Nvidia now trades at less than 17 times expected earnings over the next 12 months, according to Bloomberg data cited in market coverage this week. That is roughly half the multiple the chipmaker commanded in 2025 and far below the 50-plus times investors paid in 2023. A cheaper stock sounds like a gift, but the Nvidia forward P/E is only one side of a story that also involves memory costs, slipping margins and customers who now build their own chips. The same tension runs through our look at the AI infrastructure race, where spending keeps rising while the winners keep shifting.
The framing comes from a headline running on ad-hoc-news.de, which argued that Nvidia’s cheap multiple is only half the story. The underlying numbers are what matter. Revenue is still doubling, yet the market is paying less for every dollar of that profit, and it is doing so for reasons that show up in the margin line and in the product roadmaps of Meta Platforms and Alphabet. This analysis walks through the valuation data, the margin pressure, the custom-silicon threat, and what could change the picture over the next few quarters.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What the Sub-17x Nvidia Forward P/E Actually Says
A forward price-to-earnings ratio divides the current share price by analysts’ expected earnings per share over the coming 12 months. A reading below 17 means investors are paying less than 17 dollars for each dollar of profit Wall Street expects Nvidia to earn. Bloomberg data puts the figure under that line. For a company that has been the face of the AI buildout, that is an unusually modest price tag.
There are two ways to read a low forward multiple. The optimistic reading says earnings estimates are rising faster than the stock, so the shares have become cheaper on paper without anyone selling in a panic. The skeptical reading says the market refuses to capitalize those earnings at a high multiple because it doubts they will last. Both can be partly true at once, and the second reading is the one that gives the headline its edge.
One caution before going further. Not every October screen agrees on the number. Separate analyst commentary published by 247wallst earlier this month cited a forward multiple closer to 25 times for Nvidia, against about 40 times for AMD. The two figures do not reconcile in the sources we reviewed, and different earnings estimates or screening dates could explain part of the gap. This article uses the Bloomberg-based sub-17x figure as the headline number and treats the 25x reading as a reminder that the exact multiple depends on method.
From 50x in 2023 to Under 17x: The Valuation Timeline
The compression has been steady rather than sudden. In 2023, as the first wave of generative AI demand hit data centers, Nvidia traded at more than 50 times earnings. By 2025 the multiple had roughly doubled the level it sits at now. As recently as May 2026 it was still below 25 times. Today it is under 17 times. Each step down came while revenue was climbing, which is the part that makes the trend unusual.
| Period | Nvidia multiple | Source and context |
|---|---|---|
| 2023 | More than 50x earnings | Market coverage of the early AI boom |
| 2025 | Roughly double today’s level | Bloomberg data, as cited in market coverage |
| May 2026 | Below 25x | Bloomberg data, as cited in market coverage |
| October 2026 (Bloomberg) | Below 17x forward | Bloomberg data, 12-month expected earnings |
| October 2026 (247wallst) | About 25x forward | Analyst commentary, with AMD cited near 40x |
Reading across that table, the direction is clear even if the exact level varies by data provider. Investors have been willing to pay less each year for the same stream of Nvidia profit. That is a statement about confidence in durability, not about current performance, which is why the margin and customer data below carry so much weight.
Why a Cheap Multiple Is Only Half the Story
A low multiple can signal opportunity or warning. In Nvidia’s case, the opportunity half is easy to document. The company reported fiscal second-quarter 2027 revenue of $96.2 billion, up 106% from a year earlier, in its release dated August 26, 2026. Data center revenue reached $89.0 billion, up 117%. Few companies of any size post growth like that.
The warning half sits in the details around those results. Gross margin stood at 75% in the second quarter, yet Bloomberg-compiled analyst estimates project it below 72% by the fourth quarter. Rising costs for memory chips are cited as a pressure on profitability. And two of Nvidia’s largest customers, Meta Platforms and Alphabet, have moved to design AI chips in-house. Put those three items together and you get a business that is growing fast but may keep a smaller slice of each sale.
Eli Horton, senior portfolio manager for thematic and durable growth equities at TCW, said the valuation drop reflects investor skepticism about how long Nvidia’s current earnings power will last. That sentiment explains the paradox. The market is not doubting that Nvidia earned a fortune last quarter. It is doubting whether the next several years will look the same.
Fiscal Q2 2027 Results: Revenue Still Doubling
To judge the skeptics fairly, start with the growth. Based on the figures Nvidia has published, revenue has climbed in every recent quarter. Fiscal 2026 as a whole brought in $215.9 billion, up 65%. The fourth quarter of that year delivered $68.1 billion. The first quarter of fiscal 2027 jumped to $81.6 billion, and the second quarter reached $96.2 billion.
| Quarter | Total revenue | Data center revenue | Year-over-year growth |
|---|---|---|---|
| Q4 fiscal 2026 (ended Jan. 25, 2026) | $68.1 billion | $62.3 billion | 73% total, 75% data center |
| Q1 fiscal 2027 | $81.6 billion | $75.2 billion | 85% total, 92% data center |
| Q2 fiscal 2027 (ended July 26, 2026) | $96.2 billion | $89.0 billion | 106% total, 117% data center |
| Full fiscal 2026 | $215.9 billion | Not broken out here | 65% total |
Sequential growth was 20% in the first quarter and 18% in the second. Data center revenue now accounts for the overwhelming majority of the total, roughly 92 cents of every dollar in the latest quarter by our arithmetic on the published figures. That concentration is both Nvidia’s strength and its exposure. If a handful of hyperscalers change how they buy, the whole income statement feels it.
The spending side of that equation keeps expanding too. Meta’s capital plans, covered in our report on Meta’s raised AI capex, show why Nvidia’s order book looks so full. The catch is that the same customers pouring money into data centers are also the ones funding alternatives.
Nvidia Gross Margin: From 75% Toward Under 72%
The Nvidia gross margin is the single number that best explains the cheaper multiple. It was 75% in the second quarter. Bloomberg-compiled analyst estimates project it below 72% in the fourth quarter. That is a decline of more than three percentage points in two quarters at a time when sales are rising by double digits sequentially.
A quick illustration shows why analysts care. On the $96.2 billion of second-quarter revenue, each point of gross margin equals roughly $0.96 billion of gross profit. Three points is therefore close to $2.9 billion per quarter on that revenue base. This is simple arithmetic, not a forecast, but it shows how small-looking margin shifts become large dollar amounts at Nvidia’s scale.
Separate October analyst commentary expects margins to bottom at roughly 71% to 72% in the fourth quarter, then recover to 72% to 73% in fiscal 2028 if Nvidia can raise prices and absorb higher memory expenses. Bulls lean on that recovery path. Bears note that a recovery depends on pricing power, and pricing power is exactly what custom silicon threatens.
How Memory Chip Costs Squeeze Profitability
Rising memory prices are identified in the reports as a direct pressure on Nvidia’s profitability. High-bandwidth memory is a large share of the bill of materials for modern AI accelerators, and suppliers know their parts are scarce. When memory vendors raise prices, the cost lands first on Nvidia’s cost of goods and only later, if at all, on customers through higher list prices.
The squeeze is not limited to data center parts. Consumers have felt the same shortage in storage, as we documented in our coverage of the NAND shortage and its effect on SSD prices, and in console memory costs, covered in our look at PS6 RAM pricing. The pattern is consistent. Memory is the tightest link in the AI supply chain, and whoever sits closest to the buyer ends up absorbing the shock first.
For Nvidia the question is whether it can pass costs through. Its dominant position historically gave it room to do so. That is why the October commentary ties any margin recovery to price increases taking effect. If customers resist, the 71% to 72% floor could prove softer than hoped.
Hyperscalers Build Their Own AI Chips
The second half of the story is the customer base. Meta Platforms and Alphabet are among the major Nvidia customers that have moved to design AI chips internally. David Russell, global head of market strategy at TradeStation, said the shift toward customers building their own silicon could gradually erode Nvidia’s pricing power and market share. The key word in his assessment is gradually. Nobody is arguing that Nvidia’s data center business collapses next quarter.
Google TPUs Reach Meta’s Data Centers
Tom’s Hardware reported on October 22 that Google’s tensor processing units now serve AI workloads in production at Meta, the first reported deployment of Google’s custom TPUs by an external customer in that kind of environment. Separate market coverage on October 12 said Meta signed a multi-billion-dollar deal to rent Google’s TPUs and has also shipped its own MTIA chip designs. In other words, one Nvidia customer is renting capacity from another Nvidia customer’s in-house chips.
Meta’s MTIA, OpenAI’s Jalapeno and Broadcom
A 247wallst report from October 7 said Google is receiving production shipments of TPU v8i, OpenAI’s first-generation Jalapeno accelerator is shipping, and Broadcom expects to deliver three generations of MTIA accelerators to Meta by the end of 2027. Broadcom, which co-designs many of these chips, is the quiet beneficiary. The picture that emerges is a market where several large buyers are moving part of their spending away from Nvidia’s merchant GPUs and toward chips tuned to their own models.
Competitive Comparison: Nvidia vs TPU vs MTIA vs AMD
The table below summarizes where each rival stands based on the reports cited in this article. It does not compare raw performance, since no head-to-head benchmark appears in the sources. It compares business position, which is what drives valuation. For readers who want the memory-bandwidth angle on the GPU side, our breakdown of Intel’s D-Series against AMD’s MI355X covers that ground.
| Platform | Owner | Status in October 2026 reporting | Who buys it |
|---|---|---|---|
| Vera Rubin GPU platform | Nvidia | Full-production ramp reported May 2026, shipping since August 2026 | Broad hyperscaler and enterprise base |
| TPU (v8i in production shipments) | Alphabet | Running in Meta data centers, sold or rented externally | Google Cloud customers, Meta |
| MTIA | Meta Platforms | Shipped designs, three generations expected from Broadcom by end of 2027 | Mainly internal use |
| Jalapeno (first generation) | OpenAI | Shipping, per 247wallst | Internal use |
| Instinct accelerators | AMD | Cited at about 40x forward earnings | Merchant market competitor |
Two points stand out. First, Nvidia is the only merchant vendor with a platform in full-scale shipping that every cloud can buy. Second, the alternatives are becoming real products with named production milestones, not lab projects. The competitive threat has moved from theoretical to measurable.
What Analysts Say About the Cheaper Multiple
No directly quotable expert statements were available from our verified-quotes check for this story, so the views below are paraphrased from the reporting. Russell at TradeStation framed the issue as a slow leak in pricing power and share. Horton at TCW framed it as doubt about how durable current earnings are. Both comments describe the same fear from different angles: the profits are real, but the moat may be narrowing.
Not every analyst agrees with the cautious read. Morgan Stanley’s Joseph Moore was cited in October coverage as keeping an Overweight rating on Nvidia and stressing its faster revenue growth relative to AMD. That view treats the lower multiple as a buying chance. Readers tracking the wider sector can find more on how AI stories feed into equity moves in our AI and machine learning coverage.
The split between bulls and bears is healthy for price discovery. It also means the multiple is likely to stay sensitive to each margin print and each customer announcement. A single data point, such as a new TPU deal or a fresh memory price hike, can move sentiment more than another strong revenue number would.
Vera Rubin: The Product Cycle That Could Reset the Debate
Nvidia’s newest data center platform in the search results is Vera Rubin. It reportedly entered a full-production ramp in May 2026 and began shipping in August 2026. A new platform matters for valuation because it gives Nvidia a chance to reprice. Each generation has historically carried higher per-unit pricing, and that is the lever bulls point to when they argue margins will recover to 72% to 73% in fiscal 2028.
The risk is timing. If Vera Rubin ramps while memory costs are still elevated, the early quarters could show margin compression before pricing catches up. That would fit the Bloomberg-compiled estimate of a fourth-quarter dip below 72%. Demand outside the largest cloud buyers also matters, including sovereign and telecom projects such as the cluster covered in our report on B300 GPUs spotted at China Telecom.
Market Impact Across AMD, Broadcom and Memory Suppliers
A lower Nvidia multiple does not exist in isolation. If the market assigns less value to Nvidia’s earnings, it may also reconsider the premium on rivals. AMD’s cited 40x forward multiple looks rich next to Nvidia’s, which is the comparison Morgan Stanley’s Moore appeared to be drawing. Broadcom gains narrative strength each time a hyperscaler confirms a custom accelerator program.
Memory suppliers sit on the other side of the margin trade. Every dollar Nvidia loses to higher HBM prices is a dollar a memory vendor gains, which is why storage and DRAM makers have been among the quieter winners of the AI cycle. Capital markets are watching the same dynamic at the AI lab level, as shown by OpenAI’s reported funding talks at a $1.25 trillion valuation. Those labs generate the demand that fills Nvidia’s order book.
The near-term calendar adds noise. Several megacap tech companies report this week, and their capital spending guidance will either reinforce or challenge the thesis. Our preview of the October 28 and 29 big tech earnings lays out what to watch.
Historical Context: When Pricing Power Fades in Hardware
Hardware leaders have faced this pattern before. A supplier gains a commanding position while customers lack alternatives, then the largest buyers invest in substitutes, and the supplier’s margins and multiple drift lower even as unit volumes grow. The cycle does not require the leader to lose. It only requires the buyers to gain leverage.
Nvidia’s own path fits the early part of that script. From more than 50 times earnings in 2023 to under 17 times now, the multiple has fallen by about two-thirds. Over that same period the company moved from fast growth to enormous scale, with fiscal 2026 revenue at $215.9 billion. Scale and multiple tend to move in opposite directions. What is different here is that the earnings base is still expanding at triple-digit rates, so the compression is not simply a maturity story.
Five Predictions for Nvidia’s Valuation
These are our own judgments based on the data above. They are not forecasts from any named analyst.
- Margins will define the next move. If fourth-quarter gross margin lands near the 71% to 72% estimate, expect the multiple to stay depressed. A print above 72% would likely lift it.
- Custom silicon announcements will keep coming. With TPU v8i shipping, Jalapeno in the field and MTIA generations planned through 2027, more hyperscaler chip news is close to certain.
- Memory suppliers will keep leverage through 2027. Tight HBM supply favors vendors, which means Nvidia’s cost pressure does not vanish soon.
- Vera Rubin pricing will be the swing factor. Successful price increases would support the recovery to 72% to 73% in fiscal 2028 that bulls expect.
- The gap between data providers will narrow. The difference between a sub-17x and a roughly 25x reading should converge as analysts update estimates after the next earnings report.
What Investors and Builders Should Watch Next
For investors, three indicators matter most. Track gross margin guidance, any disclosure of customer concentration, and confirmed external TPU deals. For engineers and infrastructure buyers, the practical takeaway is different. More viable accelerator options mean more bargaining room on price and more software portability work, since models tuned for one chip family do not always move cleanly to another.
The takeaway: a sub-17x multiple is a verdict on durability, not on demand. Nvidia still posts record revenue, and the market still hesitates to pay up for it. Until margins stabilize and the customer-silicon threat becomes clearer in the numbers, the stock is likely to keep trading as if the best pricing is already behind it.
Frequently Asked Questions
What is Nvidia’s forward P/E right now?
Bloomberg data puts it below 17 times expected earnings over the next 12 months. Other October analyst commentary cites a figure nearer 25 times, so the exact reading depends on the data source and estimate date.
How does the current multiple compare with past years?
It is roughly half the 2025 level, below the sub-25x reading from May 2026, and well under the more than 50 times Nvidia traded at in 2023.
What was Nvidia’s gross margin, and where is it headed?
Gross margin was 75% in the second quarter. Bloomberg-compiled analyst estimates project it below 72% in the fourth quarter, with memory chip costs named as a key pressure.
Why are Meta and Alphabet a risk to Nvidia?
Both are major Nvidia customers that have moved to design AI chips internally. TradeStation’s David Russell said this could gradually erode Nvidia’s pricing power and market share.
Is Nvidia still growing?
Yes. Fiscal second-quarter 2027 revenue was $96.2 billion, up 106% from a year earlier, with data center revenue of $89.0 billion, up 117%, according to Nvidia’s August 26, 2026 results.
What is Vera Rubin?
It is Nvidia’s newest data center platform in the reporting reviewed here. It reportedly entered full-production ramp in May 2026 and began shipping in August 2026.
Is a low forward P/E a reason to buy the stock?
Not by itself. TCW’s Eli Horton said the drop reflects skepticism about how durable Nvidia’s earnings are. This article is news analysis, not investment advice.
