Nvidia's Doubling Pledge Rests on Memory Supply, Not Stagecraft
Published on 09/18/2026 at 18:41 | Editorial boerse-global.de
Jensen Huang used a Scottish AI summit stage to promise that Nvidia will ship twice as many chips in 2027 as it does this year. The stock's response — a 2.5% climb to $219.34 in New York, and a 2.6% advance to EUR 191.16 in German trading — was enthusiastic but hardly euphoric. That restraint is instructive, because the real test of Huang's headline number has nothing to do with the keynote and everything to do with whether the supply chain can feed it.
A Measured Benchmark Beneath the Big Number
Skeptics have learned to discount Huang's flair for round figures. This time, though, the announcement sits atop a stack of verifiable operational progress. Nvidia recently published preliminary MLPerf results showing its new Vera-Rubin NVL72 system delivers up to 3.7 times the inference throughput of the prior GB300 generation on the Qwen3-VL test — a measured benchmark rather than a marketing claim, and arguably the true foundation for the doubling target. Lift compute performance that sharply and customers keep queuing even as volumes scale.
The partner build-out reinforces the point. In Australia, Nvidia is expanding land, power and building capacity alongside a growing roster of cloud partners — a reminder that doubling shipments requires data centers, energy and floor space to exist first. Huang also appeared at Salesforce's Dreamforce alongside Marc Benioff, timed to the launch of Koa, Salesforce's first CRM product of its kind. Such appearances can look like pure PR, yet they demonstrate how deeply Nvidia's hardware has embedded itself in third-party software ecosystems. Each new alliance of this sort raises the odds that the doubling pledge is more than arithmetic.
The Gap Between Units and Revenue
The figure that actually decides the story is not the doubling itself but the distance between demand and deliverability. Nvidia has guided to roughly 70% revenue growth for the fiscal year ending January 2028 — 26 percentage points above the analyst consensus before that guidance, though the estimate has since settled near 63%. The reason unit volumes can double while revenue "only" grows 70% is straightforward: supply constraints. Unfilled demand is estimated at around $100 billion, and memory costs are expected to compress gross margin to 71–72% in the fourth quarter, down from 75% in the second.
Whether Nvidia untangles those capacity bottlenecks faster than the memory market tightens them determines if this remains a growth narrative or becomes a margin-erosion story.
Should investors sell immediately? Or is it worth buying Nvidia?
Hyperscaler Spending and the Energy Equation
The bull case leans on the sheer scale of infrastructure outlays. Huang himself projects $3–4 trillion in global AI infrastructure investment through 2030, while hyperscaler capex for 2026 is already estimated at roughly $725 billion — a 77% jump over 2025. A newly formed AI Energy Management Alliance, bringing together Google, Emerald AI, Anthropic, Analog Devices and other partners, aims to unlock up to 100 gigawatts of grid capacity in the United States by letting data centers flex their power draw around network congestion. That tackles one of the biggest structural obstacles to further growth: access to electricity.
Where the Bear Case Lives
The other side of the wager is just as tangible. HBM memory prices are climbing 30–50% per generation by market observers' reckoning, inflating the cost base of Nvidia's own customers and squeezing its margin. Then there is China: Huang has acknowledged that the country could hit a milestone in advanced lithography by 2030, which over time would erode Asian dependence on Nvidia technology. Nearer term, export controls on the H20 chip already wiped out $2.5 billion in revenue during the first quarter.
S&P Global Ratings adds a further caution — memory makers and server assemblers in Asia are considerably more exposed to an AI investment downturn than foundries such as TSMC, a signal that the supply chain as a whole is more fragile than the share prices of Samsung and SK Hynix suggest. Those two Korean names jumped 3.4% and 6.4% respectively following the announcement, underscoring how far Nvidia's remarks ripple across the sector.
Valuation, Volatility and the Next Checkpoint
Momentum has not yet stretched the valuation to breaking point. In German trading the stock sits 2.9% above its 50-day average of EUR 186.05, still 5.4% shy of the 52-week high of EUR 202.50 set in May. An RSI of 53.1 points to no overheating, and the shares trade about 12% above their 200-day average after gaining a third since their March low. Piper Sandler initiated coverage roughly a week ago with an Overweight rating and a $300 price target — a vote of confidence, though targets are opinions rather than guarantees. The consensus sits at an average of about $328 across 61 firms, with a range spanning $180 to $515.
The flip side is an annualized volatility of 39%, a measure of how twitchy the market gets around every Nvidia headline. Promising to double sales raises the bar considerably; missing that target would hit the stock harder than a routine profit warning would hit most other companies.
As long as hyperscalers hold their spending plans and Nvidia gradually clears its supply bottlenecks, the analyst consensus remains the dominant scenario. If the memory price spiral bites into margins more than expected, or if hyperscalers such as Amazon or Microsoft tap the brakes, the doubling story becomes unit rhetoric without an earnings driver. The next concrete test arrives with third-quarter guidance, for which Nvidia has already signaled roughly $108 billion in revenue — excluding China data-center sales. Hold that line, and the market will likely keep believing Huang's promise.
Ad
Nvidia Stock: New Analysis - 18 September
Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
