Failed AI Startups: 26 That Raised Millions and Still Collapsed

Failed AI startups are piling up faster than almost anyone expected. More money has poured into artificial intelligence over the past three years than into any other technology on earth, and a single AI startup can now raise more in one round than most companies used to raise in a lifetime.

The pitch decks promise to rewrite every industry, and the cheques match the ambition.

And yet a long line of well-funded AI companies has gone to zero. Some filed for bankruptcy, some were sold for scraps, and some had their teams hired away by the very investors who backed them.

The common thread is that money was never what they lacked.

This piece tracks 26 failed AI startups, from the US, the UK, Europe and India, that raised serious capital and still collapsed. Between them they took in well over $11 billion in disclosed funding, a figure that mixes venture rounds, listing proceeds and debt.

The point is not to mock them. It is to show, with names and numbers, that funding is fuel and not a finished engine, and our look at survivorship bias in startups explains why the failures teach more than the winners.

Failed AI startups: the numbers

Before the names, here is the shape of the data across all 26 companies.

The US accounts for about half the list, with the UK and India close behind. Self-driving and AI healthcare drew the biggest cheques and produced the heaviest losses.

The full list of failed AI startups

Here is every company, what it did, where it was based, what it raised, and how and when it ended. Figures are approximate and drawn from public reporting.

A handful of self-driving companies alone burned through more than five billion dollars between them.

Two consumer-AI darlings, Inflection and Character.AI, raised a combined one and a half billion and were absorbed by their own backers within two years of launching.

And the largest single collapse, Builder.ai, did not run out of a good product so much as run out of a story people believed.

The billion-dollar collapses

Two names stand apart, because they did not just fade. They fell.

Builder.ai is the cautionary tale of the AI era. The London company, founded in 2016 by Sachin Dev Duggal and backed by Microsoft and the Qatar Investment Authority, sold the dream of building an app as easily as ordering a pizza, with an AI assistant called Natasha.

It was once valued at around $1.2 billion. In May 2025 it collapsed into insolvency after reporting that it was running out of money, and a lender reportedly seized most of the roughly $40 million left in its accounts.

The deeper problem was what sat behind the brand. For years there were allegations, which the company denied, that the "AI" was largely hundreds of human engineers in India writing code by hand.

After the collapse came reports of inflated and allegedly round-tripped revenue. It had raised about $445 million.

The lesson is blunt. You cannot out-market the absence of a real product.

Babylon Health raised the same questions in healthcare. The London startup, founded by Ali Parsa in 2013, built an AI symptom-checker and the telehealth service behind the NHS "GP at Hand" practice, and at its 2021 SPAC listing it was valued at $4.2 billion.

By August 2023 it had gone into administration, and its UK assets were sold for just $620,000. It had taken in well over half a billion dollars of venture money across its life.

Babylon spent faster than any plausible revenue could support, its clinical claims drew sustained criticism, and when the public markets turned, the financing it needed never arrived.

A multi-billion valuation turned out to be worth less than a London flat.

The self-driving money pit

No category torched more capital than autonomous vehicles. The technology is real, but the timeline to profit kept stretching past what investors would fund.

Argo AI is the starkest example. Backed by Ford and Volkswagen, the Pittsburgh company raised more than $2.6 billion and was once valued at $7 billion.

In October 2022 its backers decided robotaxis were too far off, pulled their money and shut it down, splitting the staff between the two carmakers. Ford took a $2.7 billion write-down.

A whole cohort met the same end. TuSimple, once worth billions on Nasdaq, wound down its US operations amid cash burn and national-security scrutiny. Embark Trucks went public in a $614 million deal, then ran out of money and sold its assets in March 2023.

Ghost Autonomy, backed even by OpenAI, burned about $220 million before shutting in April 2024. In the UK, Cambridge's Five AI raised around $78 million chasing a British robotaxi before giving up and selling to Bosch.

The pattern is identical. Self-driving is a capital problem disguised as a technology problem, and when cheap money ended, so did the runway.

AI healthcare's hard landing

Healthcare drew huge AI bets on the promise of cheaper, smarter care. It produced some of the heaviest losses.

Olive AI raised roughly $902 million to automate hospital admin and was once a $4 billion unicorn, before winding down in October 2023, undone by over-expansion and a loss of focus.

Forward Health took about $650 million to build gleaming AI "CarePods," self-service medical kiosks, then shut down in November 2024 when the hardware proved expensive and the economics did not work.

Mindstrong raised around $160 million for an AI mental-health platform and wound down in 2023. In the UK, Sensyne Health and BenevolentAI both listed publicly, both promised AI-driven drug discovery, and both saw their value collapse after clinical and commercial setbacks.

The sector's lesson is that medicine is slow, regulated and stubborn. No amount of funding makes a clinical result or a hospital sales cycle arrive faster.

The soft landings that were still failures

The most modern form of AI failure barely looks like failure at all. In the race to build frontier models, several of the most hyped startups could not afford to keep training, and were absorbed by their largest backers.

Inflection AI raised $1.3 billion, launched a polished chatbot called Pi, and within a year had most of its staff hired into Microsoft, which paid around $650 million to license its technology.

Adept AI raised about $415 million and ended up with its founders and tech going to Amazon. Character.AI, hugely popular with users, could not fund its own model training and did a licensing deal that sent its founders back to Google.

Each of these founders landed well. Each company, as an independent business, stopped existing.

That is the cleanest proof of the thesis. More than a billion dollars could buy the models and the users, but not a standalone business that could pay for itself. Older cases like Canada's Element AI ($257 million, fire-sold to ServiceNow) and France's Snips show the same shape stretching back years.

Hardware, the harshest teacher

Two companies prove that AI plus hardware is the most unforgiving combination of all.

Humane raised around $230 million, drew comparisons to Apple, and launched its AI Pin to some of the worst reviews in recent gadget history. Returns reportedly outpaced sales before HP bought the remains for $116 million in February 2025.

Zume took roughly $445 million from SoftBank to build pizza-making robots, pivoted to packaging, and shut down in 2023.

Software can be rewritten overnight. A warehouse of robots or a boxful of returned devices cannot.

India's quieter AI shakeout

India's failed AI startups were smaller in dollar terms but just as instructive, and they came faster as the funding winter bit.

Toplyne, backed by Sequoia and Tiger Global with about $17.5 million, shut in October 2024 after failing to find durable product-market fit. OSlash, with around $7.5 million from Accel, closed in 2023, and Nintee wound down and returned money to investors.

A cluster of smaller generative-AI startups, InsurStaq.ai, subtl.ai and Astra, each shut within a year or two of starting, undone by thin funding, long sales cycles and an inability to raise the next round.

The Indian story is less about billion-dollar hubris and more about a hard truth for early founders everywhere. A wrapper around a popular AI model is easy to launch and easy to copy, and being first is not the same as being needed.

What the failed AI startups had in common

Read across all 26 and the same handful of causes keep coming back. They have little to do with the fundraising and everything to do with the business.

The first was product-market fit that never truly arrived. Many had users, demos and press, but not enough people who needed the product badly enough to keep paying. A launch spike is not retention.

The second was capital intensity with no end in sight. Self-driving, robotics and hardware swallow money for years before any return, so when the market turned in 2022 and 2023, these were the first to run dry.

The third was leaning on the AI label itself. Builder.ai is the extreme case, but plenty of companies used the word "AI" to raise, then could not show the technology did anything a rival could not copy in months.

The fourth was hype priced in and delivery priced out. A $4 billion valuation sets an expectation the company then has to grow into, and several here, Babylon above all, raised at numbers their revenue could never justify.

The fifth was dependence on a funding climate that changed. Cheap money funded long-shot bets, and when rates rose and investors demanded a path to profit, the companies built for endless rounds had no plan B.

Notice what is missing from that list: a lack of money. Almost none of these companies failed because they could not raise. They failed with hundreds of millions in the bank, or shortly after raising more.

The takeaway for founders

Funding buys you time and talent. It does not buy you a reason to exist.

The companies above had serious investors, enormous war chests and, often, genuinely clever technology, and it was not enough. The thing that makes a company survive is the same thing it always was: customers who need the product, economics that work at scale, and honesty about what you have actually built.

A big round is a vote of confidence, not a guarantee. Treating it as proof you have won is how a well-funded company ends up as a line in an article like this one.

For the other side of the ledger, the UK's broader funding picture is strong, as our look at the top UK AI startup investments shows. The discipline that separates the survivors starts long before the money arrives, with the hard work of validating that anyone wants what you are building.

The AI boom is real, and plenty of these ideas will be built successfully by someone, eventually. But raising money is a start. It was never the finish.