Nvidia flags China, custom chip surge
What happened
- Nvidia CEO Jensen Huang told investors that Chinese demand remains crucial even as US export controls and China's domestic chip push complicate market access. - Huang said AI had crossed a “critical threshold” and tokens are now profitable, while TrendForce projects custom AI‑ASIC shipments to rise 44.6% in 2026 versus 16.1% for merchant GPUs. - The comments and market forecasts point to a market splitting between hyperscalers, specialised ASICs and geopolitically constrained GPU demand. (thestreet.com) (benzinga.com) (techtimes.com) (stratechery.com)
Why it matters
Nvidia is telling investors two things at once: China still matters, and the AI chip market is no longer a one-lane road. On May 20, the company reported first-quarter fiscal 2027 revenue of $81.6 billion, up 85% from a year earlier, with data center revenue of $75.2 billion. (investor.nvidia.com) The China piece is the more politically fraught one. Nvidia has already disclosed that the U.S. government told it on April 9, 2025 that exports of its H20 chip to China would require a license, and then said on April 14, 2025 that the requirement would remain in effect for the indefinite future. In a later SEC correspondence, Nvidia said no licenses had been provided to China customers and tied that change to a $4.5 billion charge for excess H20 inventory and purchase obligations. (sec.gov) That is why Jensen Huang’s emphasis on Chinese demand matters. He is not describing a marginal market. He is signaling that one of the world’s largest AI markets remains commercially important even as Washington’s controls limit what Nvidia can ship and Chinese companies push domestic alternatives. That reading is supported by Nvidia’s own export-control disclosures and by Huang’s public positioning around the market. (sec.gov) The second message is about AI economics. Huang told investors AI had crossed a “critical threshold” and that “tokens” are now profitable, according to reports on the earnings discussion. In plain terms, Nvidia is arguing that customers are no longer buying AI infrastructure only for experimentation; they are buying because inference output can be monetized. Nvidia’s reported growth gives that claim some financial backing, especially in data center. (investor.nvidia.com) But the market around Nvidia is also changing shape. TrendForce said cloud service providers’ in-house ASICs are expected to grow 44.6% in 2026, versus 16.1% for GPUs, based on growth in AI server shipments. In a separate January forecast, TrendForce said ASIC-based AI servers would account for nearly 28% of shipments in 2026, while total AI server shipments would grow more than 28% year over year. (trendforce.com) That does not mean Nvidia is suddenly losing control of the market. TrendForce said explicitly that faster ASIC growth does not mean Nvidia’s dominance “will collapse overnight,” and described the contest as moving beyond chip performance toward interconnects, switches, software and ecosystem strength. Nvidia has been making that same case in practice: not just selling GPUs, but selling networking, system design and a software stack that is harder to replace. (trendforce.com) So the split in AI chips now looks clearer. Hyperscalers such as large cloud providers are building more custom silicon for specific workloads. Nvidia remains the main merchant supplier for broad AI infrastructure. And China sits in a separate bucket, where demand may be strong but supply is constrained by U.S. policy. That segmentation is an inference from Nvidia’s filings and TrendForce’s shipment outlook, rather than a single company statement, but it fits the direction of both. (sec.gov) The next hard checkpoint will be Nvidia’s fiscal second-quarter results and any updated disclosure on China-related licensing, inventory charges or product mix. Nvidia’s investor site lists quarterly reports, transcripts and SEC filings as the company’s official channels for those updates. (investor.nvidia.com)
Key numbers
- Huang said AI had crossed a “critical threshold” and tokens are now profitable, while TrendForce projects custom AI‑ASIC shipments to rise 44.6% in 2026 versus 16.1% for merchant GPUs.
- On May 20, the company reported first-quarter fiscal 2027 revenue of $81.6 billion, up 85% from a year earlier, with data center revenue of $75.2 billion.
- government told it on April 9, 2025 that exports of its H20 chip to China would require a license, and then said on April 14, 2025 that the requirement would remain in effect for the indefinite future.
- In a later SEC correspondence, Nvidia said no licenses had been provided to China customers and tied that change to a $4.5 billion charge for excess H20 inventory and purchase obligations.
What happens next
- On May 20, the company reported first-quarter fiscal 2027 revenue of $81.6 billion, up 85% from a year earlier, with data center revenue of $75.2 billion.
- TrendForce said cloud service providers’ in-house ASICs are expected to grow 44.6% in 2026, versus 16.1% for GPUs, based on growth in AI server shipments.
- TrendForce said explicitly that faster ASIC growth does not mean Nvidia’s dominance “will collapse overnight,” and described the contest as moving beyond chip performance toward interconnects, switches, software and ecosystem strength.
Quick answers
What happened in Nvidia flags China, custom chip surge?
Nvidia CEO Jensen Huang told investors that Chinese demand remains crucial even as US export controls and China's domestic chip push complicate market access. Huang said AI had crossed a “critical threshold” and tokens are now profitable, while TrendForce projects custom AI‑ASIC shipments to rise 44.6% in 2026 versus 16.1% for merchant GPUs. The comments and market forecasts point to a market splitting between hyperscalers, specialised ASICs and geopolitically constrained GPU demand. (thestreet.com) (benzinga.com) (techtimes.com) (stratechery.com)
Why does Nvidia flags China, custom chip surge matter?
Nvidia is telling investors two things at once: China still matters, and the AI chip market is no longer a one-lane road. On May 20, the company reported first-quarter fiscal 2027 revenue of $81.6 billion, up 85% from a year earlier, with data center revenue of $75.2 billion. (investor.nvidia.com) The China piece is the more politically fraught one. Nvidia has already disclosed that the U.S. government told it on April 9, 2025 that exports of its H20 chip to China would require a license, and then said on April 14, 2025 that the requirement would remain in effect for the indefinite future. In a later SEC correspondence, Nvidia said no licenses had been provided to China customers and tied that change to a $4.5 billion charge for excess H20 inventory and purchase obligations. (sec.gov) That is why Jensen Huang’s emphasis on Chinese demand matters. He is not describing a marginal market. He is signaling that one of the world’s largest AI markets remains commercially important even as Washington’s controls limit what Nvidia can ship and Chinese companies push domestic alternatives. That reading is supported by Nvidia’s own export-control disclosures and by Huang’s public positioning around the market. (sec.gov) The second message is about AI economics. Huang told investors AI had crossed a “critical threshold” and that “tokens” are now profitable, according to reports on the earnings discussion. In plain terms, Nvidia is arguing that customers are no longer buying AI infrastructure only for experimentation; they are buying because inference output can be monetized. Nvidia’s reported growth gives that claim some financial backing, especially in data center. (investor.nvidia.com) But the market around Nvidia is also changing shape. TrendForce said cloud service providers’ in-house ASICs are expected to grow 44.6% in 2026, versus 16.1% for GPUs, based on growth in AI server shipments. In a separate January forecast, TrendForce said ASIC-based AI servers would account for nearly 28% of shipments in 2026, while total AI server shipments would grow more than 28% year over year. (trendforce.com) That does not mean Nvidia is suddenly losing control of the market. TrendForce said explicitly that faster ASIC growth does not mean Nvidia’s dominance “will collapse overnight,” and described the contest as moving beyond chip performance toward interconnects, switches, software and ecosystem strength. Nvidia has been making that same case in practice: not just selling GPUs, but selling networking, system design and a software stack that is harder to replace. (trendforce.com) So the split in AI chips now looks clearer. Hyperscalers such as large cloud providers are building more custom silicon for specific workloads. Nvidia remains the main merchant supplier for broad AI infrastructure. And China sits in a separate bucket, where demand may be strong but supply is constrained by U.S. policy. That segmentation is an inference from Nvidia’s filings and TrendForce’s shipment outlook, rather than a single company statement, but it fits the direction of both. (sec.gov) The next hard checkpoint will be Nvidia’s fiscal second-quarter results and any updated disclosure on China-related licensing, inventory charges or product mix. Nvidia’s investor site lists quarterly reports, transcripts and SEC filings as the company’s official channels for those updates. (investor.nvidia.com)
