China drafts $295 billion plan to build national AI data center grid running on 80% homemade silicon — projected 2028 timeline could run into limits of local chip production

Beijing's $295 Billion Bet on Domestic AI Infrastructure

China is drafting an ambitious plan to invest roughly 2 trillion yuan (approximately $295 billion) over the next five years to build a nationwide grid of artificial intelligence data centers. The sweeping proposal, reported by Tom's Hardware, mandates that 80% of the semiconductors powering these facilities must be domestically produced — a target that directly tests the limits of China's indigenous chip manufacturing capabilities.

The plan, spearheaded by Beijing and expected to involve major state-owned carriers including China Mobile and China Telecom, envisions a 2028 timeline for initial completion. However, industry analysts caution that the country's semiconductor supply chain may not be ready to deliver on such an aggressive schedule.

The 80% Domestic Chip Mandate

At the heart of the proposal is a requirement that 80% of the chips used in these new AI data centers come from domestic manufacturers. This is a significant escalation from current levels, where China relies on imported semiconductors — particularly high-performance AI accelerators and advanced memory — for the vast majority of its computing needs.

The mandate is part of a broader national strategy to achieve semiconductor self-sufficiency amid escalating export controls from the United States and its allies. Washington has progressively tightened restrictions on advanced chip sales to China, targeting both the hardware itself and the equipment needed to manufacture it.

SMIC Operating at Maximum Capacity

China's primary hope for advanced chip fabrication rests with Semiconductor Manufacturing International Corporation (SMIC), the country's largest foundry. SMIC's N+2 process — roughly equivalent to 7-nanometer class technology — is currently running at 93% utilization, indicating that the foundry is effectively at maximum capacity.

This leaves little room to absorb the massive additional demand that a $295 billion infrastructure buildout would create. Analysts quoted in the report note that China's semiconductor fabrication capabilities trail the global leading edge by an estimated 5 to 10 years. Even with substantial capital investment, closing that gap in time to meet the 2028 deadline appears extremely challenging.

HBM Memory and Huawei Ascend Bottlenecks

Beyond fabrication, the plan faces a critical bottleneck in high-bandwidth memory (HBM) — a key component for AI accelerators. The HBM market is currently dominated by South Korean manufacturers Samsung and SK Hynix, leaving China dependent on foreign supply for a crucial part of the AI computing stack.

Huawei's Ascend series processors represent China's most advanced domestically produced AI chips, with approximately 812,000 units shipped to date. However, production has been constrained by limited access to HBM memory modules. The company's Kunpeng and Ascend chip lines have faced persistent challenges scaling to meet the demands of cutting-edge AI workloads.

A revealing data point comes from DeepSeek, a prominent Chinese AI startup. The company reportedly attempted to train its large language models using Huawei's Ascend hardware but found the performance insufficient and ultimately reverted to Nvidia GPUs. This anecdote underscores the performance gap between domestic AI accelerators and the industry-standard offerings from Nvidia.

Projected Supply Gap

Industry projections suggest that even under optimistic scenarios, China's domestic semiconductor supply will only cover approximately 76% of demand by 2030 — falling short of the 80% target laid out in the plan. The gap underscores the structural challenges inherent in building a competitive semiconductor ecosystem from a position of technological disadvantage.

When factoring in the power grid modernization required to support a nationwide network of AI data centers, some analysts estimate the total cost could reach up to 5 trillion yuan (approximately $740 billion) — more than double the initial headline figure.

Geopolitical Context

The plan must be understood against the backdrop of intensifying US-China technology competition. Washington's CHIPS Act and subsequent export control measures have directly targeted China's access to advanced semiconductor manufacturing equipment, effectively freezing China's ability to acquire the lithography systems needed to produce leading-edge chips.

While Beijing's strategy of massive state-directed investment has yielded results in other sectors, semiconductors present a uniquely difficult challenge. The industry requires not just capital but specialized expertise, advanced equipment, and a mature ecosystem of suppliers and customers — all of which take years, if not decades, to develop.

What This Means for the Global AI Landscape

China's ability to execute on this plan has significant implications for the global AI industry. If successful, a domestically powered Chinese AI infrastructure could accelerate the country's AI development independently of Western technology. If it falters, the gap between China and the US in AI computing capacity could widen further.

For now, the semiconductor industry and global tech observers will be watching closely to see whether China can translate its massive financial commitment into tangible progress on chip manufacturing — or whether the limits of domestic production will prove too great a hurdle.