The $2 Billion Pivot: How Killing His Own Product Built the Keyboard's Replacement

Tanay Kothari spent three years and his company’s entire first round of funding on a wristband that would let people type with their minds. It never shipped. He killed it, and six weeks later he had a new product, a Product Hunt launch, and the start of a run that took his company from $4.6 million to a $2 billion valuation in under five years, most of it in the last twenty-two months.

This is the story of Wispr Flow. It’s as much a lesson in what to do when your first idea fails as it is a lesson in how to raise money.

Kothari and Sahaj Garg met on their first day at Stanford, dropped into the same freshman dorm. Kothari had already shipped seven apps as a teenager in New Delhi, over 30,000 downloads across 17 countries before he ever got to campus. At Stanford he worked as a teaching assistant for Andrew Ng's deep learning course and did medical AI research at Stanford's AIMI lab, co-authoring a paper on pulmonary embolism detection published in npj Digital Medicine. Garg, a Henry Ford II Scholar who graduated top of his engineering class, published across SAIL, Google Research, and Harvard Medical School.

Neither of them was a first-time founder when they started Wispr. Kothari had built Convert.cc to 2.5 million monthly users without spending a dollar on marketing, then founded FeatherX, an e-commerce personalization company that got acquired within months, after which he ran product and engineering at the acquirer. Garg had led the AI team at Luminous Computing, designing the system architecture for their chip. This was not two people guessing.

What they built is a voice dictation tool that has become, by its own account, the default way tens of thousands of businesses and a majority of Fortune 500 employees put words into a computer. As of August 2026 it processes over 60 billion words, runs in 162 countries and 100-plus languages, and just closed a $280 million Series B at a $2 billion valuation.

None of that was the plan in 2021.

Wispr’s original idea was a wearable. The pitch, funded by a $4.6 million seed round from NEA and 8VC in November 2021, was a wristband that read the electrical signals your muscles send when you silently try to speak, an EMG interface, no keyboard, no voice, nothing visible at all. The seed round’s angel list tells you how credible this looked at the time: Josh Duyan, co-founder of CTRL-Labs (the neural-interface company Meta bought), Chris Manning, Stanford’s own ML and linguistics chair, and Richard Socher, Salesforce’s Chief Scientist.

They spent nearly three years on it. In October 2022 they raised a $10 million follow-on, part equity, part venture debt, from Neo, Triple Point Capital, and Fred Ehrsam of Coinbase and Paradigm, with angels including Synchron’s CEO and Dropbox’s co-founder. They hired neuroscience researchers, poached engineers from Apple, Tesla, and Sony. And at the end of it, the founders concluded the AI underneath the hardware just wasn’t good enough yet to make the thing work at the fidelity people would actually need.

Most founders in that spot keep going. Nearly $15 million raised, a real team, a real thesis, hard to admit it’s not working. Instead, they took the speech-processing stack they’d built to feed the wristband, turned it into a standalone dictation app, and ran a six-week sprint from that decision to launch. Wispr Flow hit the Mac App Store and Product Hunt in October 2024 and finished #1 for both the day and the week.

The pivot itself isn’t the rare part, pivots happen constantly. What’s rare is recognizing that the technology, not the team, was the blocker, and having the discipline to stop funding a thesis the tech couldn’t support instead of grinding on a wearable nobody outside the company could actually use.

One detail worth sitting with: seven months before Menlo Ventures formally led Wispr’s Series A, Menlo partner Matt Kraning personally wrote a $12 million check into the company. Not a fund, his own money. He wasn’t even at Menlo yet, he joined the following spring. He backed the pivot itself, before Product Hunt, before there was traction data to point to. Kraning had his own exit under his belt, he co-founded the cybersecurity company Expanse and sold it to Palo Alto Networks for $1.25 billion in 2020. When he joined Menlo and led the Series A nine months after his personal check, then the Series B a year after that, it wasn’t a fund discovering a hot deal. It was an operator following through on a bet he’d already made with his own money.

Most founders pitch investors who stopped writing checks six months ago. Fund Momentum tracks 1000+ actively deploying VC funds with real GP signals — who’s bullish, what they’re writing, and how to pitch them. Now also available as an MCP API so your AI tools can pull investor intelligence directly. Stop guessing. Start closing.

Here’s the full history, and it reconciles cleanly once you check every figure against the next: Seed, $4.6 million, November 2021, NEA and 8VC. Seed II, $10 million, October 2022, Neo and Triple Point Capital. An unannounced $12 million round led personally by Matt Kraning, September 2024, right at the pivot. Series A, $30 million, June 2025, Menlo now formally leading. Series A extension, $25 million, November 2025, Notable Capital, at a $700 million valuation, and this round came bundled with something arguably as valuable as the capital: a year-long distribution partnership on The Diary of a CEO, a podcast with 35 million subscribers. Then the Series B: $280 million in August 2026, Menlo leading again, at $2 billion.

Kraning’s own explanation for doubling down is unusually specific for a venture blog post: “What convinced us to double down was watching demand outrun the company. Flow spread to employees at the majority of the Fortune 500 before there was much of a sales team to speak of; until this year, the entire enterprise business ran through a single account executive working inbound.”

That’s the actual anomaly here. Not the dollar figure. A company with a two-year-old consumer product ended up inside most of the Fortune 500 without ever building a sales org to get there.

This isn’t an uncontested category, and it’s worth saying so plainly. OpenAI is reportedly building a voice-controlled hardware device with Jony Ive for a 2026 launch and claims over 150 million people already use ChatGPT’s voice features weekly. Google shipped Gemini Live with emotional-nuance detection. And in the six weeks before Wispr’s Series B closed, Meta acquired two speech-AI companies outright: PlayAI for speech synthesis in July, WaveformsAI for emotion recognition in August, both explicitly feeding its smart-glasses roadmap. A well-capitalized competitor is buying its way into exactly the layer Wispr sits on.

Closer to home, the standalone dictation competitors are real too. Superwhisper competes on privacy, on-device processing, cheaper. Aqua Voice, developer-focused, has published an independently verified 6.24% word error rate on the public Open ASR Leaderboard, a number that, if directly comparable, beats Wispr’s own claimed roughly 10%. Wispr hasn’t published its accuracy against an independent benchmark the way Aqua has. None of this undoes the growth story. But a fair reader should know Wispr’s edge right now looks more like distribution and habit than a settled technical lead.

Raising (soon)? I’ve helped 50+ startups to accelerate their fundraising campaigns and match them with the right investors. If you want hands-on support - from pitch refinement to warm intros to term sheet negotiation - let’s talk.

1. If the technology can’t support the thesis yet, stop funding the thesis, not the team. Wispr’s founders didn’t pivot because the market wasn’t there or the team failed. They pivoted because they honestly concluded the AI they needed didn’t exist yet. That’s a harder call than it sounds, most founders keep pushing a broken thesis because admitting it’s broken feels like admitting failure. The six-week sprint from that decision to a public launch is the actual skill on display here.

2. Turn your product into the sales team. Wispr ran its entire enterprise motion through one inbound account executive for most of its life. That only works if the product, not a pitch deck, is doing the convincing. The founders’ own metric of choice: whether users trust the output enough to hit send without editing it. Optimize for that number and the sales org becomes optional for a lot longer than conventional wisdom says it should.

3. Get the people who write checks to become daily users before they write the check. The founders personally onboarded Reid Hoffman and Marc Andreessen as early users. When a VC uses your product every day, they don’t need convincing it’s good, they push it into their own portfolio for you. That’s a distribution channel you can’t buy directly.

4. An investor’s personal conviction can outrun their firm’s process, and that’s useful information. Kraning wrote a personal check into Wispr before he had an institutional platform, and before the product that would justify the bet had even launched. If you can get an individual at a fund excited enough to invest their own money ahead of a formal process, that’s a stronger signal than a partner meeting, and it’s worth cultivating those relationships directly instead of only pitching firms.

5. Localize deeply, don’t just translate. Wispr’s India expansion wasn’t a language toggle. It shipped Hinglish support and priced at roughly $3.40 a month, a fraction of the US price, timed to the Android launch. India became the company’s second-largest market within three months, three-quarters of users on annual plans. A market that looks small at your home-market price can be your fastest-growing one at the right price and the right dialect.

40,000+ teams have already hired an AI employee. Viktor works inside Slack and Microsoft Teams, connects to 3,000+ tools, and does the work end to end, not just answers questions. Hire Viktor for your team, $100 in credits, no card required.

The $280 million is going toward closing what’s left of the accuracy gap and expanding what Wispr calls “zero edit rate,” the share of dictated text a user never has to fix. The company has also stood up the Wispr Advanced Interfaces Lab, led by Ariya Rastrow, a founding member of Amazon’s original 2012 Echo and Alexa team who went on to lead roughly 400 researchers there before a stint at Meta working on voice and multimodal models for wearables. The ambition is bigger than dictation: routing between models and modalities so voice becomes the default way people get things done with software, not just a faster way to type.

They’ve also just shipped Notetaker, free inside the same app. It names your speakers, summarizes the meeting, lets you search across every call you’ve had. Same playbook, one layer up: give away the next habit, make going back to anything else feel like a downgrade.

The open question for readers raising their own rounds isn’t whether Wispr succeeds. It’s whether you have the discipline to do what Kothari and Garg did in 2024, recognize when your current idea is the wrong one, and move fast enough on the pivot that you don’t burn the credibility of the round that got you there.

Want the full round-by-round breakdown, exactly which investors backed which stage and why, the precise mechanics behind Wispr’s 40%+ monthly growth, and a founder-ready checklist you can apply to your own raise? That’s in this week’s paid Playbook below.