C3.ai Reset Its Revenue — and Expectations. The Turn Isn't Proven Yet

C3.ai just reported a quarter in which revenue fell 52.5% year over year, to $51.6 million, and its chairman called the sales performance "entirely unacceptable, to the point of surreal." A stock like that should be broken. Instead, C3.ai trades near $10.50, up roughly 15% over the last month, with a market cap of about $1.6 billion — and a cash pile, which the founder just made bigger, worth roughly 40% of that.

That gap between a collapsing income statement and a stock that refuses to stay down is the story. C3.ai has finished resetting expectations. It has not yet proven the reset works.

Why the revenue number is lying about the business

C3.ai sells enterprise AI applications — predictive maintenance, supply chain, defense and logistics software — the kind of turnkey work that once got it compared to Palantir. It used to sell the old way: big multi-year subscription contracts signed up front. In 2022 it abandoned that, dropping what founder Tom Siebel called "elephant hunting" — three-year, million-dollar-plus deals that took forever to close — for consumption-based pricing. Customers start with a pilot, then pay as they actually use the software, at a price like 55 cents per vCPU hour. Siebel was blunt about the cost at the time: near-term revenue would take a hit.

Then the company made the income statement look worse on purpose. As it restructured through fiscal 2026, it stripped out the professional-services projects that had padded reported revenue. On the June earnings call, Siebel put it plainly: "Make no mistake. We are focused on software revenue, guys, not services."

Here is the quarterly path of the damage. Fiscal 2025 built from about $87 million, up to $94 million, $99 million, and a $108.7 million close. Fiscal 2026 went the other way, hard: $70.3 million, then $75.1 million, then $53.3 million, then $51.6 million. Full-year revenue fell from $389.1 million to $250.3 million, down roughly 36%. Underlying the whole decline is a real thing management calls out plainly: miserable sales execution.

The other side of the ledger

Now look at where the bookings went, because this is where the reported numbers and the health of the business diverge.

Federal, defense and aerospace bookings grew 89% year over year in the second quarter of fiscal 2026, then 134% in the third — climbing from 45% to 55% of all bookings. The federal mix now includes a five-year, up-to-$500 million Department of Defense agreement, an army of agency customers, a Microsoft alliance that has produced over $130 million in bookings in its first year, and a Shell expansion announced in June. The founder, in case you missed the shift in emphasis, has been refining the company toward the defense and intelligence business for two years.

That said, the leading indicator is not clean. In the fourth quarter management signed 28 agreements but said bookings were "lower than expected." So the pipeline is tilting the right way, but the cash register has not confirmed it yet.

The balance sheet is doing the convincing so far

This is where the proof actually sits today — and I want to be straight about how thin it is.

Cash and marketable securities stood at $575.4 million at the end of April. By June 3, they were $673 million, because Siebel bought 6.17 million newly issued shares at $11.16, putting roughly $69 million of his own money into the company. Folded into the ~$1.6 billion market cap, the cash is roughly $4 a share against a ~$10.50 stock. At the recent burn of about $55 million a quarter, that's three-plus years of runway.

The cost program is supposed to stretch it further. Headcount went from about 1,075 in January to roughly 700 — a 35% cut — and management says it has removed about $135 million in annual operating costs, with the savings from non-labor expenses fully showing up in the second half of fiscal 2027.

Now the candid part. There is no free cash flow to prove this yet. Trailing free cash flow is roughly negative $190 million, and each of the last two quarters burned about $55 million. Management is still guiding to a non-GAAP operating loss of $128 million to $160 million for fiscal 2027. A story carried by a cash cushion, founder capital, and bookings skew is a higher-uncertainty story than one carried by cash flow — I'm not going to pretend otherwise. The founder also has a habit of re-owning the outcome: he stepped down as CEO in September 2025 citing health, then resumed the job in May 2026 once his health was "largely resolved", with his replacement demoted to president. Around the edges, a long-tenured director left in June to take a European Commission post, and in late August an outside lawyer was added to the board. None of that is a thesis. It is a founder who keeps re-buying his own clean-up.

The concrete test, and what would break it

Here is what has to happen for the reset to become a turn, stated as plainly as I can. Management's fiscal 2027 guidance calls for revenue of $210 million to $240 million — still down 4% to 16% from fiscal 2026, so they are not yet claiming growth, only a floor. It guides the first quarter (the three months ended July 31) to $50 million to $54 million. That number gets reported Wednesday, September 2.

The one number in the file that matters most is not a blowout — it is whether revenue holds the floor and whether the operating loss actually shrinks. Guided for the year at an average of roughly $32 million to $40 million a quarter, versus the recent $55 million, the loss tightening is the income-statement evidence that the $135 million cost program is real. If revenue breaks below the $210 million floor, or burn doesn't shrink and cash starts sliding toward $400 million within a year, the pivot is just running in place — the cushion becomes a countdown rather than a bridge. That is the tripwire.

I can be wrong again; this stock touched $7.68 within the last year, and beaten-down has a way of staying beaten down. But the market is still pricing the risk profile of the old story — AInvest's aggregate signal still labels the shares a Hold, and at roughly 4.5 to 5 times forward sales on a still-shrinking base there is no bargain-multiple cushion to hide behind. The easy part is done: expectations are reset, the cash is real, the founder has his money in, and the bookings are tilting toward customers with budgets governments fund.

The hard part starts September 2, when the company has to show the floor holds and the burn is coming down. Watch that print the way you'd watch a foundation pour — not for drama, but for whether the grade is exactly where they said it would be. If the floor holds and the losses shrink, the case builds itself from there. If they don't, you've learned the answer cheaply, without trusting a decline that was priced as a turn.