China's MIIT issues computing power standards as AI costs drop 70%

China is standardizing its computing power market — a move that could widen the country's cost advantage over US AI rivals.

China's Ministry of Industry and Information Technology will issue a computing power standard system construction guide and promote market-based pricing standards, the agency said at a State Council press conference, as Beijing moves to commercialize its AI infrastructure.

"Optimizing computing power resource deployment and improving utilization efficiency are critical to China's AI competitiveness," Xie Cun, spokesperson and director of the information and communication development department at MIIT, said. The ministry will promote standards for computing power service capability evaluation and market-based pricing.

The policy follows a wave of Chinese AI models that have undercut US rivals on price. Moonshot AI's Kimi K3, released July 17, costs $15 per million output tokens — 70 percent less than Anthropic's Fable 5 at $50. DeepSeek's V4 model charges just $0.87 per million tokens. Chinese systems now occupy all five top spots on OpenRouter's weekly usage leaderboard, from Tencent and Xiaomi to DeepSeek and MiniMax.

Standardizing computing power pricing could unlock new revenue for Chinese cloud providers and data center operators while further compressing costs for AI developers. The move shows China's intent to build an alternative AI infrastructure independent of US-controlled supply chains, potentially reshaping the global cloud computing market.

Market-Based Pricing Reshapes Competition

The MIIT's push for market-based pricing standards addresses a critical bottleneck: computing power in China is abundant but unevenly distributed and priced. By creating standardized pricing mechanisms, the government aims to make compute resources more accessible to smaller AI startups, which have struggled to compete with well-funded hyperscalers like Alibaba and ByteDance.

The timing is strategic. Chinese AI labs have already shown they can produce frontier-level models at a fraction of US costs. Moonshot's K3 performed competitively with Anthropic's Fable 5 on independent benchmarks, with Arena.AI ranking it as the best model currently available. The K3 launch triggered a semiconductor selloff: TSMC fell 7 percent on July 18 despite reporting a 77 percent jump in quarterly operating profit, while SoftBank dropped 9 percent and Z.ai plunged almost 30 percent in Hong Kong trading.

Domestic Hardware and Open-Source Strategy

China's computing power standardization effort is tied to its push for domestic chip independence. Meituan trained its LongCat 2.0 model entirely on Chinese-made semiconductors — a feat that "would have been inconceivable in October 2022," according to Paul Triolo, a partner at DGA-Albright Stonebridge Group. Chinese companies can buy 10 local chips from Huawei for the price of one Nvidia processor, George Chen, a partner at the Asia Group, said.

President Xi Jinping affirmed China's commitment to open-source AI at the World Artificial Intelligence Conference in Shanghai on July 17. "We must seize this rare historic opportunity, encourage open source, openness, cooperation, and sharing," Xi said. Almost all Chinese AI companies release their models under open-source licenses, allowing third-party providers to host them for free — with costs limited to "GPUs and energy," said Ameya Karnitkar, co-founder of Larridin, an AI measurement platform.

Investment Impact

For investors, the MIIT's computing power standards represent a potential inflection point. Chinese cloud providers like Alibaba and Tencent, which already back Moonshot AI, stand to benefit from a more liquid computing power market. US hyperscalers face a different calculus: if Chinese computing power becomes standardized and cheaper at scale, the cost advantage that has driven US AI dominance could erode. Nvidia shares fell 1.2 percent on July 18, briefly forfeiting its position as the world's most valuable company to Apple, as investors reassessed the sustainability of US AI infrastructure spending.

This article is for informational purposes only and does not constitute investment advice.