Nvidia’s flagship B300 AI servers are commanding eye-popping prices in China, according to industry sources, a sign that global demand and tightening controls are colliding to create acute hardware scarcity. Reuters reported this week that B300 systems are selling for about 7 million yuan (roughly $1 million) in China — roughly double the roughly $550,000 price for comparable equipment in the United States.
The B300 is more than a rack-mounted box: each unit packs eight of Nvidia’s high-performance GPUs and is tuned for heavy AI inference workloads used to run and monetize large language models and other generative AI services. The premium being paid in China reflects three forces converging, industry sources told Reuters — surging local AI demand, U.S. export curbs on cutting-edge chips and systems, and a stepped-up Chinese and international crackdown on chip smuggling that has squeezed gray-market supplies.
The demand surge is stark. Morgan Stanley estimates Chinese AI models accounted for 32% of worldwide token use in March 2026, up from about 5% a year earlier — a near-tripling that underscores how rapidly usage, and therefore infrastructure needs, are rising. That spike has pushed some Chinese firms toward short-term rentals rather than direct purchases: Reuters said short-term leases for Nvidia technology can run as high as 190,000 yuan a month.
At the same time, enforcement actions have tightened. Reuters reported U.S. officials recently pursued a key individual connected to Supermicro, a partner in server supply chains, in moves that targeted illegal distribution channels. Nvidia itself emphasized it offers no aftermarket support for systems sold through unofficial channels and voiced confidence in enforcement, saying: “Nvidia does not provide any service or support for such systems, and the enforcement mechanisms are rigorous and effective.”
Related Recipes
Anthropic taps xAI’s Colossus 1 data center to fuel Claude amid surging demand
Intel hits all-time highs on AI infrastructure bets as skeptics warn the rally may be stretched
ServiceNow Bets on AI With Action Fabric and Usage-Based Pricing to Reach $30B Revenue by 2030
Seoul’s Jogyesa Temple Ordains Robot Monk Gabi, Sparking Debate on AI in Buddhism
For investors and policymakers the price divergence highlights a double-edged reality for Nvidia. On one hand, the company’s chips are accruing pricing power as the backbone of AI infrastructure; on the other, rising geopolitical and regulatory frictions are raising the company’s China risk and creating potential barriers to growth in one of the world’s largest AI markets. Analysts have already been recalibrating expectations for the sector as surging enterprise AI spending strains supply chains across memory, foundry and server supply segments.
The spike in Chinese prices also signals market distortions that could reshape deployment strategies for AI providers. Some corporations appear reluctant to take direct ownership of sanctioned or hard-to-source hardware and are instead relying on rental markets despite higher operating costs. That dynamic could slow the pace of some long-term, at-scale deployments even as short-term demand for inference capacity climbs.
Trending Now
Valorant: Riot Temporarily Removes Neon from PC Queues to Patch Fast Lane Exploit Ahead of Patch 12.09
Kathleen Miller Named Nevada’s 2026 Mother of Achievement as Living Grace Homes Expands to Help Pregnant Teens in Las Vegas
DC Lottery May 8 Draws: Mega Millions Numbers 37-47-49-51-58 (Mega Ball 16) and Daily DC Game Results
Olivia Munn reveals on-set clash over female agency that briefly halted production, underscoring gender tensions in Hollywood
The situation underscores how export policy, enforcement and local demand are now as determinative of the AI rollout as chip design itself. As governments continue to tighten controls and pursue illicit supply channels, the premium for authorized, supported hardware in China looks set to remain high — a new, tangible constraint on the global AI buildout.
Continue Reading
