AI stocks are everywhere. But not all revenue growth is real. Some numbers look great on paper but fall apart under closer look. This guide shows you simple ways to find the truth.
Where AI Revenue Actually Comes From
Revenue sources tell you if growth is solid or shaky. Some AI companies earn from real products. Others rely on one-time deals or accounting tricks.
| Revenue Type | What It Means | How Reliable |
|---|---|---|
| Subscription | Customers pay monthly or yearly | Very reliable |
| Usage-based | Pay per API call or compute hour | Reliable if usage grows |
| Enterprise contracts | Big deals with companies | Good but check renewal rates |
| Licensing | Others pay to use your technology | Watch for one-time spikes |
| Hardware sales | Selling chips or servers | Reliable but cyclical |
| Professional services | Consulting and setup help | Low margin, less repeatable |
Palantir makes most money from long government and business contracts. Investors like this because it repeats.
Some AI startups show huge growth from one custom project. That growth rarely lasts.
Look for income that comes back every quarter. One-time sales can hide weak business health.
Red Flags in Revenue Reporting
Companies sometimes use tricks to make growth look better. Knowing these tricks keeps your money safer.
| Red Flag | What to Look For | Why It Matters |
|---|---|---|
| Revenue recognition changes | Company changes how it counts sales | May inflate current numbers |
| Related-party transactions | Sales to sister companies or investors | Not true market demand |
| Channel stuffing | Push inventory to resellers early | Pulls future sales forward |
| Round-trip deals | Swap services with another company | No real value created |
| Deferred revenue drops | Future payments suddenly shrink | Sign of slowing new business |
| Customer concentration | One client is over 10% of sales | Lose one client, lose big |
In 2022, some AI firms booked revenue before products were delivered. When delivery failed, revenue had to be reversed.
Always check if growth comes from more customers or just bigger deals from the same few.
Smart investors read the footnotes. That is where companies hide the messy details.
Metrics That Show Real Growth
Headline revenue is not enough. These numbers dig deeper into business health.
| Metric | What It Shows | Healthy Range |
|---|---|---|
| Net Revenue Retention (NRR) | Do existing customers spend more? | Above 110% is strong |
| Contracted Remaining Performance Obligation (cRPO) | Future revenue already signed | Growing faster than revenue |
| Gross margin expansion | Is the business getting more efficient? | Stable or improving |
| Customer acquisition cost (CAC) payback | How fast new customers pay off | Under 18 months |
| Remaining Performance Obligation (RPO) | Total future contracted revenue | Growing consistently |
| Rule of 40 | Growth rate plus profit margin | Combined above 40% |
Snowflake reports NRR above 130%. This means old customers keep buying more. That is real growth, not just new logo hunting.
Compare this to companies where RPO stalls. It means customers sign up but do not renew or expand.
If existing customers spend more each year, the product is sticky. New customer growth alone is weaker.
Comparing AI Companies Fairly
Different AI business models need different checks. A chip maker is not the same as a software firm.
| Business Model | Key Check | Watch Out For |
|---|---|---|
| AI Infrastructure (chips, cloud) | Capacity utilization rates | Inventory write-downs |
| AI Software (SaaS) | Net revenue retention | Short contract lengths |
| AI Services (consulting) | Backlog and pipeline | Heavy reliance on subcontractors |
| AI Vertical (healthcare, finance) | Regulatory clearances | Pilot projects never converting |
| AI Platforms (marketplaces) | Take rate stability | Subsidy-dependent growth |
NVIDIA sells chips. Investors check data center revenue and inventory levels. Software firms like Datadog get checked for how many modules each customer buys.
Do not use the same ruler for every company. It leads to bad comparisons.
Key Takeaways
| Key Point | What It Means | Action Item |
|---|---|---|
| Recurring revenue beats one-time sales | Predictable income shows real demand | Check 10-K for subscription vs. services split |
| Net Revenue Retention (NRR) above 110% is strong | Customers stay and spend more | Find NRR in earnings slides or investor relations |
| Watch for revenue recognition changes | Companies can shift when they count sales | Compare footnotes year over year |
| cRPO growth signals future health | Shows booked but not yet recognized revenue | Track cRPO trend across quarters |
| Match metrics to business model | Chip makers and SaaS firms differ | Use sector-appropriate benchmarks |
| Customer concentration is risky | One big client leaving hurts badly | Check if any customer is over 10% of revenue |