Nvidia CEO expects chip sales to double next year driven by AI

Nvidia just posted quarterly revenue of $96.22 billion. That number is up 106% from the same quarter a year ago, and it still might not be enough to satisfy the appetite for AI chips.

CEO Jensen Huang told investors during the company’s fiscal Q2 2027 earnings call on August 26 that customer demand forecasts point to chip sales doubling next year. The catch: Nvidia probably won’t be able to deliver on all of it, because the supply chain can’t keep up.

The numbers behind the confidence

Nvidia’s data center business generated $89 billion in revenue during the quarter. That’s a 117% year-over-year increase, making it the engine behind essentially all of the company’s growth.

For the current quarter, Nvidia guided revenue of approximately $108 billion, give or take 2%. For the full fiscal year 2028, ending January 2028, the company projected about 70% year-over-year growth.

Huang reiterated these figures at a September conference, expressing confidence that Nvidia could hit 70% annual growth while noting that underlying customer demand suggests something closer to a full doubling.

CFO Colette Kress spelled it out during the earnings call: supply constraints, particularly in high-bandwidth memory (HBM), are expected to persist through the end of fiscal 2028.

Where the demand is coming from

Emerging AI labs, sovereign AI initiatives run by national governments, and enterprise customers across industries are all showing up with purchase orders, broadening Nvidia’s demand base beyond hyperscalers like Amazon, Microsoft, and Google.

One concrete example: Nvidia announced a partnership with AWS to deploy an additional 2 million GPUs over the course of 2027 and 2028.

The supply problem nobody can fix overnight

Nvidia’s margins are expected to stabilize around 71-72%, but memory pricing pressures associated with HBM scarcity could chip away at those numbers over time.

The company depends on partners like SK Hynix and Samsung to manufacture the HBM modules its chips require. When demand outstrips memory production capacity, memory prices rise, squeezing margins, and Nvidia can’t ship as many complete systems as customers want.

Kress’s acknowledgment that these constraints will last through at least the end of fiscal 2028 means this isn’t a temporary hiccup. Growing 70% when your customers want you to grow 100% leaves revenue on the table.

What this means for the AI infrastructure race

Infrastructure investments in AI are projected to reach multi-trillion-dollar levels through the end of the decade, and Nvidia sits at the center of nearly all of it.