Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Evaluate an AI company in three linked passes: test the quality of its revenue, test whether margins and cash flow improve as it grows, and test how much bargaining power sits with a few customers or suppliers. Start with the latest 10-K and 10-Q, not a presentation. Read the business description, risk factors, MD&A, segment information, revenue-recognition notes, financial statements and non-GAAP reconciliations. Record the filing date and the company’s definition for every metric. Fast AI growth can show real demand without proving durable economics; a high gross margin can hide infrastructure spending or stock-based compensation; and a major customer can change orders or pricing quickly. Your conclusion should state what must remain true for growth and margins to persist, the downside case if those assumptions fail, and the evidence that would invalidate the thesis.
What should you collect before evaluating an AI company?
Download the newest Form 10-K and Form 10-Q and note the reporting period, filing date and fiscal-year end. Public companies often describe AI activity in marketing language, while their audited statements report broader segments. Your first job is to connect the claims to reported numbers—or clearly mark claims that cannot be reconciled.
As an Amazon Associate I earn from qualifying purchases.
Read the sections that change the investment case
- Business: identify products, customer types, sales channels, contract terms and where AI is actually used.
- Risk Factors: look for dependence on hyperscalers, cloud providers, chip suppliers, energy, distributors, data licensors and a small number of buyers.
- MD&A: find management’s explanation for growth, mix, pricing, costs, cash flow and changes between periods.
- Financial statements and notes: inspect revenue categories, contract liabilities, segment results, inventory, commitments, debt and stock compensation.
- Revenue-recognition disclosures: determine when subscription, usage, project and hardware revenue is recorded.
- Non-GAAP reconciliations: identify which costs management excludes and calculate the effect on operating income, cash flow and per-share results.
- Legal proceedings and commitments: check disputes, purchase obligations, capacity reservations and other liabilities that can consume cash.
Keep a dated worksheet. A “gross margin” or “AI revenue” figure without its period, segment, currency, accounting basis and definition is not comparable evidence.
How do you tell whether AI revenue growth is real and durable?
Separate revenue that repeats from revenue that arrives once. Recurring subscriptions, renewals, repeat purchases and contracted usage provide a different foundation from a one-time model-training project, installation, consulting engagement or data-center build-out. Usage-based revenue can be durable when customers repeatedly consume the service, but it can also fall quickly if workloads are optimized, budgets tighten or a buyer switches providers.
#1 Best Overall
Classify the revenue stream
- Subscription: record contract length, renewal terms, cancellations and expansion within existing accounts.
- Usage-based: track volume, price per unit, customer retention and whether growth comes from more customers or heavier use by a few customers.
- Repeat product sales: distinguish replacement cycles and normal replenishment from a temporary capacity build-out.
- Projects and services: separate implementation, consulting, customization and milestone revenue from software or platform revenue.
- Acquisitions: identify how much growth came from acquired businesses rather than the company’s pre-existing operations.
Then reconcile the classification with the financial statements. If management says AI revenue is growing but the company reports only a broad segment, use segment trends, contract commentary and customer disclosures as imperfect proxies. State explicitly when AI revenue is not separately reported; do not turn a marketing percentage into an audited fact.
What do current growth statistics actually show?
Standard Metrics reported that in the fourth quarter of 2025, AI companies in the $1 million–$20 million annualized-revenue cohort had median year-over-year growth of 73%, compared with 39% for non-AI companies. The top-decile figures were 684% and 388%, respectively. For companies above $20 million in annualized revenue, median growth was around 27%–30%. These figures describe cohorts, not a forecast for any individual company. They show that AI demand can produce exceptional early growth, but they do not establish retention, pricing power, margins or cash generation.
| Standard Metrics cohort, Q4 2025 | AI companies | Non-AI companies |
|---|---|---|
| Median year-over-year growth, $1 million–$20 million annualized revenue | 73% | 39% |
| Top-decile year-over-year growth, $1 million–$20 million annualized revenue | 684% | 388% |
| Median growth above $20 million annualized revenue | Approximately 27%–30% | Approximately 27%–30% |
Test growth against cash and funding
For each period, compare reported revenue with operating cash flow and free cash flow. Growth that requires large infrastructure purchases, subsidies, acquisitions, debt or repeated share issuance may be demand-led but not self-funding. Examine accounts receivable and contract assets for sales that have not converted to cash, deferred revenue for cash collected before delivery, and capital expenditure for compute, networking and data-center capacity.
Calculate free-cash-flow conversion as free cash flow divided by revenue or by operating income, but keep the definition constant. A company may label free cash flow as operating cash flow less capital expenditure while excluding other cash commitments. Reconcile the company’s measure to the cash-flow statement and show both the reported and adjusted versions.
Which margin measures matter for an AI company?
Track gross margin, operating margin and free-cash-flow conversion across at least three to five reporting periods. A single quarter can be distorted by product mix, launch costs, inventory charges, cloud commitments or a large customer shipment. Explain every material change instead of treating a percentage as permanent.
Gross margin: what is left after delivering the product?
Gross margin is revenue minus cost of revenue, divided by revenue. For AI businesses, cost of revenue can include GPUs and other hardware, cloud compute, data-center power and depreciation, data licensing, inference or training costs, support and inventory charges. A rising mix of lower-margin hardware can reduce the percentage even while gross profit dollars increase. Conversely, a temporary software or service mix shift can make margins look better than the long-run economics.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Operating margin: can the business support its overhead?
Operating margin subtracts research and development, sales and marketing, and general and administrative costs from gross profit. Read the expense notes for headcount, infrastructure leases, amortization and stock-based compensation. If management highlights a non-GAAP operating margin, reconcile it to GAAP and quantify what is removed. Stock compensation is non-cash at grant, but it dilutes existing owners and can create a recurring economic cost when employees must be rewarded with new shares.
Free-cash-flow conversion: does accounting profit become cash?
Compare operating cash flow with net income and operating income, then subtract capital expenditure. Investigate working-capital movements, customer prepayments, inventory purchases, supplier financing and acquisitions. A company can report positive operating cash flow because customers paid deposits while it is committing to years of capacity, or show weak cash flow because it is building inventory ahead of demand. Neither pattern should be interpreted without the commitments and timing.
What do Microsoft and AMD illustrate?
Microsoft’s fiscal-2025 annual report reported $281.724 billion of revenue, $193.893 billion of gross margin in dollars and $128.528 billion of operating income. Microsoft Cloud’s gross-margin percentage decreased to 69% as the company scaled AI infrastructure, despite strong cloud growth. The example shows why demand growth and gross-margin percentage can move in opposite directions when compute and data-center costs rise.
AMD reported 2025 revenue of $34.6 billion, up 34% from 2024, and a 50% gross margin versus 49% in 2024. AMD also disclosed approximately $440 million of net inventory and related charges associated with U.S. export controls on MI308 GPUs. The charge demonstrates why an investor must read inventory and regulatory disclosures before assuming that a reported margin is sustainable.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches| Company and period | Revenue | Gross-margin evidence | Important qualification |
|---|---|---|---|
| Microsoft, fiscal 2025 | $281.724 billion | Gross margin of $193.893 billion; Microsoft Cloud gross-margin percentage 69% | Cloud percentage decreased as AI infrastructure scaled |
| AMD, 2025 | $34.6 billion, up 34% from 2024 | 50%, versus 49% in 2024 | Approximately $440 million of net inventory and related MI308 export-control charges |
How much customer concentration is too much?
There is no universal customer-concentration cutoff. The risk depends on the customer’s bargaining power, contract duration, switching ability, segment economics and whether the buyer is direct or hidden behind a distributor or cloud provider. A large account can be attractive if it signs a long contract with minimum commitments; the same percentage can be dangerous when orders are cancellable and the buyer can build internally.
Read concentration footnotes precisely
Record each customer’s share of revenue, the period covered, whether the number is estimated, the affected segment and whether the customer is direct or indirect. Do not compare a quarterly percentage with an annual percentage. A customer that represents 20% of one quarter may be less significant over a full year, while a customer hidden in an indirect channel may still control the end demand.
NVIDIA disclosed that in the second quarter of fiscal 2026, one direct customer represented 23% of total revenue and a second represented 16%. Its filing warned: “We have experienced periods where we receive a significant amount of our revenue from a limited number of customers, and this trend may continue.” The figures are a reminder that even a company with broad end-market demand can depend heavily on a few direct buyers in a particular period.
Intel reported that its three largest customers accounted for 43% of net revenue in 2025, 45% in 2024 and 40% in 2023. Intel warned that losing key customers, materially reducing sales or changes in order timing can reduce revenue, increase volatility and harm results and financial condition. The multi-year percentages show why concentration should be tracked as a trend rather than judged from one quarter.
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| Company and period | Concentration disclosed | How to interpret the figure |
|---|---|---|
| NVIDIA, fiscal Q2 2026 | One direct customer: 23% of total revenue; second direct customer: 16% | Quarterly direct-customer concentration; order timing and bargaining power matter |
| Intel, 2025 | Three largest customers: 43% of net revenue | Annual concentration across the company |
| Intel, 2024 | Three largest customers: 45% of net revenue | Annual comparison with 2025 |
| Intel, 2023 | Three largest customers: 40% of net revenue | Annual comparison with 2024 and 2025 |
Ask who has the leverage
- Can the customer switch suppliers without rewriting software, retraining models or redesigning hardware?
- Can the customer delay orders, cancel capacity or negotiate lower prices at renewal?
- Could the customer build the capability internally or use a competing cloud, chip or model provider?
- Does the company rely on one hyperscaler, distributor, foundry, cloud provider, energy supplier or data source?
- Are commitments take-or-pay, cancellable, short-term or dependent on the customer’s own end demand?
- Would losing the account leave unused inventory, leased capacity or specialized employees?
Concentration becomes more dangerous when the same counterparty is both a major customer and a strategic supplier, lender or investor. Map those relationships in the notes and related-party disclosures, not only in the customer table.
How should you compare two AI companies?
Use identical periods and definitions before ranking companies. One company may report adjusted gross margin excluding stock compensation while another reports GAAP; one may include hardware in a cloud segment while another reports it separately. Rebuild the comparison from filings where possible.
| Comparison axis | Questions to answer |
|---|---|
| Revenue growth | What is the rate, and how much is recurring, usage-based, project-based or acquisition-driven? |
| Gross margin | What is the level and direction, and are changes explained by compute, data-center costs, inventory, mix or pricing? |
| Operating margin and cash | Does operating income convert to free cash flow after capital expenditure and stock compensation? |
| Customer concentration | Which named customers, segments, geographies and direct or indirect channels drive revenue? |
| Retention and expansion | Are renewals, repeat usage, contract duration and switching costs disclosed? |
| Partner dependence | How dependent is the company on hyperscalers, chip suppliers, cloud providers, energy, distributors or a few partners? |
| Funding and dilution | How much debt, capital expenditure and share issuance is required to sustain growth? |
| Accounting consistency | Are GAAP and non-GAAP definitions comparable across companies and periods? |
When a metric is missing, do not fill the gap with an industry average. Mark the limitation in your model and lower the confidence of the conclusion. A company that does not disclose AI revenue may still be investable, but the thesis must rely on less direct evidence such as segment trends, contracts and cash generation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What is a practical step-by-step diligence process?
- Download the newest 10-K and 10-Q. Note the filing date, reporting period, fiscal year and any subsequent events.
- Read Business, Risk Factors, MD&A, financial statements, notes, segment data and legal proceedings. Highlight every statement about AI demand, capacity, pricing, customers, suppliers and commitments.
- Build a three-to-five-period table. Include revenue, gross margin, operating income, operating cash flow, free cash flow, capital expenditure, stock compensation and share-count dilution.
- Classify revenue. Separate subscriptions, usage, repeat purchases, projects, installations, consulting, hardware and acquired revenue. Tie each category to a reported segment or explain why it cannot be tied.
- Reconcile AI claims. Flag percentages and customer counts that have no corresponding financial-statement line, definition or period.
- Explain margin movement. Attribute changes to product mix, compute, data-center costs, cloud costs, inventory charges, regulation, pricing, sales efficiency or stock compensation.
- Map concentration and counterparties. Record customer shares, direct or indirect status, segment, geography, contract terms and supplier dependencies.
- Normalize peer comparisons. Use the same period, GAAP basis, revenue definition and free-cash-flow formula. Keep reported and adjusted figures side by side.
- Stress-test the thesis. Model slower growth, lower renewal or usage, a major customer order cut, higher compute costs, inventory write-downs and additional share issuance.
- Write a dated conclusion. State the assumptions that must remain true, the evidence supporting each one, the downside case and the specific filing disclosure or operating result that would invalidate the thesis.
Which warning signs deserve immediate follow-up?
- Revenue growth accelerates while operating cash flow and free-cash-flow conversion deteriorate.
- Management highlights an AI percentage that cannot be reconciled to a reported segment or revenue category.
- Gross margin falls without a quantified explanation for compute, cloud, inventory or product mix.
- Non-GAAP profit rises mainly because recurring stock compensation, amortization or restructuring costs are excluded.
- Accounts receivable, contract assets or inventory grow faster than revenue without a clear timing explanation.
- One customer or one channel supplies a large share of revenue and can delay, cancel or vertically integrate.
- Capacity commitments, debt or share issuance rise faster than the cash produced by operations.
- Definitions change between periods, making retention, bookings, adjusted profit or free cash flow appear better without a like-for-like bridge.
None of these signs proves that an investment will fail. They identify the exact disclosure, contract term or sensitivity analysis that deserves more work before you rely on the growth story.
Recommended Free Tools
How should the final investment conclusion be written?
Use an evidence-linked statement rather than a score based on one ratio. For example: revenue is growing at a stated rate, but the filing reports only a broad segment; gross margin is moving in a specified direction because of identified infrastructure or inventory costs; free cash flow covers a stated portion of investment; and named customers account for a documented share of sales. Then state what must happen next—such as renewal, price, utilization, capacity or cost conditions—and what evidence would disprove the thesis.
Refresh the analysis with each new 10-Q and 10-K because customer shares, segment definitions, infrastructure costs and contract economics can change quickly. The most useful output is a dated view of durability, not a permanent label such as “high growth” or “high margin.”
Best Value
Frequently Asked Questions
How often should I refresh an AI-company investment analysis?
Review it after every quarterly filing and immediately after a material acquisition, financing, capacity commitment or major customer announcement. Recheck the definitions of revenue, margins and customer shares each time because companies can change segment presentation or non-GAAP measures.
How should acquisitions affect an AI company’s reported growth rate?
Separate organic growth from revenue added by acquired businesses using the acquisition date and purchase-accounting disclosures. A high headline rate may fall sharply once the acquired contribution is removed, and integration costs or purchase-price amortization can affect both margins and cash flow.
Is there a standard customer-concentration percentage that makes an AI company unsafe?
No single percentage works across business models. Evaluate the customer’s contract duration, cancellation rights, switching cost, ability to build internally, renewal history and effect on unused capacity; the same share can be manageable in one contract structure and dangerous in another.
Can I evaluate a private AI startup without public-company filings?
Ask for monthly revenue by product, cohort retention, gross-margin calculations that include compute and data costs, cash burn, customer contracts and cap-table dilution. Request definitions and source periods for every metric, then apply the same separation of recurring revenue, project work, concentration and funding needs used for public companies.
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.
