Nvidia Doubles Production Capacity for Late 2026 as AI Demand Goes Through the Roof

Nvidia Supercharges Its Supply Chain

Nvidia CEO Jensen Huang has capped a whirlwind two-week tour of Taiwan with a striking announcement: the company has already doubled its production capacity for the second half of 2026, and its suppliers should brace for demand in 2027 that could double again.

Speaking to partners and reporters after his GTC Taipei keynote on 1 June and a string of meetings across the island's semiconductor ecosystem, Huang said Nvidia now works with approximately 120 Taiwanese suppliers and that second-half 2026 supply volumes have been raised to twice prior levels. "The second half of this year will be the largest ever for Nvidia and Taiwan," he declared, adding that capacity is tight across the entire supply chain because demand is growing so fast.

Vera Rubin Enters Full Production

The expansion is driven by the production ramp of Vera Rubin, Nvidia's next-generation AI platform, which Huang called "the largest product launch, probably in the history of Taiwan." The five-rack platform is being manufactured by more than 350 factories across 30 countries, with 150 ecosystem partners in Taiwan alone. The supply chain built for Vera Rubin is twice as large as that of its predecessor, Grace Blackwell.

On 5 June, Huang confirmed that all three major memory makers — Samsung, SK Hynix and Micron — have been qualified and are in active production of HBM4 memory for Vera Rubin, with the first systems scheduled to ship in Q3 2026. Analysts estimate SK Hynix holds roughly 60-70% of the allocation, with Samsung taking around 25-30% and Micron the remainder. AWS, Google Cloud, Microsoft, OCI, CoreWeave and Lambda are among the first cloud providers lined up to deploy Vera Rubin instances.

The Money Behind the Ramp

The scale of investment is staggering. Nvidia's annual spending in Taiwan has grown from $10-15 billion five years ago to approximately $150 billion today, and at GTC San Jose in March, Huang projected at least $1 trillion in revenue from the Blackwell and Vera Rubin platforms through 2027 — more than double the company's previous $500 billion estimate.

The wider ecosystem is scrambling to keep pace. SK Group Chairman Chey Tae-won has pledged to double SK Hynix's wafer production capacity within five years, while TSMC continues expanding advanced packaging and reportedly plans to raise 3nm prices by up to 15%. Huang, who described Taiwan as "the epicentre of the AI revolution", departed this week for South Korea to prepare partners for what he calls an "incredibly busy" 2027.