Futures
Access hundreds of perpetual contracts
CFD
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
CFD
Stock CFD Derivatives
US Stocks
Access real US stocks and ETFs
HK Stocks
Trade quality Hong Kong-listed stocks
Korean Stocks
SK Hynix
Real Korean stocks and top assets
Stock Futures
High leverage, 24/7 trading
Tokenized Stocks
Backed by real stock assets
IPO Access
Unlock full access to global stock IPOs
GUSD
3.8%
Mint GUSD for Treasury RWA yields
Stocks Activities
Trade Popular Stocks and Unlock Generous Airdrops
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
IPO Access
Unlock full access to global stock IPOs
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
Promotions
AI
Gate AI
Your all-in-one conversational AI partner
Gate AI Bot
Use Gate AI directly in your social App
GateClaw
Gate Blue Lobster, ready to go
Gate for AI Agent
AI infrastructure, Gate MCP, Skills, and CLI
Gate Skills Hub
10K+ Skills
From office tasks to trading, the all-in-one skill hub makes AI even more useful.
China has approved the release of Nvidia H200 chips for Alibaba, ByteDance, and DeepSeek after Trump loosened the restrictions for half a year.
Chinese officials have reportedly notified leading AI companies such as Alibaba, ByteDance, and DeepSeek that they will be allowed to purchase limited quantities of NVIDIA H200 chips, but must first declare the required quantity and purpose to get approval.
(Previous Summary: Huang Renxun declines to testify before Congress; US lawmakers mock: "Free time for noodles, but no time to explain NVIDIA's China business") (Background Supplement: DeepSeek reveals its own chip-making plan: one-year layout, aiming to ditch both NVIDIA and Huawei)
Table of Contents
Toggle
The Trump administration cleared the sale of lower-end H200 chips to China last December, but Chinese authorities had been slow to approve. Now the tide has finally turned.
According to a report from The Information, Chinese officials have informed leading AI companies including Alibaba, ByteDance, and DeepSeek that they may obtain purchasing permits for a certain number of NVIDIA H200 chips for training and running AI models.
Who can buy, and how much
Citing two sources, the report said companies must first declare the number of chips needed and their intended use before they can obtain approval. It is not an open-ended quota nor available to everyone.
In fact, the US Commerce Department approved about 10 Chinese companies (including Alibaba, Tencent, ByteDance...) to purchase H200 chips late last year, with a cap of around 75,000 chips per company. However, the real bottleneck has always been on China's side: the procurement volumes applied for by Chinese companies early this year far exceeded that ceiling, and Beijing is now inclined to approve a total volume likely less than half of the applications.
Computing power shortage forces Beijing to soften stance
Beijing has long been wary of US-made AI chips entering the country. On one hand, it fears that a flood of US-designed processors could slow down the long-term development of China's homegrown chip industry. On the other, it worries that opening the door to foreign chips could create security vulnerabilities. These two reasons combined kept H200 chips stuck at customs for the better part of a year (even though we know China already holds a large number of smuggled NVIDIA chips).
Yet surging demand for computing power has eventually outweighed these concerns. Like their US counterparts, China's AI labs are struggling with insufficient computing power. Training the next generation of large models requires ever-increasing amounts of computation, and domestic chips cannot close the gap in capacity and performance anytime soon. The urgency of the chip shortage has thus overshadowed the long-term plan to support the domestic supply chain.
Blackwell and Rubin series chips still banned for export
The H200 belongs to NVIDIA's Hopper architecture. Simply put, it is NVIDIA's previous-generation chip series for training and running AI models. Before Blackwell was launched at the end of 2024, the H200 was the most powerful AI chip on the market. NVIDIA is now racing to develop its next-generation product line, Rubin, which is expected to launch in the second half of this year. Washington, citing national security concerns, continues to restrict exports of more advanced chips like Blackwell and Rubin to China. The H200 is essentially the ceiling of what China can buy and NVIDIA is willing to sell.
The cap on H200 purchases approved by China this time is likely less than 200k chips, a relatively small number in the context of AI infrastructure—a single data center can use more than 400k Blackwell chips. The huge gap suggests this approval is more of a symbolic easing than a real solution to the shortage. NVIDIA CFO Colette Kress admitted in May that H200 has yet to generate any revenue for the company in the Chinese market: "We are not certain whether these chips will actually be allowed to be imported."