Friday, August 7, 2026
DarkSubscribe
AI Infrastructure · News & Analysis
HomeChips & HardwareReport
Chips & Hardware · Report

Nvidia secures H20 GPU export approval to China and announces a new GPU specifically engineered for the Chinese market.

Demonstrates Nvidia's strategy to navigate US export controls through customized chip design, opening substantial new revenue stream in China's AI infrastructure buildout.
Trade pressSlicast · July 15, 2025 · Global · Source: wccftech.com
importance 80

NVIDIA CEO Jensen Huang announced a significant reversal in US export restrictions following a meeting with President Trump at the White House. The company has filed an application with the US government to resume sales of H20 GPUs in China and reports that US officials have assured NVIDIA it will receive the requisite authorization promptly. The company hopes to "soon" commence deliveries of the H20 GPUs to China. Alongside this development, Huang announced a new, dedicated RTX PRO variant AI GPU for China that meets the Trump administration's requirements and is "ideal for digital twin AI for smart factories and logistics."

This reprieve comes after the Trump administration formally communicated to NVIDIA on April 9th that its China-specific H20 GPU would be subjected to an "indefinite" export licensing requirement. That initial restriction prompted NVIDIA to announce charges of up to $5.5 billion for its fiscal Q1 2026, which concluded on April 27th, incurred from writing off "inventory, purchase commitments, and related reserves" for the H20 GPU. NVIDIA also completely wrote off its China-origin TAM following the imposition of US licensing requirements, vowing to treat any residual sales in China as an undiscounted windfall rather than a solid figure factored into its guidance.

Despite the challenges posed by China restrictions, NVIDIA continues to demonstrate robust momentum in its core markets. According to a recent UBS assessment, NVIDIA currently has visibility into a whopping $1.5 trillion in data center revenue. This aligns with NVIDIA's own statement during its Q1'26 earnings call, where it declared that it had visibility into "tens of gigawatts" of AI infrastructure projects. Speaking to reporters in Washington DC, Huang emphasized the company's strategic positioning, stating that "general-purpose, open-source research and foundation models are the backbone of AI innovation" and that "we believe that every civil model should run best on the U.S. technology stack, encouraging nations worldwide to choose America."

Read the original
Nvidia secures H20 GPU export approval to… · Slicast