C3.ai Q2: beat but revenue declines, wider losses; stock falls on cautious Q3 view

What happened

C3.ai posted a quarter that beat near‑term estimates yet showed a sharper year‑over‑year revenue drop, wider net losses and margin pressure, and the stock sold off after management guided cautiously for Q3. Management emphasized strong federal bookings, but those deals are lumpy and long‑lead — useful for RPO headline numbers but not a substitute for measurable subscription consumption (vCPU/vGPU) that drives durable revenue. The company still holds a meaningful liquidity cushion (~$712m) and RPO (~$235m), and its partner ecosystem and production deployment counts validate product interest. That said, persistent negative free cash flow, steep EPS losses and long sales cycles keep execution risk high and leave valuation unsupported absent an operating‑leverage inflection. Investors should be skeptical of filings that highlight bookings and deployments as progress without clear, sustained conversion into consumption revenue and improving FCF/EPS trends.

Implication

Q2’s beat is not proof of durable improvement: management’s emphasis on federal bookings and deployments raises RPO headlines but not necessarily near‑term revenue or margin durability. The critical watch items remain subscription mix vs. services, sequential RPO and production deployment trends, and tangible vCPU/vGPU consumption growth that converts deployments into recurring revenue. Cash on the balance sheet reduces immediate capital risk, but continued wide losses and volatile free cash flow mean funding or dilution risk could re‑emerge if conversion stalls. If subscription growth and consumption accelerate materially and FCF turns sustainably positive, upgrade toward BUY; if these metrics stagnate or worsen, the case deteriorates toward SELL. For now, maintain a cautious HOLD and use rallies to reduce exposure until operational evidence of scaled consumption appears.

Thesis delta

No change to our HOLD thesis. The quarter reinforces the report’s central gate: bookings and deployments validate product demand but do not yet prove scaled consumption or operating leverage. We slightly increase emphasis on monitoring consumption metrics (vCPU/vGPU), subscription mix and RPO conversion while remaining skeptical of booking‑centric narratives in filings.