Global AI server shipments are forecast to rise 36% next year, to 3.83 million units, as major cloud providers keep investing heavily in infrastructure. The projection points to a build-out still gaining pace, even as investors question whether the spending will pay off.
Watch Desk analysis
What happened
TrendForce research manager Frank Kung gave the forecasts at a forum in Taipei, according to the Taipei Times. The nine major cloud providers it tracks are projected to lift combined capital expenditure by 60% to US$1.52 trillion next year. TrendForce also expects the AI server market to grow 92% to US$919 billion, or 81% of the overall server market’s value.
Nvidia is forecast to remain the leading AI server supplier, with 100,000 rack shipments and a 62.6% market share. TrendForce expects Vera Rubin racks to account for about 70% of Nvidia’s rack shipments next year, while Rubin Ultra shipments are forecast to remain in the single digits as a share of the company’s total. AMD’s Helios systems are estimated at about 5,000 racks this year, with that figure potentially doubling next year; TrendForce cautions that deployment and delivery challenges could stop orders becoming shipments.
Why it matters
These figures put a scale on the infrastructure race: the forecast is not just for more AI servers, but for a market expected to dominate the value of server sales. That matters to chip and equipment suppliers, cloud providers and anyone trying to judge whether the current investment wave is translating into physical deployments.
The forecast is not a delivery tally or proof that every planned dollar will be spent. The Helios caveat is a useful reminder that orders, production and installed racks are three different milestones, however neatly they fit in a chart.
Our read
This is a substantial forecast, not a verdict on the eventual returns from AI infrastructure. The useful takeaway is that TrendForce expects server demand and cloud-provider spending to grow sharply next year, with Nvidia still out front and AMD’s rack-scale effort facing a ramp-up test.
What to watch
- Whether cloud providers follow through on the projected capital spending.
- How many forecast server orders become delivered and deployed systems.
- Whether Vera Rubin and Helios shipments meet their expected ramps.
Discussion spark: If AI server shipments and cloud spending keep climbing, what evidence would convince you that the infrastructure is earning its keep rather than getting ahead of demand?
Sources and evidence
- Global AI server shipments expected to jump 36% – Taipei Times (5 November 2026, 16:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.