Nvidia CEO Says Chip Unit Sales Should Roughly Double in 2027; U.S. Stocks Rebound Sept. 17 After Fed-Driven Drop
Jensen Huang told reporters in Scotland that Nvidia expects to sell about twice as many AI chips next year; U.S. indexes and semiconductor shares rose the same day, recovering ground lost after the Federal Reserve's rate increase.
Two numbers, one company, and a market trying to decide which one matters
On September 17, 2026, Nvidia chief executive Jensen Huang told reporters in Scotland that his company expects to sell about twice as many AI chips next year as it's selling this year[1][6]. He said it standing outside Dumfries House in Cumnock, ahead of an AI summit hosted by King Charles III[6]. Nvidia stock rose more than 2% that day. A broad index of chipmakers gained about 3%. Arm Holdings jumped 8.6%[4][11][17].
That's one number. Here's the other: on August 26, at its actual earnings call, Nvidia's own guidance to investors was about 70% revenue growth for fiscal 2028[2]. Not a doubling. Both numbers are true. Both came from the same company, three weeks apart. The gap between them is where this story lives.
Layer on one more fact: the day before Huang's remark, the Federal Reserve raised interest rates for the first time in three years, and stocks fell hard[3]. September 17 was, in large part, a bounce-back day. Oil prices dropped. Treasury yields eased off their highest level since 2007[3]. So a reader has to hold three things at once: a striking CEO quote, a more modest official forecast, and a market that had its own reasons to rally that day.
What "double" actually means
Huang's math was about chips shipped, not dollars earned[6]. Those are different measurements, and the difference matters. If Nvidia sells twice as many units but at a lower average price, or spends more to make each one, revenue doesn't double even though shipments do.
Nvidia's formal guidance — the 70% revenue growth figure — is the number the company is legally on the hook for, the one built into SEC filings and analyst models[2][15]. Huang's Scotland comment was more casual, made to reporters, not written into guidance. Bulls argue that's actually the more honest number: Nvidia's official forecast reflects what it thinks it can physically build, not what customers actually want to buy[10]. On that reading, the smaller official number is a ceiling set by factories, not a forecast of weak demand.
Skeptics see it differently. To them, a striking round number like "double" delivered informally, days after a market selloff, reads as sentiment management — a reason for investors to feel better about a stock that had just dropped alongside the rest of the sector[13]. Nobody disputes Huang said it. What's disputed is how much weight it should carry next to the company's own signed guidance.
The case for taking Huang at his word
Nvidia's last earnings report gives the bull case real numbers to stand on. The company brought in $96 billion in quarterly revenue, with $89 billion of that from data-center chips — up 117% from a year earlier[2]. Huang has said cloud companies' GPUs are sold out entirely[2]. If that's true, growth isn't limited by how many customers want the chips. It's limited by how many Nvidia can physically produce.
Huang also pointed to breadth: AI investment showing up in nearly every country where Nvidia does business, not just a handful of American tech giants[1]. That matters for the bull case, because it's an argument against the idea that this is a narrow, circular trade between a few companies. Nvidia has locked in about $279 billion in supply and capacity commitments to back that growth up, much of it in memory chips[7].
There's a real incentive behind Nvidia talking up demand publicly, and it's not just about the stock price. Chip factories and memory production lines take years and billions of dollars to build. Nvidia needs suppliers like Samsung and SK hynix to commit to new capacity now, based on a promise of future orders[7][16]. A confident public forecast is partly a message to those suppliers, not just to shareholders.
The bet against the boom: an accounting argument, not a "AI is fake" one
Investor Michael Burry, known for shorting the 2008 housing bubble, isn't arguing that AI chips are useless. His argument is about how cloud companies count their costs. He says hyperscalers — the big cloud firms buying Nvidia chips — are spreading the cost of those chips over five to six years on their books, when the chips are really only useful for two or three[8][9].
Here's why that matters. Depreciation is the accounting method companies use to spread the cost of expensive equipment across the years they expect to use it, rather than counting the whole cost the year they buy it. Stretch that timeline out, and this year's reported costs shrink, and this year's reported profit grows — without a single dollar actually changing hands differently. Burry estimates this understates costs across the industry by roughly $176 billion between 2026 and 2028[8][9].
He also points to scale: the five biggest cloud buyers are on pace to spend about $805 billion on capital projects in 2026, up from $261 billion just two years ago, in 2024[8]. To skeptics, that kind of spending growth, combined with deals where Nvidia helps finance the very customers buying its chips, makes demand look more independent than it really is[8]. It's worth noting Burry has disclosed short positions betting against Nvidia and related stocks, so he has a direct financial stake in this argument being right[8]. That doesn't make him wrong. It does mean his interest should be weighed alongside his math.
A different country, a different story entirely
Read the Korean press that day, and Huang's quote isn't a stock story at all — it's a supply story. Seoul Economic Daily's headline said Nvidia's chief expects sales "volume" to double, and built the piece around what that means for Samsung and SK hynix, the memory makers who supply Nvidia[6]. SK hynix reportedly supplies about two-thirds of the high-bandwidth memory Nvidia needs for its newest chips[12].
Korean outlets had reason to focus there. On September 15, two days before Huang's remark, Korean chip stocks rose even as U.S. chip stocks fell roughly 6%[13]. That split suggests investors in Seoul were pricing in the order book — actual contracts for memory chips — rather than reacting to swings in American sentiment. The U.S. market's day-to-day mood barely shows up in that coverage at all[6][7].
Back in the U.S., the mechanics of the rally deserve their own note. The Dow rose 0.6%, the S&P 500 rose 1.1%, and the Nasdaq rose 1.7% on September 17[11]. The tech-heavy Nasdaq gaining nearly three times what the Dow did fits a story where chip stocks led the day. But outlets split on how much credit to give Huang versus the Fed-driven rebound already underway — Bloomberg's headline stuck close to what Huang actually said, while Yahoo Finance's "demand is through the roof" framing leaned harder into the CEO's line as the cause of the rally, without mentioning the prior day's selloff[1][5]. Nvidia's contracts with its suppliers are locked in regardless of which framing wins. Whether the demand behind them holds up over the next two years is the part nobody in this story has settled yet.
Summary
On Sept. 17, 2026, Nvidia chief executive Jensen Huang told reporters in Scotland that the company expects to sell about twice as many AI chips next year as it is selling this year[1][6]. He spoke before an AI summit at Dumfries House in Cumnock, hosted by King Charles III[6]. Nvidia shares rose more than 2% that day[4]. A widely watched gauge of chipmakers gained about 3%[11]. Arm Holdings rose 8.6%[17].
The broader U.S. market rose too. But Huang's comment was not the only thing moving it. The Federal Reserve had raised interest rates on Sept. 16 — its first increase in three years — and stocks fell sharply that day[3]. Sept. 17 was largely a rebound. Oil prices also dropped on news that Saudi crude was starting to flow again, and 10-year Treasury yields fell back from their highest level since 2007[3]. Reporting that day credited all of these, not Nvidia alone[3].
The main point of genuine dispute is not whether Huang said it. It is what the word "double" covers, and whether the buying behind it lasts. Huang was talking about chip units shipped[6]. Nvidia's formal guidance to investors, given on its Aug. 26 earnings call, was about 70% revenue growth for fiscal 2028 — a big number, but not a doubling in dollars[2]. Bulls say the gap exists because Nvidia cannot build fast enough, so the official figure is a supply ceiling rather than a demand estimate[10]. Skeptics, including investor Michael Burry, argue the demand is partly an accounting illusion: they say cloud companies stretch out how long they claim their Nvidia chips will last, which flatters their profits and keeps the orders coming[8][9]. Nvidia and its largest customers reject that reading and point to sold-out capacity[2].
The Event
On Sept. 17, 2026, Nvidia CEO Jensen Huang told reporters in Cumnock, Scotland, ahead of an AI summit at Dumfries House hosted by King Charles III, that Nvidia expects to sell roughly twice as many chips next year as this year[1][6]. He said AI investment was appearing in nearly every country where Nvidia operates[1]. Nvidia shares closed up more than 2%, and a semiconductor index rose about 3%[4][11]. U.S. indexes rose across the board the same session, one day after the Federal Reserve raised interest rates and stocks fell[3].
Undisputed Facts
- Huang made the remark on Sept. 17, 2026, to reporters in Scotland before an AI summit at Dumfries House in Cumnock[1][6].
- He said Nvidia expects to sell about twice as many chips next year as it sells this year[1][6].
- Nvidia shares rose more than 2% on Sept. 17, 2026[4].
- A gauge of chipmaking stocks rose about 3% that session, and Arm Holdings rose 8.6%[11][17].
- On Sept. 16, 2026, the Federal Reserve raised interest rates for the first time in three years, and U.S. stocks fell that day[3].
- On its Aug. 26, 2026 earnings call, Nvidia reported quarterly revenue of $96 billion and data-center revenue of $89 billion, up 117% from a year earlier, and guided to about $108 billion in the following quarter, plus or minus 2%[2].
- On that same call, management forecast roughly 70% revenue growth for fiscal 2028[2].
- SK hynix is reported to supply about two-thirds of Nvidia's HBM4 high-bandwidth memory, with Samsung also seeking orders[12].
The Pressure
Strip away the moralizing and blame. What structural realities persist regardless of which narrative wins?
- Book the capacity before you need it
- Chip fabs and memory lines take years and billions to build. Nvidia has to talk up multi-year demand in public to get Samsung, SK hynix and TSMC to commit lines now[7][16]. A CEO forecast is partly a signal to suppliers, not only to investors.
- Concentration risk
- About 92% of Nvidia's revenue comes from its data-center unit, sold to a small group of very large cloud buyers[2]. Anything that slows those few buyers hits nearly all of the company's income at once.
- The depreciation lever
- How long a buyer says its chips will last changes its reported profit without changing a single dollar of cash. That makes useful-life assumptions the single most contested number in the AI build-out[8][9].
- Macro noise drowns single-company news
- A Fed rate move, an oil-price swing and a bond-yield reversal all landed in the same 48 hours as Huang's remark[3]. Attributing a whole-market day to one quote is a narrative choice, not a measurement.
Material realityNvidia is selling every AI accelerator it can make and reported $89 billion in data-center revenue in a single quarter, up 117% from a year earlier[2]. The bottleneck is physical: advanced packaging, high-bandwidth memory and electrical power. SK hynix supplies roughly two-thirds of Nvidia's HBM4, and Nvidia has committed about $279 billion in supply and capacity, much of it memory[7][12]. Those commitments are real contracts regardless of what happens to the stock. Separately, U.S. index moves on Sept. 17 followed a Fed rate increase the day before, falling oil prices and 10-year yields retreating from a 2007 high[3]. Both things are true at once: chip demand is currently unmet, and the market's one-day direction had more than one cause.
Narrative as a weaponThree groups are shaping how this day is read. Nvidia wants you to believe AI compute is infrastructure — permanent, supply-limited, and worth booking years early; the 'double' framing serves that, and it is worth noting the doubling was in units while the company's formal revenue guidance was about 70% growth[1][2]. Retail-investor outlets want a clean cause for a green day, so the CEO quote gets promoted over the Fed rebound that also explains it[5][17]. Short-sellers and AI skeptics want you to believe the buying is financed and accounted into existence rather than demanded, and they have a disclosed financial stake in that view[8]. Korean outlets are running a fourth story entirely, about whose fabs fill the orders[6][7]. No side disputes what Huang said; they dispute what it measures and how long it lasts.
How Each Side Sees It
Each major actor’s view — how it frames things, its underlying incentive, and how it’s materially affected. Tap a side to read it.
Frames it asNvidia's case is that AI compute is a new category of infrastructure, not a product cycle. Huang argues the constraint is supply, not appetite: cloud GPUs are sold out and Blackwell sales are 'off the charts'[2]. On that view, a forecast of doubling unit shipments is simply a statement about factory output finally catching up. Nvidia also points to breadth — investment showing up in nearly every country it operates in, and across industries, not just in a handful of U.S. cloud firms[1]. Its next platform, Vera Rubin, is pitched as carrying far more revenue per gigawatt of data-center power than Blackwell, so each new site is worth more to Nvidia than the last[14].
WhyNvidia needs customers to commit capital years ahead of delivery. Public confidence from the CEO helps lock in orders, supplier capacity and financing before rivals or in-house chips take share[7][16].
Impact on themNvidia gets about 92% of its sales from its data-center unit[2]. That makes the company unusually exposed to a small set of very large buyers. If their spending slows, there is little else to fall back on.
Frames it asSkeptics do not mainly argue that AI is useless. Their argument is about accounting and circularity. They say cloud companies write off Nvidia chips over five to six years when the chips are really useful for two or three — which makes reported profits look better than they are, by roughly $176 billion across 2026 to 2028 on Burry's math[8][9]. Depreciation is simply how a company spreads the cost of expensive equipment across the years it expects to use it. Stretch that period, and this year's costs look smaller and this year's profit looks bigger. Skeptics also point to deal structures in which Nvidia helps finance the customers who buy its chips, which they say makes demand look more independent than it is[8]. They note the five biggest cloud buyers are on track to spend roughly $805 billion in 2026, up from $261 billion in 2024[8].
WhyBurry has disclosed short positions against Nvidia and related names, so he profits if the trade unwinds[8]. That is a real financial interest, and readers should weigh it — it does not by itself make the depreciation argument wrong or right.
Impact on themIf the skeptics are right, the losses land first on the cloud buyers and on investors, and only later on Nvidia's order book.
Frames it asThe Korean memory makers see this as confirmation of orders they have already booked, not as a surprise. Their case: AI accelerators are useless without high-bandwidth memory stacked alongside them, so any doubling of Nvidia units is a doubling of memory demand. SK hynix is reported to supply about two-thirds of Nvidia's HBM4[12]. Nvidia has more than doubled its supply and capacity commitments, to $279 billion, much of it memory[7]. Korean coverage frames the day's news as an output question — can the fabs fill the orders — rather than a stock-market question[6].
WhyBoth firms are spending heavily to expand capacity and want long, binding commitments before they commit to new lines[7].
Impact on themKorean chip shares had already risen on Sept. 15 even as U.S. chip stocks fell about 6% — a split that suggests Korean investors were pricing the order book rather than the U.S. sentiment swing[13].
Frames it asTraders and market strategists argue the single-day move should not be read as a verdict on Nvidia. The Fed raised rates on Sept. 16 for the first time in three years, and stocks dropped[3]. The next day brought a recovery in beaten-down technology names, helped by falling oil prices and by 10-year Treasury yields easing off their highest level since 2007[3]. On this reading, Huang's quote supplied a reason for a bounce that was already underway.
WhyFund managers and brokerages benefit from clean, single-cause market stories; those stories are easier to sell than 'several things happened at once.'
Impact on themThe Dow rose 0.6%, the S&P 500 1.1% and the Nasdaq Composite 1.7% on Sept. 17[11]. The spread matters: the tech-heavy Nasdaq gained nearly three times what the Dow did, which is what you would expect if chips were a driver — but it also means the Dow's move was the smallest of the three.
Like this article?
The Bias Ledger average rating 3.7
The same story, as framed by outlets across the spectrum, ordered least to most biased. The bias score (1 = straight, 10 = heavily spun) is an AI assessment of that framing — click an outlet to see its track record. The tell is the word choice or omission that reveals the angle.
| Outlet | Vantage | Bias | How they frame it | The tell |
|---|---|---|---|---|
| Bloomberg | U.S. center, financial | 2 | "Nvidia's Huang Expects to Sell Twice as Many Chips Next Year" | Keeps the unit framing ('sell twice as many chips') and attributes it to Huang rather than to the company. Straight, but does not put the number next to Nvidia's own 70% revenue guidance. |
| Seoul Economic Daily | South Korean business daily, conservative-leaning | 2 | "Nvidia CEO Says Chip Sales Volume Will Double Next Year" | The most precise headline of the set — it says 'volume,' which is what Huang actually claimed. Frames the story around Samsung and SK hynix output, with the U.S. index reaction largely absent. |
| CNBC | U.S. center, business | 3 | "Nvidia earnings takeaways: Huang forecasts 70% fiscal 2028 revenue growth, far above estimates" | 'Far above estimates' is an analyst-relative frame that reads as a win. The underlying figures are company-reported and specific. |
| TheStreet | U.S. center, retail-investor finance | 3 | "Stock Market Today (Sept. 17, 2026): Nasdaq, S&P 500 surge on post-Fed rate hike buying" | Credits the Fed rebound, not Nvidia, for the same session — a direct contrast with the chip-led framing elsewhere. 'Surge' is a strong verb for a 1.1% S&P move. |
| Korea JoongAng Daily | South Korean, conservative-leaning English edition | 4 | "Nvidia's $279 billion memory push boosts Samsung, SK hynix outlook" | National-champion framing: the story is what Korean firms gain. 'Boosts' states the effect as settled rather than as an expectation. |
| Yahoo Finance | U.S. center, aggregator newsroom | 5 | "Chip Stocks Soar as Nvidia CEO Huang Says Demand Is 'Through the Roof'" | 'Soar' plus a quoted superlative in the headline. The causal 'as' links the sector move to the quote without noting the prior day's selloff. |
| The Motley Fool | U.S. retail-investor advisory, subscription-driven and structurally bullish | 7 | "Nvidia's 70% Growth Forecast Is a Supply Ceiling. Demand Is Actually Stronger Than That." | Reframes conservative company guidance as understatement. 'Actually' asserts the writer knows the real demand better than the filing does. |
References
- Nvidia's Huang Expects to Sell Twice as Many Chips Next Year — Bloomberg · U.S. center, financial newswire owned by Michael Bloomberg
- Nvidia earnings takeaways: Huang forecasts 70% fiscal 2028 revenue growth, far above estimates — CNBC · U.S. center, business network owned by Comcast
- Stock Market Today (Sept. 17, 2026): Nasdaq, S&P 500 surge on post-Fed rate hike buying — TheStreet · U.S. center, retail-investor finance site
- Why Nvidia stock is up over 2% on Thursday — Invezz · UK-based retail trading and investing news site
- Chip Stocks Soar as Nvidia CEO Huang Says Demand Is 'Through the Roof' — Yahoo Finance · U.S. center, ad-driven finance aggregator
- Nvidia CEO Says Chip Sales Volume Will Double Next Year — Seoul Economic Daily · South Korean business daily, conservative-leaning
- Nvidia's $279 billion memory push boosts Samsung, SK hynix outlook — Korea JoongAng Daily · South Korean conservative-leaning English-language daily
- Michael Burry reveals his verdict on the ongoing AI bubble — TheStreet · U.S. center, retail-investor finance site
- Michael Burry sounds alarm on $176B depreciation gap among tech giants — TipRanks · Investment-analytics firm, subscription-driven
- Nvidia's 70% Growth Forecast Is a Supply Ceiling. Demand Is Actually Stronger Than That. — The Motley Fool · U.S. retail-investor advisory, subscription-driven and structurally bullish
- How major US stock indexes fared Thursday 9/17/2026 — Associated Press · U.S. nonprofit cooperative wire service; this copy was read on the Washington Post site
- SK hynix Reportedly to Supply About Two-Thirds of NVIDIA HBM4; Samsung Targets Early Delivery — TrendForce · Taiwanese semiconductor market-research firm, industry-funded
- Samsung and SK Hynix defy a 6% US chip rout: what is Korea seeing differently — Invezz · UK-based retail trading and investing news site
- Nvidia Expects Chip Sales to Double in 2027 — 24/7 Wall St. · U.S. retail-investor finance site, traffic-driven
- NVIDIA CORP Form 10-Q, quarter ended July 26, 2026 — U.S. Securities and Exchange Commission · U.S. federal regulator; primary filing
- Nvidia's Huang asked SK hynix to bring forward supply of HBM4 chips by 6 months, SK's chairman says — Reuters · UK-based international wire service owned by Thomson Reuters
- Chip Stocks Rally: Stock Market Recap September 17, 2026 — Trading Strategy Guides · U.S. trading-education site, course-and-subscription funded