Innventure's Crash Is a Lesson in the Difference Between a Booking and a Customer

On November 17, 2025, Innventure announced something that looked like a company crossing a line: a subsidiary, Accelsius, had signed an agreement for a Canadian outfit called DarkNX to deploy its liquid-cooling technology across a new 300-megawatt AI data center campus in Ontario. Management described it as a major commercial milestone on the road to a $100 million annualized revenue run rate and cash flow positivity by the end of 2026. Nine months later, on August 13 of this year, Innventure filed its quarterly report and admitted the deployment site in that purchase order "is no longer available," pulled the project out of its bookings, and suspended the 2026 targets. The stock fell 55% the next day, to $1.62, and it has kept sliding — it trades around $1.10 now, down about three-quarters on the year.

There is a securities class action attached to all this, with a lead-plaintiff deadline of October 27, 2026, and a passel of plaintiff firms reminding investors about it. That is the billboard. Underneath is a cleaner lesson, which is about the difference between a booking and a customer, and how much of a company's entire valuation one unperformed contract can carry.

What you actually bought

Innventure is a "company creation" vehicle. It licenses undeveloped technologies from big multinationals, builds operating companies around them, and takes them public as a package. It came public itself in October 2024 through a SPAC. Its flagship is Accelsius, a direct-to-chip liquid cooling startup built on tech licensed from Nokia — cooling is one of the hottest problems in the AI data center buildout, which is why the Accelsius story could run so hard. Accelsius essentially is the company: in the second quarter of 2026 Innventure's total revenue was $953,000, and 96% of it came from Accelsius.

That means the bull case was, in large part, a single-asset story, and the asset's story was one customer. Accelsius had raised a $65 million Series B late last year led by Johnson Controls, and the round valued it at $665 million post-money. The company leaned on a claimed $1 billion-plus sales pipeline and "over $50 million in first-quarter 2026 bookings" as evidence of a commercial inflection point. But the bookings largely trace back to the DarkNX purchase order.

Now, a purchase order is a strange thing to put a whole valuation on. It is a contract, not revenue. Its value is whatever the counterparty can be made to pay, which depends on whether the counterparty actually exists as a going concern and can perform. The class action — filed in the Southern District of New York under the securities-fraud provisions, covering the period from the November announcement through August 13 — alleges the company knew the deal was unlikely to materialize and talked it up anyway, that there was no evidence DarkNX was building or facilitating a large-scale data center.

That allegation traces to a May 28 short-seller report from Morpheus Research, which called DarkNX a shell: incorporated barely a year before the deal was announced, registered to a suburban house, a claimed data-center address that turned out to be a trucking company, nine LinkedIn employees with other full-time jobs and no data-center credentials, no institutional funding, and "strategic partners" who said they had no relationship with it. Innventure called the report "selective and misleading", and for three months the stock held a version of the story. Then, on August 13, the company itself confirmed the practical version of the short seller's thesis: no site, no project in bookings, no 2026 targets.

I want to be careful about what is proven here. The short-seller charges are allegations, folded into a complaint that is also, at this point, allegations. What is documented is what Innventure said publicly, and then what it took back. The structural point does not depend on whether anyone committed fraud: management told shareholders the $100 million run-rate and the cash-flow-positivity target were achievable on the back of a purchase order, and then the purchase order turned out not to correspond to a deployable site. The whole "commercial inflection point" was, as far as the disclosure trail shows, largely one contract that did not perform.

The incentive plumbing

This is the part that makes it funny (in the cold sense). The company's executives had a direct financial incentive to announce a big binding contract, because their compensation was wired to it. The short-seller report and the lawsuit both point to an earnout where insiders received 2 million shares — worth around $9.2 million — tied to Accelsius signing a binding contract for more than $15 million, a milestone the DarkNX deal is said to have triggered. Whether or not that is exactly how the numbers break down, the shape is familiar: a company that creates its own milestones, books a headline customer to trip them, and solves the current funding problem by selling a story about future revenue.

The plumbing also shows up in who pays for the story. Note the financing history: roughly 88% more shares outstanding in about eighteen months, an overhang of warrants and a standby equity arrangement with Yorkville Advisors that could add another chunk of dilution. The shareholders are effectively the funding source for a venture builder that needs continuous capital while its operating companies generate almost no revenue.

What is left

Strip out the DarkNX narrative and the disclosed math is blunt. Innventure ended the second quarter with about $46.5 million of cash, equivalents, and restricted cash, on a market capitalization of about $93 million — so cash is roughly half the company, with the rest being stakes in a barely-revenue liquid-cooling startup and two younger ventures (AeroFlexx and Refinity). Against that cash, the company lost $34.9 million in a single quarter, or about $22.6 million of adjusted EBITDA. That is a cash pile being consumed in roughly two quarters at the headline burn, before accounting for the fact that some of the loss is non-cash.

There is a legitimate way to read this as cheap: the stock trades below book value, cash covers a big share of the market cap, and Accelsius' institutional investors valued it at $665 million not long ago. There is a more honest way to read it, which is that book value at a venture builder is only as good as its ability to stop burning cash on assets that are not yet earning, and the one asset that made the story exciting is the one whose customer just vanished. Two quarters of runway and a history of dilution is a short leash; whether the remaining shareholders get asked to write new checks is a live question, and the existing standby-equity facility is the mechanism already in place to do it.

The class action, briefly

The lead-plaintiff deadline is the one date being hammered into the news cycle, so let's be clear about what it does. In a federal securities class action, the October 27 deadline is for investors who want to be lead plaintiff — the person who gets to choose the lawyers and steer the case. If you bought INV shares during the class period and did nothing, you are still a class member and covered by any settlement; you do not need to file anything, and the case does not turn on the reminder emails. So the deadline matters mostly to institutions and large holders who want to run it. The economic reality for the ordinary holder is simpler and colder: the value of whatever the company actually owns is not going to be fixed by the lawsuit. The diligence question was always whether DarkNX was a customer or a watermark on a press release, and on that question the company's own August disclosure is the most recent and most reliable word.

A booking and a customer are different nouns, and the gap between them was the entire business of Innventure shares. When a company's flagship "customer" is a purchase order from an outfit with no site, the correct price for the stock is not ninety percent of a $100 million run-rate — it is whatever the disclosed balance sheet and a realistic view of a pre-revenue portfolio will support. The rest was a projection riding on a contract that, in the end, the company itself said was not going to happen.