在高端情境下，这可能成为迄今最大规模的国家支持半导体激励计划，背景是各国为 AI 与国安加速在地化供应。知情人士称该方案将独立于既有资金安排，例如 500 亿美元的「大基金三期」。政策动因与美国三届政府以来的出口管制不确定性相连，并伴随「全国一盘棋」式动员、要求优先采用国产元件，以及劝阻使用为符合管制而定制、较弱的 H20；Nvidia 高层则称其在中国 AI 晶片市场的份额已降至 0。
产业端已出现明显的资本与能力变化：中芯国际作为华为主要合作伙伴在缺少先进设备下仍持续扩张；AI 加速器设计商摩尔线程自上海上市以来股价累计上涨逾 600%（且曾因交易风险提示出现 19% 回落的报导）。同时，整体制程技术仍被描述为落后台积电所掌控的最先进水准约 6 年，呈现「大额补贴与国产替代加速」对照「前沿工艺差距仍在」的趋势。
China is weighing chip-sector incentives worth up to about $70 billion (Dec 12, 2025, 6:02 PM GMT+8). Officials are discussing subsidies and financing support of 200 billion to 500 billion yuan (about $28 billion to $70 billion), with exact totals and target firms still unsettled. Even the low end starts to approach the capital Washington set aside for the Chips Act, signaling Beijing’s drive to cut reliance on foreign suppliers such as Nvidia and to keep backing domestic players like Huawei and Cambricon, even after US approval to sell more powerful H200 chips into China.
At the high end, the proposal would be the largest-ever state-backed semiconductor incentives program, as countries race to localize supply for AI and national security. People familiar say it would operate separately from existing plans such as the $50 billion Big Fund III. The push is tied to uncertainty from export clampdowns under three US administrations and includes a “whole-nation” mobilization, instructions to prioritize domestic components, and guidance to avoid the weaker, China-tailored H20; Nvidia executives say their share of China’s AI chip market has fallen to zero, while China still hasn’t publicly agreed to allow H200 imports despite the policy shift.
Market and industry signals show sharp swings: SMIC keeps expanding as Huawei’s key partner despite lacking the most advanced equipment; Moore Threads’ shares are up more than 600% since its Shanghai debut (with reports of a 19% drop after warning of trading risks). Yet the underlying gap remains: China’s technology is described as about six years behind the cutting-edge processes controlled by TSMC, creating a trend where massive funding and domestic substitution accelerate while frontier-node capability still lags.
