AI — C3.ai, Inc.

C3.ai built the enterprise AI category's most recognizable independent brand, assembled a library of 130 prebuilt applications, and signed distribution alliances with Microsoft and AWS — then watched its quarterly revenue fall 46% year-over-year while GAAP gross margins collapsed to 17%, its founder-CEO stepped down due to declining health, and the company cut 26% of its workforce and slashed full-year revenue guidance nearly in half. At roughly $10 per share and 3.7 times enterprise value to revenue on a business with no independent sales motion, no path to positive cash flow, and a founder's voting control that prevents the governance corrections a new CEO would need to make: avoid.

---

The enterprise AI software market is growing at roughly 19% annually toward an estimated $115 billion in 2026 and will reach several hundred billion by the end of the decade. That fact alone has sustained the valuations of virtually every company with "AI" in its product description or investor relations presentation. The relevant question for any specific enterprise AI company is not whether the category is growing — it is — but whether the company has built a structural position within it that generates durable economics. Enterprise software at its best creates customer lock-in that produces high gross margins, predictable renewal rates, and revenue that compounds without proportional increases in cost. The evidence from C3.ai's fiscal third quarter of 2026 indicates it has built neither of those things.

The companies capturing durable economics from enterprise AI are the ones whose AI capabilities are entangled with data that cannot easily be moved. Microsoft's Copilot and Azure OpenAI Service are embedded in the productivity software used by 350 million commercial users, with direct billing relationships across every significant enterprise on the planet. Salesforce's Agentforce is delivered through the same platform that holds fifteen years of enterprise customer data — customers cannot decouple the AI from the CRM without abandoning both. ServiceNow is building AI natively into IT service management workflows where the switching cost is organizational paralysis. Palantir has embedded its Gotham platform into the decision infrastructure of U.S. government agencies in ways that make replacement a national security discussion rather than a vendor evaluation. Each of these businesses compounds its AI advantage by deepening lock-in that existed before AI became the priority. C3.ai began the AI era without that pre-existing lock-in and has not built it.

C3.ai was founded in 2009 by Tom Siebel, who previously built Siebel Systems from inception to $2 billion in annual revenue before its acquisition by Oracle. The company went public in December 2020 at $42 per share, reached approximately $170 per share in early 2021 during the initial AI enthusiasm, and has declined toward $10. It sells enterprise AI software across more than 130 prebuilt applications addressing use cases in energy, manufacturing, defense, financial services, and utilities — predictive maintenance, supply chain optimization, fraud detection, reliability forecasting, and others. The platform has been rebranded the "C3 Agentic AI Platform" to address the generative and agentic AI narrative. Customers include Shell, Bank of America, the U.S. Air Force, and Koch Industries. The revenue model is subscription-based, with subscription revenue representing approximately 90% of the total.

The moat case for C3.ai rests on three claims: that its prebuilt application library represents a durable time-to-value advantage over building AI applications in-house; that its federal and defense customer relationships create sticky recurring revenue; and that its partner network with Microsoft and AWS amplifies distribution beyond what the company could achieve independently. The Q3 fiscal 2026 results contradict all three simultaneously.

The prebuilt application advantage is the first to dissolve on examination. As large language model APIs have become commoditized at sub-cent-per-token pricing, the cost to replicate any individual C3.ai application has converged toward the cost of a developer-month of effort. The 130 applications that once represented years of domain-specific development are increasingly fungible against what a competent enterprise development team with GPT-4 API access can assemble in weeks. Microsoft, Salesforce, ServiceNow, and AWS are all embedding AI application capabilities natively into platforms that already serve the same enterprise customers C3.ai targets, at no incremental per-seat cost to customers already paying for the base platform. The displacement does not require a dedicated competitive campaign against C3.ai; it happens as a byproduct of the hyperscalers deepening their existing relationships.

The gross margin data makes the structural reality undeniable. A software business with durable switching costs should generate gross margins in the 70-80% range — the cost of serving an incremental customer approaches zero because the software already exists. C3.ai's GAAP gross margin in Q3 fiscal 2026 was 17%. Palantir's gross margin exceeds 80%. ServiceNow's gross margin is 78%. The gap is not a matter of scale or maturity. It reflects the structural reality that C3.ai's application deployments require substantial human customization and services delivery for each enterprise engagement. The "prebuilt application" framing describes the starting point; the path from deployment to production value requires C3.ai personnel to configure, customize, and support each installation. This is a professional services model with software company valuations applied to it.

Company	GAAP Gross Margin	Revenue Growth (YoY)	Sales Motion
C3.ai (Q3 FY2026)	17%	−46%	89% through Microsoft/AWS partners
Palantir (Q3 2025)	~80%	+30%	Direct enterprise and government
ServiceNow (FY2025)	~78%	+21%	Direct platform renewal

The partner-dependency reveals the second structural problem. 89% of C3.ai's Q2 fiscal 2026 bookings were driven "with and through" the partner ecosystem — primarily Microsoft and AWS. This concentration looks like leverage on first reading: the world's two largest cloud providers co-selling C3.ai products to their enterprise bases. In practice it means C3.ai has no independent sales motion. When the Microsoft sales team decides to prioritize its own Copilot Studio offering, or when Azure AI's own application layer matures to cover a use case C3.ai currently fills, C3.ai has no fallback channel. The alliance that generated more than 100 joint agreements and $130 million in C3.ai bookings in one year is a business development relationship, not a moat. It persists exactly as long as Microsoft's commercial interests align with sending customers to C3.ai rather than retaining them on Azure's own AI services. Microsoft's $13 billion investment in OpenAI and its Copilot products suggest the direction of that alignment.

C3.ai's fiscal third quarter 2026 results delivered $53.3 million in total revenue — 46.1% below the same period a year earlier and 30% below the $75.9 million consensus estimate. GAAP gross margin was 17%; on a non-GAAP basis stripping out stock-based compensation, it was 37% — a fifteen-point sequential collapse from Q2's 54%. Free cash flow was negative $56.2 million. GAAP net loss was $133.4 million for the single quarter, equivalent to a $534 million annualized loss rate. Cash and marketable securities totaled $621.9 million at quarter-end, implying fewer than three years of runway at the current burn rate before a capital raise becomes unavoidable. The full-year fiscal 2026 revenue guidance has been revised three times: from $447.5 million to $484.5 million when the year was set, to $289.5 million to $309.5 million midyear, to $246.7 million to $250.7 million after Q3. The midpoint has moved from $466 million to $248.7 million — a 47% reduction inside a single fiscal year. A company that cannot predict its own revenue within 50% over twelve months does not have a business model that produces predictable, contractible demand. It has episodic, project-specific revenue dependent on individual deal close rates — which is what professional services businesses generate.

Tom Siebel's departure compounds the structural problem with a governance one. In August 2025, Siebel was diagnosed with an autoimmune disease that has left him visually impaired; he transitioned from CEO to Executive Chairman while the company conducts an ongoing CEO search. Siebel controls approximately 52.4% of voting power through Class B shares. The new CEO — whoever is ultimately selected — will operate the company without the founder's operational leadership while subject to the founder's governance constraints. Strategic alternatives, capital allocation changes, or organizational restructurings that Siebel would not sanction cannot occur regardless of the new CEO's mandate. The result is a leadership vacuum with a structural ceiling on how it can be resolved. Siebel also sold $7.6 million in company stock during the period of declining performance — an act of insider selling at a company generating no free cash flow, by the controlling shareholder who has the most complete information about why the revenue trajectory is what it is. The 26% global workforce reduction — approximately 280 employees — targeting $135 million in annual savings is the largest capital allocation decision of the past twelve months, and it is a defensive response to missed guidance, not a strategic investment made from a position of strength.

Quarter	Revenue ($M)	GAAP Gross Margin	Non-GAAP Gross Margin	Free Cash Flow ($M)
Q3 FY2025	~99	~50%	~55%	negative
Q4 FY2025	~105	~50%	~55%	negative
Q1 FY2026	~67	~50%	~52%	negative
Q2 FY2026	75	~52%	~54%	(46.9)
Q3 FY2026	53	17%	37%	(56.2)

The quarterly trajectory in this table is more alarming than the absolute numbers. Revenue had been building through the late fiscal 2025 quarters — the period of peak AI investment enthusiasm when C3.ai's partner ecosystem was generating qualified pipeline — and then collapsed sequentially from Q4 FY2025 through Q3 FY2026 to a level 50% below the peak. The simultaneous gross margin collapse from 50% GAAP to 17% in a single quarter is not a step-down in the cost of a product or service — it reflects a quarter in which the company delivered a significantly higher share of its revenue through high-cost human services work rather than software licensing. This happens when deal mix shifts toward larger, more complex deployments, which is precisely what C3.ai said it was targeting: "large-scale enterprise-wide transformations." The larger the deal, the more customization required, and the more the economics look like a systems integration project. That is the growth strategy revealing its structural consequence.

The penetration argument for C3.ai is difficult to construct honestly. The company has not publicly disclosed a customer count or annual recurring revenue figure in the way that SaaS businesses typically do — metrics that would allow a calculation of market share. What is known: 89% of bookings are partner-mediated (meaning C3.ai does not independently own the customer relationship), the largest customers in the $1 million-plus deal tier numbered 17 in Q2, and a significant share of revenue is generated by the federal government business (which represented 45% of total Q2 bookings). The federal business — C3.ai's strongest segment, growing 89% year-over-year in bookings — is the company's best argument: large agencies like the U.S. Air Force and defense contractors are multi-year recurring customers with budget certainty and high barriers to competitive displacement. But the federal vertical is approximately $100 billion of the $115 billion total enterprise AI market, and C3.ai is capturing a single-digit percentage of it. The commercial market, which is where the bulk of enterprise AI spend will flow, is the market where C3.ai's distribution disadvantage is most acute.

At roughly $10 per share and approximately $1.5 billion in market capitalization, C3.ai trades at 6 times trailing twelve-month revenue. With $621.9 million in cash and no debt, enterprise value is approximately $900 million, representing 3.6 times the annualized revenue run-rate implied by Q3's $53.3 million quarterly result. Against Palantir at 30 times trailing revenue, C3.ai's multiple looks superficially attractive. The comparison is misleading. Palantir is generating 30% revenue growth with 80% gross margins and positive free cash flow. C3.ai is generating revenue that declined 46% in Q3, GAAP gross margins at 17%, and negative free cash flow of $56 million in the same quarter. The discount to Palantir's multiple is not an arbitrage opportunity — it reflects an accurate market assessment of structurally different businesses. A company burning $200 million or more annually with revenue in decline is not cheap at 3.6 times enterprise value to revenue; it is a company whose revenue multiple will fall further as the cash burns down and dilutive financing approaches.

For the investment case to reverse, three things would need to change simultaneously: the gross margin collapse would need to prove transitory rather than structural, with GAAP margins recovering toward 50% in subsequent quarters as the deal mix normalizes; revenue would need to stabilize and reaccelerate despite the competitor dynamic and the leadership disruption; and the CEO search would need to produce a candidate with both the operational credibility to rebuild the sales organization and the political ability to operate constructively under Siebel's voting control. Each of these is possible individually. All three occurring simultaneously, without a partner disruption from Microsoft or AWS, represents a combination of outcomes that the current evidence does not support betting on.

The intelligent bear argues that even the analysis above may be too generous: the 17% GAAP gross margin in Q3 is not an anomaly but a preview of what C3.ai's unit economics look like when AI application building becomes trivially cheap — at which point every marginal deal requires more human services to differentiate, compressing margins further as the category scales. The counter, for what it is worth, is that the federal business does not follow the same competitive dynamics as commercial and may continue to generate sticky recurring revenue at improving margins as deployment matures. That is a real argument. It is an argument for a much smaller company focused exclusively on federal and defense customers, not for the $1.5 billion market cap enterprise AI platform story C3.ai is currently selling.

The enterprise AI opportunity is real; C3.ai's position within it is not.