How to Value a Stock With AI: DCF Explained (2026)

Federico Kunze Küllmer is the founder of Moneda, one of the tools compared here.

Short answer: A discounted cash flow (DCF) model values a company by estimating the cash it will make in the future and converting that cash into what it is worth today. An AI can pull the figures from company filings, run the maths and show you how sensitive the result is to each guess. It cannot know the future. Your job is to check the inputs, question the growth and discount rate assumptions, and treat the output as a range, not a single "true" price.

What a DCF actually is

A DCF, or discounted cash flow model, starts from a simple idea: a business is worth the cash it will hand back over its life, and money later is worth less than money now.

So you do two things:

1. Forecast free cash flow. Free cash flow is the cash a company has left after paying its running costs and the investment it needs to keep growing (equipment, buildings, software). It is closer to "real money" than reported profit.
2. Discount it. You shrink each future year's cash by a rate that reflects risk and the time value of money. A euro in ten years counts for much less than a euro today.

Add up the discounted cash flows, add a terminal value (a single number for everything after your forecast period), subtract debt, add cash, and divide by the number of shares. That gives you an estimated value per share. Compare it with the market price.

That is the whole method. The hard part is the guesses.

Why use AI for this

Building a DCF by hand means reading a 10-K (the annual report US companies file with the Securities and Exchange Commission); see our guide to reading a 10-K with AI, copying numbers into a spreadsheet and checking formulas. That takes hours, and most errors are copying errors.

An AI can do the tedious parts in minutes: find revenue, margins, capital spending and share count; lay them out by year; and run the model under several scenarios. That leaves you more time for the judgement calls, which is where valuations are won or lost.

The risk is that an AI can also produce a confident number from figures it made up. Any tool you use should show where each figure came from. If it cannot, treat the answer as a draft.

Step 1: Get the historical numbers right

Every forecast rests on history. Start by asking the AI to lay out the last five years.

Example questions:

• "Show me revenue, operating margin, capital expenditure and free cash flow for the last five years, with the source for each figure."
• "How has the share count changed over that period, and why?"
• "Were there any one-off items that make a year look better or worse than it really was?"

What to verify yourself: open the filing and spot-check two or three numbers. If revenue or free cash flow is off, everything after it is off too.

In Moneda, the agent works figures out from company filings and live market data and shows the source behind each one, so you can click through and check. That is the behaviour to look for in any tool.

Step 2: Check the quality of the earnings

Before you project profits forward, ask whether they are real. Some companies report strong profits while cash flow lags behind, which can be a warning sign.

Example questions:

• "Does operating cash flow keep pace with net income? Where does it diverge?"
• "How much of the profit depends on stock-based pay, acquisitions or accounting changes?"
• "Are receivables growing faster than sales?"

What to verify yourself: if the gap between profit and cash is large, read the explanation in the filing. A good reason (a big prepaid contract, say) is different from a vague one.

Moneda AI runs a quality of earnings check as part of its research, which flags these gaps for you.

Step 3: Decide on growth, and argue with it

This is where most DCFs go wrong. A small change in growth moves the final value a lot.

Example questions:

• "What revenue growth did analysts and management expect five years ago, and what actually happened?"
• "What growth rate is the current share price implying?"
• "Give me a bull case and a bear case for growth over the next five years, with the reasoning for each."

I would ask that second question every time. Instead of guessing the future, you work backwards: the market price already assumes some growth rate. Is that rate plausible?

What to verify yourself: compare the growth rate with the company's own history and with the size of its market. A company cannot grow 30% a year forever.

Does the company have a moat?

A moat is a lasting advantage that protects profits from competitors, such as a strong brand, network effects or low costs nobody can match. Companies with moats can keep margins high for longer, which justifies longer growth periods in your model.

Moneda AI scores a company's 7 Powers, a framework for the main sources of lasting advantage. It is a structured way to ask "why won't competitors take these profits away?"

Step 4: Choose a discount rate

The discount rate is the return you need to make holding this stock worthwhile, given its risk. Many models use the company's weighted average cost of capital (a blend of what its debt and its shares cost). Others just use a personal hurdle rate, such as 8% to 10%.

Example questions:

• "What discount rate would a typical DCF use for this company, and how is it worked out?"
• "Show me the value per share at discount rates of 7%, 9% and 11%."

What to verify yourself: check the risk-free rate the AI used (often the 10-year US Treasury yield) against today's figure. Stale rates are a common mistake.

Step 5: Set the terminal value carefully

The terminal value often makes up more than half of the total. It is usually worked out with a long-term growth rate that runs forever after your forecast.

Example questions:

• "What share of the total value comes from the terminal value?"
• "What terminal growth rate are you using, and why?"
• "What happens to the value if terminal growth is 1% instead of 3%?"

What to verify yourself: a terminal growth rate above long-run economic growth (roughly 2% to 3%) means you are assuming the company eventually becomes bigger than the economy. That rarely holds.

Step 6: Run a sensitivity table

Never trust one number. Ask for a table that shows value per share across a grid of growth and discount rates.

Example question:

• "Build a sensitivity table of value per share with revenue growth from 4% to 12% and discount rates from 7% to 11%."

If the stock looks cheap only in the most optimistic corner of the table, it is not cheap. If it looks fair across most of it, that tells you something.

Step 7: Cross-check with simpler measures

A DCF should agree, roughly, with simpler ways of looking at price.

Example question:

• "How do the P/E and free cash flow yield compare with this company's own five-year average and with its closest peers?"

If your DCF says the stock is worth twice its price while every simple measure says it is expensive, look again at your growth assumptions.

Moneda's Moneda Ratio is a descriptive 1 to 7 scorecard that summarises a company on several factors at once. It describes the company. It does not tell you to buy or sell.

What AI does well, and where it falls short

A note on accuracy. Every AI tool can make mistakes, and benchmark results do not guarantee correct answers on your specific question. Moneda publishes its own test results from October 2026: groundedness 97.9%, claim verifiability 92%, 80% on the public set of the Vals AI Finance Agent Benchmark v2 (run and graded by Moneda; Vals AI's top published score is 65.4% on its private set, which is a different set), and 95.6% on FLARE-CFA. These are Moneda's own tests, not independent audits, so check the numbers that matter to you.

Using Moneda for a DCF

What Moneda is built for: answers that show the source behind every figure, then invest in the same app. You can ask the agent to value a company in plain language, for example "Value Microsoft with a DCF and show me the bull and bear case." It researches stocks, ETFs (exchange-traded funds, baskets of assets that trade like one share), a few commodities through ETFs (gold, silver, oil) and crypto. It shows the source behind each figure, so you can check the inputs, and it pulls data straight from company filings: SEC filings and earnings-call transcripts are searchable, and earnings summaries give you the latest quarter in plain words.

From there you can:

• Use the stock screener to find companies worth valuing.
• Build an illustrative portfolio from a goal, track it and get told when something changes.
• Stress-test that portfolio against 2008, 2020 and 2022.
• Benchmark it against gold, an index or any fund (see how to compare two stocks with AI for side by side research).
• Invest in US-listed shares and ETFs (fractional) and crypto in the same app. You approve every trade with your passkey (face or fingerprint), and the agent never tells you to buy or sell.

If you prefer to work in Claude or ChatGPT, Moneda MCP connects those agents to your Moneda account with the permissions you set, so you choose what each agent can see and do. Payments are still approved in the app. The REST API and CLI are on Premium and Ultra.

Plans: Standard is free. Premium is from €9.99 a month and Ultra from €19.99 a month, with more AI usage and 1, 5 or 10 commission-free trades a month depending on the plan.

Honest limit: Moneda invests in US-listed shares and ETFs only, so you can't value and buy a European-listed stock there.

A quick checklist before you trust a valuation

• Historical figures match the filing.
• Earnings are backed by cash.
• Growth is in line with history and market size.
• The discount rate uses today's rates.
• Terminal growth is at or below long-run economic growth.
• The stock still looks reasonable across most of the sensitivity table.
• Simple measures (P/E, free cash flow yield) roughly agree.

FAQ

Can AI tell me what a stock is worth?

It can work out a value from your assumptions and the company's figures. It cannot know the future, so the result is only as good as the inputs. Treat it as a range.

Is a DCF useful for every company?

It works best for companies with steady, positive cash flow. For young companies that burn cash, or banks and insurers, other methods often fit better.

What is the most common DCF mistake?

Growth that is too high for too long, plus a terminal value that quietly dominates the result. Always check how much of the value comes from the terminal value.

How do I know the AI did not invent a number?

Ask for the source of each figure and check a few against the company's filing. If a tool cannot show sources, verify everything yourself.

Do I need to understand the maths?

You need to understand what each input means, not do the arithmetic. If you can explain why you chose your growth and discount rates, you understand your model.

Can I value ETFs with a DCF?

Not directly. An ETF is a basket of holdings, so you would look at the holdings, fees and what you are paying for the underlying earnings instead.

This guide is for education only and is not investment advice.