Nvidia at 13x Forward Earnings: Cheap Multiple, Expensive Assumptions

Nvidia (NASDAQ:NVDA) trades at $186.64, and its forward price-to-earnings multiple of roughly 13 only looks cheap if Wall Street's earnings forecasts survive the next phase of the AI buildout. That is the core tension. The multiple is low. The assumptions behind it are not, and the fiscal third-quarter earnings report due later this month will test them directly against guidance of $108.0 billion in revenue and a 74.0% gross margin.

The numbers on the surface support the bulls. Fiscal second-quarter revenue reached $96.22 billion, up 105.8% year over year, and Data Center sales rose 117% to $89.02 billion, according to the company's reported results. Nvidia has beaten EPS and revenue estimates in six straight quarters. The consensus estimate for fiscal 2028 EPS stands at $15.9146, drawn from 52 analysts, and it has risen from $12.8148 just 90 days ago, with zero downward revisions in the past month. Management expects revenue to grow about 70% in fiscal 2028 and describes that outlook as supply-constrained, because customer demand is growing at roughly 100%.

Yet the stock fell 21.4% over the past month during a broad chip selloff that Barron's described as leaving chip stocks at fire-sale prices. Nvidia is up 0.31% so far this year while the S&P 500 has gained 7.83%, and over one year the stock is down 0.53% against the index's 9.69% gain. A beta of 2.247 means the stock tends to move more than twice as much as the broader market in either direction.

What does the new Vera Rubin platform change?

Product economics are improving on paper. Vera Rubin, Nvidia's next-generation platform, raises revenue per gigawatt to $40 billion from $25 billion for Blackwell, according to the company's disclosures. That metric matters because AI data centers are increasingly constrained by power availability rather than by chip count alone. More revenue per gigawatt means each new data center contract is worth more to Nvidia's top line.

The company also retains $99 billion on its buyback authorization, a balance sheet signal that management sees value in its own shares at current prices. Nvidia makes the accelerators, networking gear and CUDA software at the center of AI data centers, which gives it pricing power that competitors have not yet broken.

Why do the forecasts carry hidden risk?

A forward P/E rests on forecasts, and Nvidia's forecasts depend on AI infrastructure spending continuing at a historic pace across Microsoft, Alphabet, Amazon, Meta and Oracle. That is the fragility. Hyperscalers have been spending more than 100% of cloud revenue on capital expenditure in recent quarters, an arrangement that requires continued access to financing and continued belief in AI returns.

The arithmetic of a multiple reset is unforgiving. If expected earnings fell 30% with the share price unchanged, the forward multiple would rise to about 18. A 40% cut would push it to about 21. The low-end fiscal 2028 EPS estimate of $12.8484 already implies roughly 15 times earnings at the current price. The stock is not cheap if the estimates are wrong.

Nvidia's own balance sheet is absorbing more of the risk in the meantime:

• Supply commitments rose to $279 billion.
• Guarantee obligations reach $108.5 billion.
• Days sales outstanding stretched to 60 days from 45.
• The company has invested nearly $50 billion in frontier AI labs.

That last item has drawn criticism. CEO Jensen Huang acknowledged that some observers call the arrangement "circular financing" — Nvidia effectively funding the customers who buy its products. Gross margin is expected to bottom at 71% to 72% in the fiscal fourth quarter, and the current outlook includes no China data center compute revenue, which removes a potential source of upside.

Customer behavior also shows how quickly computing needs turn over. Microsoft's reported decision against extending an Oracle cloud arrangement built on older Nvidia chips demonstrates that yesterday's flagship hardware can lose its anchor tenants with a single procurement choice.

What will the third-quarter report settle?

The fiscal third-quarter earnings report, due later this month, is the near-term verdict on three questions: whether revenue exceeds the $108.0 billion guidance, whether gross margin tracks toward management's 74.0% guidance and its 71% to 72% floor, and whether management confirms the fiscal 2028 growth target of roughly 70%. Execution remains strong even as the growing role of financing raises valuation risk.

Investors watching the report should track three specific signals, according to the analysis: revenue above guidance, gross margin tracking toward management's floor, and days sales outstanding no longer rising. If all three appear and fiscal 2028 estimates keep rising, the case for owning the stock improves. If guidance falls, or more of the business depends on Nvidia's own balance sheet, the bear case improves.

How much upside do analysts see?

The average analyst target of $327.14 implies about 75% upside from the current $186.64, though targets are one data point and guarantee nothing. The ratings distribution among 62 analysts covering the stock:

• Strong Buy: 10
• Buy: 49
• Hold: 2
• Sell: 1

That is 59 of 62 analysts at Buy or better, a lopsided consensus that historically moves violently when it is wrong. Upside surprises can move the stock fast in both directions. A strong report could quickly lift a stock that has dropped 8.8% in a week; a weak one could extend a drawdown that has already reached 21.4% in a month.

Is Nvidia actually cheap?

At $186.64, Nvidia's valuation depends entirely on whether the estimates hold. The multiple only looks low if the earnings behind it prove durable. Evaluating the stock before the earnings report means relying on forecasts that can be cut far faster than the underlying business can rebuild them — analyst revisions travel at the speed of spreadsheets, while supply commitments and factory ramps move at the speed of physics.

The stakes extend beyond one ticker. Nvidia is the single largest beneficiary and the single largest exposure in the AI capital cycle, and its results function as a real-time read on whether hyperscaler spending remains rational. At 13 times forecasts that depend on borrowed AI capital, one more quarter of results will show whether the stock is truly cheap.