【科创之声】技术封锁终将成自主创新“磨刀石”

Core Viewpoint - The recent U.S. policy allowing NVIDIA to sell H200 AI chips to "approved customers" in China, with a 25% revenue share to the government, raises questions about the value of this transaction for Chinese buyers, focusing on cost-effectiveness, reliability, and competition [2] Cost-Effectiveness - The policy shift indicates a move from a comprehensive blockade to a limited opening with economic ties, as the U.S. previously halted H200 chip exports, causing NVIDIA's market share in China to plummet from 95% to 0% [3] - The H200 chip, while being released, is considered a "second-tier" product compared to NVIDIA's Blackwell series, which remains banned for sale in China, and the 25% revenue share further diminishes its cost-effectiveness for Chinese users [3] Reliability Concerns - The U.S. government's inconsistent policies have heightened global supply chain risks, and the temporary lifting of the H200 chip ban does not restore trust among Chinese companies, who fear potential supply disruptions or scrutiny after purchase [4] - There are suspicions regarding the possibility of "backdoors" in the H200 chips, as NVIDIA's technology for "tracking" and "remote shutdown" has raised security concerns, further eroding market confidence [4] Competition - There are domestic alternatives to NVIDIA's chips, as Chinese AI chips are now included in official procurement lists, and the development of a self-sufficient technology ecosystem is progressing from emergency breakthroughs to systematic formation [5] - Support from over a trillion yuan in industrial funds and deepening collaboration between academia and industry is enhancing the competitiveness of domestic chips in various sectors [5] - The H200 chip may serve as a catalyst in the Chinese AI chip market, but it will not slow down the development of domestic chips, as Chinese companies have recognized the necessity of not relying on imports for core technologies [6] - The future market landscape is expected to evolve into a dual-track model of "diverse procurement + self-reliance," with short-term reliance on imported chips for specific scenarios while long-term efforts focus on advancing domestic technology and ecosystem [6]