Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Before buying an AI company’s stock, verify what the company actually sells, whether customers are paying for it repeatedly, how much it costs to deliver, how long its cash can last, and whether competition or regulation can erode its advantage. Start with the issuer’s latest annual and quarterly filings, then test revenue quality, margins, infrastructure spending, financing and dilution, risk disclosures, valuation, and your own ability to tolerate loss. The framework below applies differently to a pure-play AI vendor, a diversified technology company, and a business that mainly uses AI internally.
1. Identify what kind of AI company you are evaluating
“AI company” can describe several very different businesses. A specialist may sell models, software applications, data services or infrastructure. A diversified technology company may earn revenue from many products while investing heavily in AI. An insurance, manufacturing, healthcare or services company may use AI to reduce costs without selling AI as a separate product. Those models have different revenue drivers, capital needs and risks, so classify the issuer before comparing its stock with anything else.
| Business type | What usually drives results | Questions to ask first |
|---|---|---|
| Pure-play AI vendor | Subscriptions, usage fees, licenses, implementation and support | Is demand recurring or project-based? How concentrated are customers? Can the company price above infrastructure and support costs? |
| Diversified technology company with AI products or investment | A mix of established businesses plus AI-related product and cloud demand | How much of growth and capital spending is genuinely AI-related? Are AI investments lifting returns across the group or mainly increasing costs? |
| Company using AI in its operations | Revenue from its traditional business, with AI affecting productivity, service quality or risk | Is there measurable operating improvement? What new privacy, compliance, workforce or model-dependence risks are introduced? |
Do not treat an AI announcement, a model release or a large capital-spending plan as proof of an attractive stock. The investment case must connect the technology to durable customer demand, cash generation and an acceptable valuation.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →2. Read the primary filings in context
The latest annual report is the anchor document. Use the most recent quarterly report to update the picture for changes since the year-end filing. Read the business description, management’s discussion and analysis, financial statements and notes, and every risk-factor section that relates to the company’s AI activities.
#1 Best Overall
- Comes with secure packaging
- Easy to read text
- It can be a gift option
- Confirm the reporting period and issuer. Record the fiscal year-end, filing date, legal entity and business segments. A comparison is unreliable if one company reports a calendar year and another reports a different fiscal period.
- Map the business model. From the business section, list products, customer types, geographies, distribution channels and revenue sources. Separate AI products from cloud, hardware, consulting or non-AI operations.
- Reconcile management’s narrative with numbers. Compare claims about adoption, growth or efficiency with reported revenue, deferred revenue, remaining performance obligations, customer counts, usage measures and cash flows where disclosed.
- Read the notes, not only headline totals. Accounting policies, segment reporting, stock compensation, commitments, debt terms, legal proceedings and customer concentration can materially change how a headline result should be interpreted.
- Read risk factors in full. A summary is not a substitute for the detailed disclosure. Veritone’s FY2025 filing explicitly says its risk summary does not contain all information that may be important and directs readers to the detailed risks and the rest of the annual report. See the Veritone FY2025 Form 10-K.
Use investor presentations and earnings-call transcripts as indexes to questions, not as replacements for audited filings. When management uses a new metric, locate its definition, reconciliation and period-to-period history before relying on it.
3. Test whether revenue reflects durable demand
Separate reported growth from revenue quality
Chart revenue by quarter and year, then split it by product, geography, segment and customer type when the issuer provides that detail. A fast-growing total can conceal a shrinking core product, a one-time implementation project or an acquisition that has not yet produced organic growth. Note whether growth comes from higher prices, more customers, greater usage, new products or acquisitions.
Recurring subscription revenue is not automatically safe: customers can cancel, reduce seats or negotiate lower prices. Usage-based revenue can expand rapidly but may fall when customers optimize workloads. Services revenue can help adoption while carrying lower margins. Identify the mix and track how it changes rather than assigning a quality label to the company as a whole.
Recommended Free Tools
Look for evidence that customers are actually using the product
Read customer-contract descriptions, renewal statistics, dollar-based net retention, backlog, bookings, remaining performance obligations and concentration disclosures if they are reported. Ask whether the company counts pilots, free users, paid production deployments or consumption. A large announced partnership is weaker evidence than recognized revenue, repeat purchases and cash collected from customers.
Check receivables and deferred revenue alongside sales. Receivables growing much faster than revenue can indicate slower collections or aggressive terms; deferred revenue can show that customers paid before delivery, but its release into revenue still depends on performance obligations. These signals require the issuer’s accounting context, not a mechanical rule.
Examine margins and operating expenses together
Gross margin shows what remains after direct delivery costs. For AI products, direct costs may include model-compute charges, data licensing, cloud capacity, hardware depreciation, implementation labor and customer support. A high gross margin can deteriorate if usage rises faster than pricing or if the company absorbs infrastructure costs to win customers.
Rank #2
- Ideal for Gifting
- Ideal for a bookworm
- Comes with Proper Binding
Track research and development, sales and marketing, and general and administrative expense as a percentage of revenue. Management may rationally invest ahead of demand, but the filing should explain the milestones expected to convert that spending into revenue or improved margins. Distinguish cash expense from non-cash stock compensation and assess both when estimating the economic cost of growth.
Free tools Windows power users keep installed
One-click scans. No signup required.
Understand what would have to change for sustainable profitability
Ask management’s stated path to profitability: higher gross margin, slower hiring, lower customer-acquisition costs, more efficient inference, improved retention or a larger revenue base. Then test whether the required change is visible in historical trends and whether it depends on assumptions the company does not control, such as rapid adoption or falling compute prices.
C3.ai’s FY2026 filing illustrates why a compelling product story is not the same as profitable demand. The company reported net losses of approximately $470.4 million in FY2026, $288.7 million in FY2025 and $279.7 million in FY2024. These are company-reported, issuer-specific historical figures, not an AI-sector benchmark. Its filing discusses demand, consumption, competition, forecasting, expenses and the uncertain timing of profitability; read the C3.ai FY2026 Form 10-K for the company’s definitions and qualifications.
“We have a history of losses, we anticipate our operating expenses will continue to increase, and we may not be able to achieve or maintain profitability in the future.”
— C3.ai, Inc., FY2026 Form 10-K risk disclosure
4. Test competitive durability, not just technical capability
Define the customer’s reason to stay
Identify the workflow the product improves and the measurable cost, revenue, speed, accuracy or compliance benefit it provides. Then ask what would make a customer replace it: a cheaper model, an integrated platform, an internal team, an open-source system or a change in regulation. A technically impressive feature without switching costs can be vulnerable to rapid price competition.
Assess dependence on platforms and suppliers
Determine whether the company relies on a small number of cloud providers, chip suppliers, data licensors, distribution partners or foundation-model providers. Dependence can affect gross margin, availability, bargaining power and product roadmap. Conversely, a company that owns scarce data, controls a distribution channel or embeds software deeply in a regulated workflow may have more durable advantages. Treat each claimed advantage as a hypothesis to verify in contracts, retention data and margins.
Read the issuer’s own competition warnings carefully
Microsoft describes AI as a highly competitive, rapidly evolving market and discusses risks involving demand, costs, pricing, customer adoption and returns from substantial AI and cloud investment in its FY2026 filing. Veritone states that the market for AI-based software applications is relatively new and unproven and may decline or experience limited growth. These are company risk disclosures, not predictions that every AI business will fail or succeed. Review the Microsoft FY2026 Form 10-K alongside the issuer you are evaluating.
Compare the company with realistic alternatives, not only with other public AI stocks. The relevant alternative may be a customer’s existing software suite, an internal engineering team, an open-source model or a conventional non-AI process. If customers can switch without meaningful data migration, retraining or regulatory requalification, assume pricing power is limited unless evidence shows otherwise.
5. Calculate the cost-and-return equation
AI economics can fail even when demand is real if the cost of serving each request rises with usage. Build a simple unit-economics view using the issuer’s disclosures:
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minute- Revenue per customer, seat, transaction or unit of usage.
- Direct compute, hosting, data, hardware depreciation and support costs.
- Gross profit after those costs.
- Sales, implementation and customer-success cost needed to win and retain the account.
- Research and development required to keep the product competitive.
- Capital expenditure and working capital required before revenue arrives.
Use a range rather than a single forecast. Model what happens if model inference becomes more expensive, customers consume less than expected, prices fall, or a new model requires another round of training and deployment. Also model the opposite case: higher utilization may spread fixed infrastructure costs over more revenue, but only if pricing and service levels remain acceptable.
Separate operating expense from infrastructure investment
Research and development appears in the income statement, while data centers, servers and other equipment may appear in capital expenditure and depreciation. A company can report improving operating margins while consuming substantial cash for infrastructure. Microsoft’s FY2026 filing describes substantial AI and cloud investment and related risks from evolving demand, costs, pricing, adoption and returns. Scale those questions to the issuer’s own model: a software specialist may rent capacity, while a diversified provider may own or finance infrastructure.
Ask when returns arrive
For every major AI investment, identify the expected revenue stream, utilization assumption, useful life, financing source and return hurdle. If management provides no measurable milestones, treat the spending as an uncertainty rather than an asset that will automatically produce growth. Compare cash paid for capital expenditure with cash generated from operations, not only with non-GAAP earnings.
6. Review cash, debt, financing and dilution
Measure liquidity and runway
Start with cash and short-term investments, then subtract restricted cash and obligations that cannot readily fund operations. Compare the result with operating cash burn and committed capital spending. Runway is not a promise: it changes when hiring, infrastructure purchases, collections or customer demand changes.
Read debt terms and financing plans
List principal maturities, interest rates, covenants, secured assets, conversion features and refinancing dates. Rising rates or a weak share price can make the next financing expensive. Veritone identifies indebtedness and liquidity among its risk categories; use that disclosure as a prompt to inspect the specific issuer’s debt notes rather than assuming the same exposure elsewhere.
Account for dilution
Track basic shares, diluted shares, options, restricted stock, warrants, convertible securities and shares reserved for employee plans. A company can report revenue growth per share that is weaker than total revenue growth if the share count rises. Include likely equity compensation and future capital raises in your ownership analysis. C3.ai’s filing reports an accumulated deficit of approximately $1.8 billion as of April 30, 2026; that is a historical company-reported balance, not a forecast of its future financing needs.
7. Read risk disclosures beyond product performance
Organize risks by the company’s role as AI developer, infrastructure provider, deployer or user. The relevant exposure depends on products, customers, geography and data flows.
- Intellectual property: Check ownership of training data, model weights, code and customer outputs; review infringement claims, indemnities and licensing restrictions.
- Data rights and privacy: Determine what personal or confidential information is processed, where it is stored, how consent is obtained and whether customers can delete or export it.
- Cybersecurity and misuse: Look for incidents, safeguards, abuse monitoring, model access controls and obligations to notify customers or regulators.
- Regulation: Identify rules affecting employment, healthcare, finance, consumer protection, copyright, export controls and automated decision-making in the company’s markets. Microsoft discusses AI misuse, regulation and public scrutiny in its filing.
- Customer concentration: A few large contracts can make revenue, renewals and bargaining power fragile.
- Execution and leadership: Review departures, restructuring, delayed products, internal-control weaknesses and whether promised milestones were met.
- Governance: Examine board oversight, related-party transactions, voting rights, incentive structures and how management reports failures or model limitations.
Risk-factor language is often broad because companies must describe uncertainties. Give greater weight to risks that appear in the financial statements, have already caused an incident, or are tied to a concentrated customer, supplier or contract.
8. Compare companies on consistent measures
When comparing issuers, use the same fiscal periods, accounting definitions and denominator. A diversified company’s total revenue and margin cannot be compared directly with a specialist’s AI segment unless the segment boundaries are clear.
Best Value
| Comparison axis | Evidence to collect | Why it matters |
|---|---|---|
| Revenue mix and growth | Product, segment, geography, recurring versus usage-based revenue, organic growth and customer concentration | Shows where demand comes from and how repeatable it may be |
| Profitability | Gross margin, operating loss, adjusted measures, stock compensation and operating cash flow | Separates accounting presentation from cash economics |
| Cash and capital structure | Cash, debt, maturities, covenants, share count, options and convertibles | Indicates financing risk and potential dilution |
| Infrastructure and research | Research expense, capital expenditure, depreciation, hosting commitments and energy or component exposure | Shows what growth costs before it produces returns |
| Customer durability | Retention, renewals, backlog, implementation time, switching costs and alternative suppliers | Tests whether competitive advantage can persist |
| Legal and regulatory exposure | IP disputes, data rights, privacy, cybersecurity, compliance and geographic restrictions | Identifies liabilities that may impair growth or margins |
| Valuation | Market value or enterprise value relative to growth, margins, cash needs and realistic scenarios | Determines what expectations are already embedded in the price |
Do not fill missing cells with assumptions. If an issuer does not disclose a comparable measure, mark it as not stated and lower confidence in the comparison.
9. Consider valuation only after operating analysis
Valuation is the last major check, not the first. A high growth rate does not establish fair value if gross margin is falling, cash burn is rising or the growth requires repeated share issuance. Conversely, a low sales multiple may reflect weak retention, legal exposure or a business model that needs large infrastructure investment.
Choose a metric that fits the company’s maturity. Revenue or enterprise-value-to-revenue can be useful for an early-stage loss-maker, but only alongside growth quality, gross margin and cash burn. Earnings multiples require sustainable earnings. Free-cash-flow measures require attention to capital expenditure, working-capital swings and stock compensation. Compare scenarios rather than presenting a single precise target: slower adoption, lower prices, higher compute costs and additional dilution should each have a visible effect.
No issuer or current market price is specified here, and no comparable valuation evidence supports a buy or sell conclusion. Before acting, date the price, share count, financial period and currency used in your calculation.
10. Decide whether the risk fits you
An attractive company can still be unsuitable for an investor who cannot tolerate large drawdowns, years without profits or dilution from future financing. Write down the maximum loss, holding period and portfolio exposure you can accept before reading the stock’s promotional narrative.
Separate three decisions: whether the business is improving, whether the current price compensates for uncertainty, and whether the position fits your finances and risk tolerance. A “yes” to the first question does not answer the other two. Avoid treating a company’s AI capability, a famous customer, or a rapidly rising share price as a substitute for this separation.
11. A repeatable diligence workflow
- Capture the baseline. Save the latest annual and quarterly filings, fiscal dates, share price date, market capitalization and enterprise-value inputs.
- Describe the business in one sentence. State who pays, for what product, under which pricing model and in which segment.
- Build a five-year operating table. Enter revenue, gross profit, operating income or loss, operating cash flow, capital expenditure and diluted shares where available.
- Trace one dollar of revenue. Estimate direct delivery cost, sales and support cost, research spending and capital required to serve it.
- Verify demand. Match customer, retention, backlog and usage claims to recognized revenue and cash collections.
- Stress the model. Reduce growth and price, increase compute and infrastructure costs, delay profitability and add a financing round with dilution.
- List disqualifying uncertainties. Record unresolved IP, privacy, regulatory, concentration, debt or governance issues and what evidence would resolve each one.
- Set a decision rule. Define the valuation range, evidence threshold and maximum position size before placing an order; document why the rule would change.
12. Red flags that deserve a pause
- Management emphasizes model quality or partnerships while giving little evidence of paid production use.
- Revenue grows, but receivables, concessions or customer concentration rise faster.
- Gross margin declines as usage increases, with no credible explanation for improvement.
- Capital spending and cash burn accelerate without stated utilization or return milestones.
- Non-GAAP profitability excludes recurring stock compensation, infrastructure costs or restructuring that investors must still fund.
- The share count rises steadily while management describes financing as optional.
- Risk disclosures mention data rights, cybersecurity, regulation or IP disputes that are absent from the investment thesis.
- The valuation depends on a distant profitability assumption that fails under modestly slower growth or higher compute prices.
A red flag is a reason to investigate, not automatic proof of failure. The key question is whether the filing provides quantified evidence, a credible mitigation plan and progress over time.
13. What a convincing case would need to show
A stronger investment case links a defined customer problem to repeat purchases, stable or improving unit economics, a funded path to profitability, manageable infrastructure commitments, sufficient liquidity and a defensible competitive position. It also explains why the current price leaves room for execution errors and why the risks are acceptable for your portfolio.
AI capability and investment alone do not demonstrate profitable demand. C3.ai’s reported losses show why issuer-specific financial statements matter; Microsoft’s and Veritone’s disclosures show how competition, infrastructure spending, uncertain adoption, regulation and intellectual-property issues can affect both growth and cost. Use those filings as examples of questions to ask, then rely on the named company’s own current reports for the answer.
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.
