Nvidia chief executive Jensen Huang said the company expects to sell roughly twice as many chips next year as it did in 2026, a forecast that reinforced Wall Street's view that the AI hardware boom is still accelerating rather than cooling.
Speaking at a technology gathering in Scotland on Thursday, Huang said shipment and sales volumes are projected to increase twofold in 2027, according to reporting from Bloomberg and CNBC journalists at the event. Nvidia shares rose about 2% to $219 in Thursday trading, moving back within reach of the stock's 52-week high of $236, as investors digested the remark. For more context on this story, see our ongoing latest AI developments.
A forecast that matches the numbers
The doubling comment is consistent with what Nvidia's management told investors on the company's August 26 earnings call. At the time, Nvidia's chief financial officer said "customers' forecasts point to our growth doubling next year," and framed the company's fiscal 2028 outlook — which calls for roughly 70% revenue growth — as "a supply-constrained outlook." Huang added that "the unconstrained is significant," a signal that demand still exceeds what the company can physically build.
In other words, Thursday's comment was less a new prediction than a public restatement: the constraint on Nvidia's growth remains manufacturing capacity, not customers.
Vera Rubin becomes the growth engine
The driving force behind the doubling forecast is the Vera Rubin platform, which has entered full production alongside continued deployment of Blackwell-generation systems. Huang has described Vera Rubin as expected to be "the fastest product ramp in NVIDIA's history," and the economics explain why: the platform carries a per-gigawatt revenue opportunity of roughly $40 billion, up from about $25 billion for Blackwell, according to company figures cited in a 24/7 Wall St analysis of the remarks.
Vera Rubin has been making the rounds publicly all week. On Tuesday, NVL72 systems built on the platform debuted in MLPerf inference results, posting up to 3.7x faster performance than GB300 configurations on some workloads. On Wednesday, Nvidia confirmed that pharmaceutical giants Lilly, Roche and Bristol Myers Squibb are all building AI supercomputers on its hardware, extending the platform's reach into drug development.
Supply deals are locking in the demand
The demand backdrop is visible in Nvidia's own filings. The company's most recent 8-K shows supply obligations that swelled to $279 billion, largely tied to memory procurement for Vera Rubin — a measure of how much Nvidia has already committed to spend on the components needed to double its output. Management also disclosed guarantee obligations capped at $108.5 billion for AI cloud partners.
At the same time, financing platforms involving Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR are expected to mobilize more than $500 billion of third-party capital for AI infrastructure, according to the filing. Those arrangements matter because they convert hyperscaler demand into funded projects — the difference between a customer saying it wants 100,000 GPUs and one with the balance sheet to take delivery.
The 'invisible half': sovereign and enterprise AI
Huang has repeatedly pointed to sovereign and enterprise AI as what he calls the invisible half of Nvidia's business, and it is growing quickly. Sovereign AI revenue grew 35% sequentially in the most recent quarter and more than tripled year over year, while the broader sovereign and enterprise segment is expanding at roughly 100% per year, according to company disclosures.
That segment matters for the 2027 doubling because national AI buildouts — governments buying their own compute rather than renting it from US hyperscalers — are largely still in their early innings. Every new national program is incremental demand that did not exist in 2026.
A safety caveat from the same stage
Huang also used the Scotland appearance to address AI safety, drawing a line on product readiness: "If it's not ready, just hold it back," he said, according to Fox Business. The comment places Nvidia — usually cast as the pure supply-side beneficiary of the AI race — squarely inside an increasingly loud debate about the pace of frontier development, which has drawn warnings from rival CEOs, safety researchers and, this week, King Charles, who told technology executives gathered in Scotland that AI's dangers must be confronted head-on.
It is a notable evolution in tone. Two days earlier at Dreamforce, Huang called the speed-versus-safety framing a "false choice," arguing that the two goals are not in conflict. The new comment goes further, conceding that specific products could be too risky to ship — a standard few hardware executives have articulated publicly.
What analysts are watching
Consensus earnings estimates for Nvidia's fiscal 2028 have jumped from about $13 to $16 per share over the past 90 days, with 39 upward revisions and zero cuts, according to data cited by 24/7 Wall St. That kind of one-directional revision cycle is rare for a company of Nvidia's size, and it suggests the Street has already largely priced in a strong 2027.
The risk, as ever, is concentration. A doubling of chip sales depends on a small number of hyperscalers continuing to absorb systems at today's extraordinary prices, plus the memory supply chain delivering on schedule. If either falters, the fiscal 2028 outlook tightens quickly.
But for now, the message from the company's chief executive is unambiguous: demand is not the problem, and next year Nvidia intends to ship everything it can build — twice as much as it did this year.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
