How Humane Burned $230M on the AI Pin — and Sold the Wreckage for $116M

Humane raised $230M, ran five years of stealth, and launched the AI Pin to a TED stage. Sixteen months later it sold to HP for $116M and bricked every device. The postmortem.

TL;DR

• Humane raised $230M across three rounds, spent five years in stealth, and shipped exactly one product — the $699 AI Pin. Sixteen months after launch it sold to HP in a $116M asset deal, roughly 50 cents on the dollar of capital raised.
• The collapse was fast and public. First units shipped April 11, 2024, the same day The Verge's review landed; three days later MKBHD called it "The Worst Product I've Ever Reviewed." By August, Humane had shipped about 10,000 units against a 100,000-unit yearly target — with returns exceeding sales.
• The structural error was visible on launch day, not at the brick date: five stealth years produced no predecessor product, no installed user base, no creator pipeline — nothing for the launch demand-shock to land on.

The Numbers

By August 2024, The Verge had Humane's internal numbers: roughly 10,000 AI Pins shipped, about $9M in total device revenue, around $1M in returns processed — and only about 7,000 units still in customer hands. The cap table was the opposite of a weakness: Tiger Global, SoftBank, Sam Altman, Microsoft, Qualcomm Ventures, LG, Volvo's tech fund, and Marc Benioff all backed Humane. Founders Imran Chaudhri (22 years at Apple as Human Interface design lead) and Bethany Bongiorno (an iOS and macOS PM director) had pedigree most hardware founders never reach.

What they bet on

Bet 1: A founder-led category reveal could substitute for a product launch

Humane's reveal was, by any measure, one of the best founder-led AI product moments on record. On April 20, 2023, Chaudhri delivered a TED2023 talk — "The disappearing computer" — a 13-minute stage demo with laser palm projection and a live phone call to Bongiorno. Five months later, Naomi Campbell opened the Coperni SS24 runway wearing the still-unannounced device, producing a Vogue editorial moment no AI hardware competitor had touched.

The reveal grammar was a genuine format invention — and it was also the failure mechanism. Stage demos, invite-only press hands-on, and curated palm-projection moments are not user-replicable. A buyer could not test the experience in five minutes the way they could paste a tagged sentence into ElevenLabs or run a Zoom recording through a meeting summarizer. The demo grammar promised a category; the device had to deliver it, alone, on launch day.

When it didn't, the same channels that lifted Humane turned. KOL credit transfer runs both directions. The Vogue and Coperni and TED moments compounded for Humane in 2023; in April 2024 the same graph — including MKBHD — went the other way. Borrowed credibility is not a one-way ratchet.

Bet 2: Phone-replacement framing — which forced a phone-class verdict

Humane positioned the AI Pin as a phone replacement, not an accessory. The launch announcement on November 9, 2023 priced it at $699 plus a $24/month subscription for cellular service over a T-Mobile MVNO.

That framing did one structural thing, and it was the wrong thing: it anchored reviewers on phone-class expectations — battery life, reliability, latency, call quality. The same hardware sold as an "AI accessory" would have been judged on a gentler curve. Instead, the first customer units shipped April 11, 2024, the same day David Pierce's Verge review — "not even close" — went live, cataloguing failure rates on outbound calls and noting the only function reliably working was the time. Engadget, Wired, and TechRadar piled on within 72 hours.

Then on April 14, Marques Brownlee — roughly 18M subscribers — published "The Worst Product I've Ever Reviewed... For Now." The title became a meme. By April 23, NPR's Marketplace ran "Why the Ai Pin fell flat" — the moment the failure crossed from tech press into general-audience discourse. The launch curve was killed before it ever formed.

Bet 3: Doing hardware, cellular, AI, and a new OS at once — with no substrate underneath

Five stealth years produced exactly one product launch on a single day. There was no predecessor product, no installed user base, no content-creator pipeline, and no cellular-reseller experience. The funding was raised on the promise of a launch event, so the launch event had to perform alone.

The consequences stacked. Humane built its own T-Mobile MVNO from scratch and inherited a technical limitation it only discovered post-launch: returned pins could not be reassigned to new owners. Each return was a destroyed unit plus a refund — returned pins became e-waste, not refurbished inventory. By August 2024, returns were exceeding sales. The internal sales briefing The Verge obtained framed the gap bluntly:

Humane had hoped to sell 100,000 units in its first year. As of early August, it had sold about 10,000 — and recent returns were outpacing new sales.

— The Verge internal-data report, via 9to5Mac, August 2024

A monthly burn likely north of $6M against $9M of lifetime device revenue left no headroom. The plan to monetize during a launch peak was structurally impossible — there was no peak. Reviews killed the demand spike before it formed.

What you can copy

1. Treat stealth as a period that produces substrate, not concealment. Five quiet years should leave something in the world — a smaller product, an installed base, a creator pipeline. Humane's five years produced one launch day. Use stealth to build the thing that absorbs the launch.
2. Ship a smaller, lower-stakes product first. A quieter predecessor surfaces fulfillment problems, support patterns, and limitations — like the MVNO reassignment trap — under a small blast radius, before the high-stakes launch.
3. Make your demo replicable by the user in five minutes. A stage demo performs but does not constrain. If buyers cannot reproduce the experience cheaply, the demo becomes evidence of hype the moment the product underperforms it.
4. Frame your category off the incumbent's curve, not on it. "Phone replacement" forced phone-class judgment on Humane. Pick framing that de-anchors from a category-defining claim you cannot fully deliver yet.
5. Build a creator pipeline before launch, so reviews land alongside existing-user testimony. When KOLs encounter a product cold at launch, they judge it cold — and there is no installed-user counterweight to balance a launch-week verdict.

What probably won't work for you

This is a cautionary tale, so this section is the most useful one. Several preconditions made Humane's failure especially structural — and the lessons cut against the instincts a strong cap table encourages.

Founder gravity is not product readiness. Two ex-Apple senior leaders with a $230M cap table can attract press, fashion coverage, and a TED stage before the product is real. The market treats founder-as-IP as evidence the product is ready. It is not. Humane executed the founder-reveal move at the highest level of any 2023 hardware launch, and it did not save the company. Pedigree buys attention; it does not buy a working device.

A category-defining promise forces a category-defining bar. Each year of stealth raised the implicit promise of the launch event. By year five, "show us what you've been building" was an impossible question to answer with a single product. If you make a five-year promise, you owe a five-year answer — and one device cannot be it.

Doing four hard things at once means each must work on day one. Hardware, cellular service, an AI assistant, and a new operating system are each a large bet alone. Humane attempted all four with no predecessor product, so every one of them had to work on launch day. They didn't.

The exit was not a recovery. HP's $116M asset deal on February 18, 2025 was a reverse-acquihire: IP, the CosmOS operating system, 300-plus patents, and roughly 80 employees, with the device line explicitly excluded. Twelve months earlier the sale target had been $750M-$1B. On February 28, 2025 at noon Pacific, every consumer AI Pin lost its server connection — calling, messaging, AI queries, and customer data all gone. A failed hardware bet does not get a soft landing for its customers; it gets a brick date.

Sources & references

• Humane Series C announcement — PR Newswire, March 2023
• Imran Chaudhri's TED2023 talk — "The disappearing computer"
• Humane launches the AI Pin — PR Newswire, November 2023
• Humane AI Pin review: not even close — The Verge
• The Worst Product I've Ever Reviewed... For Now — MKBHD
• Why the Ai Pin fell flat — NPR Marketplace
• Returns exceed sales, May-August 2024 — 9to5Mac
• HP buys Humane's assets for $116M — TechCrunch
• Full timeline: The Humane Growth Story — every round, launch, and inflection charted

This case study is part of GrowthHunt's growth teardown series. Compare it with the Cursor teardown — a product that earned word-of-mouth before it asked for attention — or the Lovable teardown, then track the fastest-growing tools and founders live on GrowthHunt Velocity.

More from the growth playbook.

See it run in the wild.