Research methods
How AI competition changes a software company’s valuation
The short answer
AI changes a software valuation when it changes the cash a business can generate, the investment needed to generate it, or the risk attached to those cash flows. Product launches and user counts are evidence to investigate; they do not, by themselves, establish a higher or lower value.
Start with what customers pay for
A software product can contain features that AI makes cheaper while still offering a workflow customers value. The opposite is possible too: an impressive new feature can attract users without preserving the paid product’s economics. Begin with the transaction. Who pays, what problem is solved, and what would make the buyer switch?
Separate the seat, the task and the broader workflow. If customers need fewer employees to produce the same output, seat-based revenue may face pressure. If the company charges for usage or valuable completed work, adoption may have a different effect. Those are hypotheses until the disclosed pricing, customer behavior and financial results support them.
An annual filing’s business and risk sections help identify the company’s own description of competition, pricing and technological change. They are management disclosures, not an independent experiment proving that a product is protected.
Read the source: Adobe FY2025 Form 10-K: business risks and financial statements
Translate the argument into four model inputs
The useful question is not whether a company “has AI.” It is which assumption should change. Demand affects revenue. Pricing and compute costs affect margins. Product development and infrastructure affect reinvestment. Greater uncertainty can affect the return investors require.
Keep those channels separate. Counting an AI revenue opportunity while ignoring revenue it might replace produces an incomplete bridge. Treating every infrastructure dollar as permanently recurring can be equally misleading. Describe the timing and why the spending is necessary.
- Revenue: are paid customers, paid usage or realized prices changing?
- Operating costs: what happens to inference costs, support, sales and product development?
- Investment: does growth require more capacity, capital spending or working capital?
- Risk: which assumptions are less dependable, and what evidence would resolve them?
Read the source: Aswath Damodaran, NYU Stern: valuation approaches
Use a revenue-and-margin bridge before a price target
Consider the invented software business below. Annual revenue starts at $1 billion and operating margin at 30%, producing $300 million of operating profit. The scenarios change only revenue and margin; they are not forecasts for Adobe or another issuer.
In the adoption case, higher revenue more than offsets a smaller margin. In the pressure case, weaker revenue and heavier costs compound the reduction in profit. Neither result tells you a share price. Interest, taxes, investment, cash timing and share count still sit between this operating illustration and a per-share valuation.
| Scenario | Revenue | Operating margin | Operating profit |
|---|---|---|---|
| Starting business | 1,000 | 30% | 300 |
| Paid adoption grows; delivery costs rise | 1,100 | 28% | 308 |
| Pricing pressure and heavier spending | 950 | 24% | 228 |
Read the source: SEC: Beginners’ Guide to Financial Statements
Write down the evidence that would change the view
Look for a comparable paid-revenue series, customer retention measures when disclosed, pricing changes, margin trends and cash conversion. Preserve the metric definition and period. A free-user milestone and recurring contracted revenue are different signals.
When evidence is missing, keep the gap visible. Do not manufacture paid-seat counts, AI unit margins or retention to make a model look complete. A useful research note identifies the next disclosure that could challenge the current assumption.
Our dated Adobe sample illustrates this approach: a business argument sits beside its strongest contrary reading, with explicit valuation assumptions and conditions for revisiting the thesis. It is a worked research example, not proof that future results will follow the model.
Read the source: C&C Research: selected Adobe report excerpts