’s invisible design crack the red ocean track dilemma?

Introduction

THECAPITAL

The Red Ocean Track, How to Compete

This article is 4,263 words, about 6 minutes

Author | Lü Jingzhi Editor | Wuren

Source | #Rongzhong Finance and Economics

(ID: thecapital)

An unexpected unicorn has emerged in the AI conference recording track.

This company is called Granola, headquartered in London, with two founders, less than three years old. The team is working on a well-known red ocean: helping you record meetings. Otter.ai, Fireflies, Microsoft, Google… this track is not short of brands, funding, or users. When Granola entered the scene, few outside thought there was a new story to tell here.

Recently, Granola announced it had completed a new round of financing, raising $125 million, with a valuation of $1.5 billion, becoming a new unicorn. Just ten months after the previous Series B round, its valuation has increased sixfold. The lead investor is Index Ventures, with Kleiner Perkins participating, and previous shareholders also followed on.

Supporting all this is a small product-level decision: not to deploy a bot into meetings. Other tools invite a digital assistant into your video conference, visible to everyone. Granola, on the other hand, does the recording locally on the user’s computer, with no participation from the other side. This single decision opened the door for lawyers, investors, executives—professional groups that previously had no suitable tools—and word-of-mouth spread organically within the circle, without any advertising spend.

Granola’s story, on the surface, is a funding event; at its core, it’s a reflection on “finding the right friction point.” The AI meeting recording track itself is evolving from a simple transcription tool toward deeper enterprise data infrastructure. This upgrade has only just begun.

******** An “Invisible” Decision ********

AI, once again, produces a unicorn.

Recently, a London-based company called Granola announced it had completed a $125 million Series C funding, with a valuation of $1.5 billion.

This figure isn’t unusual; AI unicorns have been emerging in batches over the past two years. What makes people pause and look more closely is its timeline: just ten months since the last Series B; and its valuation has increased six times since then.

What Granola does, three years ago, would hardly warrant a separate project: AI meeting recording.

This track has never lacked players. Otter.ai has been operating since 2016, with over 16 million users; Fireflies is valued over $1 billion; Microsoft has integrated meeting AI directly into Teams and Copilot; Google has similar features in Workspace. There are countless tools capable of transcribing meetings into text. Logically, a two-person team entering this track in 2023 would have missed the window.

But the two founders of Granola, Chris Pedregal and Sam Stephenson, saw not just the “meeting recording market,” but a friction point that everyone accepted but no one had solved.

They reportedly met in a small circle discussing “how to work well with tools.” During conversations, they repeatedly observed the same phenomenon: professionals are not opposed to AI helping record meetings; they are opposed to a conspicuous bot sitting in the meeting. Lawyers don’t want clients to see a recording robot; salespeople worry that a “OtterPilot has joined the meeting” alert might change the client’s attitude; executives, during internal discussions, also don’t want a digital spectator always present. This discomfort is well known in the industry, but no one treated it as a real problem to solve. Everyone defaulted to the idea that this is how AI meeting tools should look.

Granola’s solution was to move the recording from the cloud to local. The app runs directly on the user’s computer, listening to device audio output, without joining the meeting or sending any prompts, so the other party is unaware it’s running. After the meeting, the user clicks “Optimize Notes,” and AI merges scattered handwritten points with full transcriptions, generating structured meeting summaries, highlighting decisions, action items, and key quotes. It also supports natural language search across past meetings, like asking “What did Director Zhang say about the project budget last time?” and the system automatically retrieves the relevant content.

This decision isn’t technically difficult or complex in logic, but it opened the door for a group of users who previously couldn’t use competing products. Lawyers, salespeople, investors, executives—these professions all rely heavily on meetings, but are highly sensitive to being recorded. For them, Granola’s “invisible” approach isn’t just a gimmick; it’s a prerequisite for use. As a result, the product quickly spread by word of mouth within VC and startup circles, without any advertising spend—entirely through personal recommendations.

What fueled the valuation surge was a rapidly expanding list of enterprise clients. Fast-growing tech companies like Vanta, Gusto, Asana, Mistral AI, Cursor, etc., are already using it, and the reputation continues to grow within this circle. Pricing-wise, Granola chose to lower the entry barrier: Business version at $14 per user per month, Enterprise at $35, compared to Otter’s enterprise price of $16.99 and Fireflies at $19. In the same category, Granola’s prices are the lowest.

A two-person team, with a small product-level judgment, met the right user group at the right time. That’s how Granola’s first chapter was quietly written in a track many thought the pattern was already set.

******** Non-Consensus in the Red Ocean ********

The real challenge in this track isn’t building a usable product, but finding a corner in a market already dominated by major players that no one has seriously addressed.

Describing the AI meeting recording track as a “red ocean” isn’t an exaggeration. Otter.ai has been deeply cultivated for ten years, with brand recognition in North America almost synonymous with the category itself; Fireflies has accumulated extensive integrations with enterprise workflow tools like Salesforce, HubSpot, Notion; Zoom has launched its own AI Companion embedded directly into the meeting software, eliminating the need for downloads. Not to mention Microsoft and Google, both turning meeting AI into a core battlefield for office suite upgrades, leveraging hundreds of millions of users via Teams and Workspace.

Against this backdrop, Granola’s chosen approach seems somewhat odd. The mainstream strategy is to compete on features, integrations, and accuracy—making bots smarter, faster, and cheaper. Granola, however, removed the bot feature altogether.

This decision isn’t complex logically, but it requires founders to have a clear judgment about users: the group willing to pay for meeting AI and the group willing to accept a bot in the meeting are not the same. Lawyers, investment firms, high-end sales teams—these professional scenarios have paying capacity and strong recording needs, but they cannot allow a robot to sit in the meeting with clients. This demand was previously unaddressed, not because no one thought of it, but because most teams believed the market was too small to justify a dedicated solution.

Granola’s non-consensus is that they believe this market isn’t small. Once friction is eliminated, these users will be highly sticky because there are no substitutes. This judgment proved correct. The product first spread in VC circles, where investors can get a clean summary right after meetings, and find it hard to go back. These users then recommend it to founders of portfolio companies, who recommend it to sales teams, who recommend it to lawyers… Word of mouth spreads along professional chains, with weekly user growth around 10%, all without any advertising.

This growth method is characterized by slow but sticky growth. Users who come through word of mouth are those who find it useful because their peers use it; their churn rate is naturally low. More importantly, Granola’s data retention mechanisms make it increasingly difficult for users to leave. Historical meeting records, cross-meeting context retrieval, team knowledge stored in the workspace—these accumulate over time, raising switching costs. A sales team that has used Granola for two years, when considering switching tools, must think about more than just features; what about those two years of meeting records?

Of course, this track isn’t without pressure. Granola’s biggest threat isn’t Otter or Fireflies, but Microsoft and Google. Both are working on the same thing: integrating meeting context into their own AI workflows. Microsoft Copilot already offers meeting summaries, action item extraction, cross-meeting retrieval in Teams, without requiring users to install anything extra. Google Gemini’s path in Workspace is similar. Their advantage isn’t better products but distribution. Enterprise employees already work in Teams and Gmail; if Microsoft and Google decide to do this seriously, Granola’s enterprise customer acquisition costs will rise sharply.

In response, Granola is accelerating its deepening strategy. In February 2026, Granola launched an MCP server allowing external AI tools like Claude, ChatGPT, Replit, Lovable to directly access meeting data; in March, its Series C financing also opened up personal and enterprise APIs, enabling enterprise admins to connect their entire team’s meeting context into their AI workflows. The logic is: rather than competing head-on with Microsoft and Google on meeting summaries, it’s better to become a data layer that others can call upon, rather than be replaced.

******** Transcription Isn’t Enough; Recording Is the Battlefield ********

The AI meeting recording track is experiencing a quiet watershed today.

On one side is transcription. Converting spoken words into text, generating summaries, and marking action items—by 2026, this has been done well enough. Otter, Fireflies, Granola, and even native tools from Microsoft and Google have negligible differences in accuracy. The user experience gap is shrinking.

On the other side is what can be done after transcription. This is the real battleground for the next phase.

Data speaks first. According to Market Research Future, the AI meeting assistant market will grow from about $3.5 billion in 2025 to over $34 billion in 2035—almost tenfold in ten years. But a closer look at this growth source shows that simply selling transcription features can’t support this number. The real incremental value comes from activating meeting data and extracting value from it, which most companies haven’t yet systematically explored.

Meetings are among the highest information density scenarios in enterprises. They contain genuine customer demands, internal decision logic, team friction points, product evolution processes. These insights have historically been buried in unread meeting minutes or lost in individual memories. AI now offers the first opportunity to systematically structure, retrieve, and utilize this data.

The next evolution in this direction roughly follows three paths.

First, from recording to action. Currently, all meeting AI tools aim to produce a summary. But the real value lies in whether the action items are followed up, customer commitments fulfilled, risks tracked. The next generation should answer: after the meeting, can the tool automatically sync action items to project management systems, write customer commitments into CRM, and push follow-up tasks to responsible parties? This isn’t a small feature iteration but a fundamental shift in product positioning—from “help you record” to “help you act.” Products that cross this threshold will command higher user payments because they deliver not just a document but a quantifiable efficiency outcome.

Second, verticalization. General-purpose meeting tools will hit a ceiling. The real space is in vertical industries with strict compliance needs. Legal firms require traceable, evidence-grade structured records; healthcare needs HIPAA-compliant diagnosis records that can generate standard medical summaries; finance needs client communication records that meet regulatory standards and automatically flag compliance risks. In these scenarios, dependency on the tool is much higher, willingness to pay stronger, and once barriers are built, it’s hard for competitors to replace.

Third, becoming data infrastructure. This is the direction many leading players are betting on, and the highest valuation premium. The logic: when all meeting records of a company are stored in one system, that system holds a core part of the company’s knowledge assets. Building on this, the possibilities extend far beyond summaries: new employees can quickly understand project backgrounds by reviewing past meetings; management can retrieve relevant discussions for decision-making; AI agents can review relevant historical meetings before acting. In this framework, meeting records are no longer just tool outputs but the foundational data source of the enterprise’s knowledge system.

Investors viewing this track see three distinct valuation logics behind these three paths. Action-oriented solutions essentially sell efficiency, with high per-user value but requiring deep integration with downstream tools and higher implementation difficulty. Vertical solutions have clear barriers, healthy cash flow, but limited market size—best to dominate a niche before expanding. Data infrastructure offers the greatest potential but is also the hardest, requiring a critical precondition: users trusting a third-party platform with their most sensitive conversation data over the long term. This trust isn’t built by features but by time and reputation.

# 线索爆料 # rzcj@thecapital.com.cn

Media cooperation: 010-84464881