Before investing in an AI company, test whether it can repeatedly solve a valuable customer problem, charge enough to cover the full cost of delivering that solution, and turn revenue into cash without relying on continual outside funding. Start with the company’s own latest filings and audited statements. Separate evidence of customer demand from announcements, and separate business quality from whether the stock’s price is attractive.
What does the company actually sell, and who pays for it?
“AI company” describes a broad category, not a business model. A company may sell access to a foundation model, an application built on a model, cloud or compute infrastructure, implementation and consulting, or a bundle of software and services. Those products can have very different buyers, costs, and revenue patterns.
As an Amazon Associate I earn from qualifying purchases.
Write a one-sentence description: “The company sells [product or service] to [buyer] to solve [problem], and charges by [subscription, seat, usage, license, project, or other method].” Identify the budget the buyer uses and what alternative the customer would choose without the product. If you cannot fill in the sentence from filings and product materials, the business model is not yet clear enough to assess.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Also distinguish revenue specifically attributable to AI products from revenue generated by a broader software or services portfolio. An AI label, product announcement, or management target is not the same thing as disclosed AI revenue or a realized customer outcome.
#1 Best Overall
How should you reconstruct the revenue model?
Read the business description, management’s discussion and analysis, audited income and cash-flow statements, revenue-recognition notes, customer-concentration disclosures, and risk factors in the annual report. Review quarterly filings too, since they can show whether recent trends differ from the full-year picture.
Map the sources of reported revenue and the obligations behind them. A subscription may be recognized over time, while a usage charge can vary with customer activity; a license, implementation project, or support contract may be recognized on a different basis. Check contract duration, renewal terms, committed versus variable usage, and whether hosting or services are bundled. Remaining performance obligations or backlog can offer context, but they are not interchangeable with recognized revenue, cash collected, or guaranteed future sales.
| Revenue type | What to establish | Why the distinction matters |
|---|---|---|
| Subscription or seat-based | Contract length, renewal behavior, included usage, and whether the customer can reduce seats or cancel | A recurring label does not by itself establish that revenue will renew or grow. |
| Usage-based or consumption | What is metered, whether customers commit to a minimum, how variable usage is, and how pricing compares with delivery cost | Revenue may rise with adoption, but it can also fluctuate or fail to cover heavy-use costs. |
| License or other product sale | When revenue is recognized, whether maintenance or support is separate, and what future obligations remain | A sale recognized upfront can produce a different pattern from a continuing service relationship. |
| Services or implementation | How much delivery depends on people, project timing, and customer-specific work | Services may help win deployments, but labor-intensive delivery can limit repeatability and margins. |
Company-specific disclosures illustrate why labels are not enough. In its fiscal year ended April 30, 2026, C3.ai reported subscriptions as 91% of total revenue; that is a figure for C3.ai and that fiscal year, not an industry benchmark. Its filing also describes consumption-based pricing and subscription revenue recognized on a ratable or consumption basis. Those details show why an investor should inspect the underlying recognition and pricing terms rather than infer revenue quality from the word “subscription.” C3.ai’s fiscal 2026 Form 10-K.
Recommended Free Tools
Is customer demand durable and repeatable?
Look for evidence that customers use the product in production and keep using it—not just that they signed a pilot, attended a demonstration, or appeared in a press release. Where disclosed, examine renewals, expansion within existing accounts, usage trends, customer outcomes, contract duration, and churn. Ask how much implementation is required before the product becomes useful and whether it becomes embedded in a critical workflow or remains easy to replace.
- Compare customer concentration across reporting periods. A large customer can validate demand, but dependence on a few accounts makes the business more exposed if one customer cuts spending or does not renew.
- Keep announcements, letters of intent, bookings, backlog, remaining performance obligations, recognized revenue, and cash receipts distinct. They describe different stages or measures of a commercial relationship.
- Check whether reported customer wins translate into production use and continued spending. A pilot or contract announcement alone does not establish long-term retention or profitable delivery.
Read a company’s own concentration and renewal disclosures alongside its numbers. C3.ai identifies customer concentration and renewals among risks in its fiscal 2026 filing, but that issuer-specific disclosure should not be generalized to every AI business. C3.ai’s fiscal 2026 Form 10-K.
What does it cost to deliver each additional unit of AI?
AI revenue can carry costs that grow with customer activity. Depending on the product and what the company discloses, examine inference, GPUs or other compute, cloud hosting, data licensing, human review, implementation, support, and model development. Determine which costs the company pays, which it passes through to customers, and which are included in a fixed-price contract.
Compare the cost to serve a customer or usage unit with the revenue earned from it. Then ask what could improve the economics: price increases, more efficient models, routing tasks to less costly models, infrastructure scale, or customer charges that better reflect usage. Do not assume a particular cost structure merely because a business uses AI; confirm it in the issuer’s filings.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →A 2026 SEC filing by GridAI Technologies Corp. identifies possible risks in AI infrastructure, including volatile usage-based revenue, subscription pricing that may not capture heavy usage, pricing below inference costs, and commoditization that can pressure prices and gross margins. These are mechanisms to investigate, not evidence that every company—or any particular target—has those problems. GridAI Technologies Corp.’s Form 10-K for the year ended December 31, 2025.
Does growth translate into margins, cash, and a viable funding plan?
Compare multiple reporting periods rather than treating one growth rate as a verdict. Put revenue next to gross profit and gross margin, operating expenses, operating cash flow, capital expenditure, stock-based compensation, debt, cash balance, and financing needs. Read management’s account of what must change before the company can fund itself. Growth alone does not show that a model is profitable or self-funding.
Losses are not automatically proof that a business model cannot work, but their size, persistence, cash impact, and funding source matter. For context—not as a current result or a sector comparison—C3.ai’s fiscal 2025 Form 10-K reported net losses of $288.7 million in fiscal 2025, $279.7 million in fiscal 2024, and $268.8 million in fiscal 2023, plus an accumulated deficit of $1.4 billion as of April 30, 2025. Those historical company figures should be read with the filing’s other financial statements and subsequent company reporting, not treated as a forecast. C3.ai’s fiscal 2025 Form 10-K.
For a private company, audited statements and consistent operating metrics may not be public. Mark unavailable figures as unknown; do not assume favorable margins, retention, or cash generation in their absence.
Free tools Windows power users keep installed
One-click scans. No signup required.
What could make the company hard to replace—and what could undermine that advantage?
Look for reasons customers would continue paying even if model capabilities and prices change. Potential sources of durability include measurable customer outcomes, workflow integration, distribution, rights to useful data, reliability, or scale. A claimed advantage matters only if it is specific, supported by evidence, and valuable to the buyer; access to a popular model alone may not distinguish an application if competitors can offer a similar feature.
Best Value
Assess alternatives and dependencies: open models, rival applications, cloud providers, model suppliers, chip suppliers, and data providers. Review privacy, security, intellectual-property, regulatory, and execution risks in the target’s own filings. Microsoft’s fiscal 2025 Form 10-K describes substantial AI development and operating costs and a rapidly evolving, competitive market. That is useful context for the market’s uncertainty, not a substitute for target-specific diligence. Microsoft’s fiscal 2025 Form 10-K.
Stress-test the business against concrete downside cases: customers delay production adoption; inference or hosting costs rise; a major customer reduces spend or does not renew; a model or cloud provider bundles a competing feature; a lower-cost or open alternative weakens pricing; or a model failure, privacy issue, or legal restriction raises costs or weakens demand. Consider what each case would do to revenue, margin, cash needs, and the company’s ability to retain customers.
How can you make the final business-model assessment?
Use the same questions for each company you compare, and base answers on disclosed evidence rather than a general impression that AI is a growing market.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- Offer and buyer: Can you name the product, the customer, the problem, and the budget that pays for it?
- Revenue quality: Can you separate recurring, variable, license, and service revenue and explain when each is recognized?
- Customer evidence: Do production use, renewals, expansion, and disclosed outcomes support repeat demand, and how concentrated is that demand?
- Cost to serve: Are AI-specific costs visible enough to judge whether pricing covers delivery as usage scales?
- Financial resilience: Do margins and cash flow support the growth plan, or does the company need continuing external funding?
- Durability and downside: Is there an evidenced reason customers stay, and can the model withstand the risks in its filings?
An unclear answer is a diligence finding, not a reason to fill the gap with an optimistic assumption. A business-model assessment also does not determine whether a stock is attractively priced: valuation, dilution, governance, investment horizon, and risk tolerance require separate analysis.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
