Nvidia Trades at 15 Times Earnings and Is Still Expensive

Nvidia closed at $219.53 on 26 August 2026, the day it reported its second quarter. It is the most valuable company in the world, at about $5.3 trillion.

Take our forecast for the year ending January 2028. Core earnings of $13.99 a share. At $219.53 that is 15.7 times earnings. The S&P 500 trades higher than that. Broadcom trades higher than that. For a company growing revenue more than 50%, 15.7 times looks like a mistake.

Now put the same forecast through a discounted cash flow at a 12% cost of capital. The answer is $168 a share, which is 23% below the market price.

Same revenue. Same margins. Same company. Two answers, 30% apart.

That gap is worth understanding, because it is not really about Nvidia. It is about what a P/E ratio quietly leaves out, and the same blind spot applies to the income stocks most of our readers hold.

What Nvidia just reported

The quarter was enormous, and it beat.

• Revenue of $96.2 billion, up 106% on a year earlier.
• Data centre revenue of $89.0 billion, up 117%.
• Gross margin of 75.0%, up 2.6 points.
• Earnings of $2.46 a share, up 128%.
• Guidance for the next quarter of $108.0 billion, which would be growth of about 89%.

Then Nvidia did something it had never done in 33 years as a public company. It guided a full year ahead, telling investors to expect revenue growth of about 70% in the year ending January 2028. That implies roughly $690 billion to $700 billion of revenue, more than $100 billion above what analysts were modelling. The shares rose 4%.

For scale: Nvidia sold $60.9 billion of product in the year ending January 2024. Two years later it expects to sell more than ten times that.

Our earnings forecast to 2031

We model 55% growth in the year to January 2028, not the 70% management points to, and then we fade growth hard. Revenue, growth and core earnings a share:

• Year to January 2026, actual: $215.9 billion, up 65%, $4.55 a share. That is 48 times the current price.
• Year to January 2027: $403.8 billion, up 87%, $9.00 a share. 24.4 times.
• Year to January 2028: $626.0 billion, up 55%, $13.99 a share. 15.7 times.
• Year to January 2029: $731.0 billion, up 17%, $16.62 a share. 13.2 times.
• Year to January 2030: $803.0 billion, up 10%, $18.58 a share. 11.8 times.
• Year to January 2031: $851.0 billion, up 6%, $19.90 a share. 11.0 times.

Two checks on that list.

It reconciles with the market. Our 2027 estimate produces a forward P/E of 24.4 times. Data providers currently report Nvidia's forward P/E at about 24.5 times. So the disagreement that follows comes from assumptions, not from arithmetic.

Most of 2027 is already known. Two quarters are reported, the third is guided, and only the fourth is a real estimate. There is very little forecasting risk left in that $403.8 billion.

Read the last line again. On these numbers you are paying 11 times earnings, four years out, for the company at the centre of the largest infrastructure build in the history of computing. That is the bull case in one sentence.

The case for cheap

Nvidia trades below the median forward multiple of its own peer group while growing about twice as fast as the fastest company in it. Forward price to earnings, with revenue growth beside it:

• AMD: 76.9 times, growing 38%.
• Arista Networks: 43.9 times, growing 30%.
• Marvell: 28.0 times, growing 25%.
• Broadcom: 24.1 times, growing 48%.
• Taiwan Semiconductor: 19.2 times, growing 35%.
• Micron: 12.0 times, growing 60%.
• Peer median: 26.1 times, growing 33%.
• Nvidia on 2028 estimates: 15.7 times, growing 55%.

Then add the margin. Nvidia earns about 64 cents of operating profit on every dollar of revenue. Nothing else in that list is close. Add the balance sheet: $117 billion of net cash and investments after subtracting every dollar of debt.

On that evidence the stock is not expensive. It is the cheapest large AI name relative to its growth.

The case for expensive

Now the cash flow model. Three assumptions do all the damage, and none of them is aggressive.

The risk-free rate is 4.65%

The 10-year US Treasury yield sits near a 20-month high, pushed up partly by the debt that AI companies are issuing to fund this very build-out. With a beta of 1.65 that gives a cost of equity of 12.5%. We use 12%.

This is the input most models still get wrong. Plenty of published Nvidia valuations discount at 8% or 9%, which is a bull case hiding inside a spreadsheet. Lower the discount rate far enough and any growth company is cheap.

Growth consumes cash

This is the part a P/E cannot see. Nvidia's receivables reached $63.1 billion last quarter, and days sales outstanding jumped from 45 to 60 in three months, because the company extended payment terms on large multi-quarter agreements. Inventory rose to $31.6 billion ahead of its next product ramp. Working capital now runs at about 21% of revenue.

On our 2028 revenue, that consumes $46.7 billion of cash in a single year. It never touches the earnings line.

You can already see it in the reported figures. Operating cash flow fell from $50.3 billion in the first quarter to $24.1 billion in the second, while reported profit rose.

Earnings tell you what a company recorded. Cash flow tells you what it kept.

Everything after 2031 grows at 3%

Standard, and unavoidable in any long-term valuation. It is also where most of the answer lives.

Run those three assumptions and the enterprise value is $3.97 trillion. Add net cash and investments of $117 billion, divide by 24.3 billion shares, and you get $168.18 a share.

Now the uncomfortable detail. Of that value, 71% is terminal value, the part that comes after the five years we forecast. You are not buying the next five years. You are buying what happens after them.

How sensitive is that answer?

Very. Value a share at different discount rates, holding 3% long-term growth:

• 10.5%: $203
• 11.0%: $190
• 11.5%: $178
• 12.0%: $168
• 12.5%: $159
• 13.0%: $151

Nothing there reaches today's price. To justify $219.53 on cash flows you need a 10.5% cost of capital combined with 4% growth for ever, which produces $228. That combination is possible. It is not conservative.

So which is it?

Both, and the reconciliation is simple once you see it.

A multiple values one year of accounting profit. A cash flow model values the money that profit actually delivers to an owner, after the receivables that finance customers, after the inventory that pre-funds the next generation, and after charging a real cost of capital in a 4.65% world.

Nvidia's earnings are real. Its cash conversion is deteriorating while those earnings grow. Both facts sit in the same filing.

Three scenarios frame the range:

• Bull. Management's 70% growth lands and margins hold near 74%. Worth about $280.
• Base. Growth of 55% and a margin trough at 72%. Worth about $168.
• Bear. Growth stalls at 25% then declines, margin falls to 65%. Worth about $76.

Minus 65% to plus 28%. That is not a defensive holding. It is a leveraged position on how long one capital expenditure cycle lasts, and it deserves to be sized that way. Blending the methods gives us about $240, or 9% above the price. Close enough to fair value that the honest answer is neither buy nor sell.

Three facts that rarely make the headlines

Two customers were 36% of revenue last year. One was 22%, another 14%. The year before, the top three were 12%, 11% and 11%. Concentration is rising, and the biggest customers all fund rival chip programmes of their own.

Nvidia guarantees up to $108.5 billion of other companies' obligations, including $105 billion tied to a data centre campus in Ohio leased to OpenAI. It also holds about $94 billion of equity stakes, many in customers, close to 30% of its total assets. When a supplier finances its buyers, revenue quality becomes a fair question.

It paid $17 billion for Groq. In December 2025 Nvidia licensed that company's inference technology and hired staff, for $13 billion in cash plus $4 billion payable, booking $14.4 billion of it as goodwill attributed to the workforce. Eight months later the product ships. Nvidia lists the deal as a risk factor in its own annual report.

Why this matters if you own income stocks

Nvidia pays a dividend of $0.25 a quarter. Against a $219.53 share price that is a yield of about 0.5%, so nobody owns this for the income.

The method still transfers directly. The test that separated cheap from expensive here was not the multiple. It was whether the profit turns into cash, and a dividend is paid out of cash, not out of earnings. A company can report rising profits for years while its receivables and inventory swallow the money that was supposed to fund your payout. That is how a dividend that looked well covered on earnings gets cut anyway.

Two numbers do most of that work on any company you hold. Compare free cash flow with net income. Then watch days sales outstanding. If the first is falling behind and the second is climbing, the earnings are being funded rather than collected.

Where these numbers come from

Every Nvidia figure is from the company's own filings: the annual report for the year ended 25 January 2026, the quarterly report filed on 26 August 2026, and the second quarter results and CFO commentary. Share prices and peer multiples were taken from stockanalysis.com in August 2026.

One thing we owe you. The forecast is ours, and it is a forecast. Every figure from the year to January 2028 onward is an assumption we chose, and a different set of assumptions produces a different answer, which is exactly what the range above shows. The peer multiples come from free data sources that disagree with each other: three different providers reported Nvidia's forward P/E as 21, 24.5 and 45 on the same day, depending on which earnings estimate they used. Treat any single multiple you read as a question to check, not as a fact.

The bottom line

A P/E ratio is a sentence with the verb missing. It tells you what a company earned and says nothing about what it kept.

Nvidia is the clearest example available: 15.7 times earnings and 23% above a reasonable cash flow value, at the same time, on the same numbers. The multiple is not lying. It is answering a narrower question than the one you are asking.

For your own holdings, ask the second question too. Does the cash follow the profit? For a growth company that decides whether the valuation is real. For an income company it decides whether the dividend survives.

Once you know which companies keep their cash, the next question is what that income compounds into. Our free dividend reinvestment calculator projects the net income in 10, 20 or 40 years with no sign-up, the dividend income simulator guide explains what the projection assumes, and dollar-cost averaging covers building the position in the first place.

This article is education, not investment advice, and it is not personalised to your circumstances. We are not licensed investment advisers. Do your own work before you buy or sell anything.