Chinese Cloud Providers Undercut AWS, Offering Nvidia AI Servers for Just $6 Per Hour

Smaller Chinese cloud providers offer Nvidia A100 GPU servers for $6 per hour, compared to AWS rates of up to $32.

Introduction

The rapidly expanding demand for artificial intelligence (AI) infrastructure is driving companies worldwide to seek more affordable and scalable solutions. While leading U.S. cloud providers like Amazon Web Services (AWS) charge between $15 and $32 per hour for access to Nvidia’s powerful AI GPUs, smaller Chinese cloud service providers are offering comparable AI computing power for a fraction of the cost—just $6 per hour for 8-way Nvidia A100 servers. This pricing disparity has created a competitive edge for Chinese firms, even as U.S. export controls attempt to limit access to high-end AI hardware in China.

Broadcom Market Cap Soars from $230 Billion to Over $700 Billion, Becoming One of the World’s Most Valuable Companies | by techovedas | Sep, 2024 | Medium

Key Takeaways:

  1. Smaller Chinese cloud providers offer Nvidia A100 GPU servers for $6 per hour, compared to AWS rates of up to $32.
  2. Nvidia GPUs are smuggled into China, bypassing U.S. export restrictions and feeding the resale market.
  3. Large Chinese companies like Alibaba comply with U.S. regulations but charge higher rates for GPU servers.
  4. Nvidia’s modified A800 and H800 GPUs, designed to comply with export laws, have also been restricted by the U.S.
  5. Chinese firms continue to find ways to acquire Nvidia AI GPUs despite tightening U.S. export controls.

Global Foundry Industry Soars with 23% YoY Revenue Boost in Q2 2024 – techovedas

U.S. vs. China: The Price War Over AI GPU Rental

Chinese Cloud Providers Undercut AWS, Offering Nvidia AI Servers for Just $6 Per Hour

Despite U.S. efforts to restrict Nvidia’s A100 and H100 AI GPUs from reaching China, smaller Chinese companies have found ways around these restrictions.

They are acquiring these GPUs through smuggling and grey markets, enabling them to offer cloud services at a fraction of the cost charged by U.S. providers.

Smuggling Fuels China’s AI GPU Market

Since October 2022, the U.S. has imposed export restrictions on Nvidia’s A100 and H100 GPUs to prevent their sale to Chinese companies.

However, these GPUs are still widely available on Chinese e-commerce platforms like Taobao and Xiaohongshu.

Smuggling routes through Japan, Malaysia, and Indonesia, with Hong Kong as a key transit hub, keep these GPUs flowing into China.

Smaller Chinese cloud providers are benefiting from this influx. While AWS and other U.S.-based providers charge up to $32 per hour for Nvidia A100 GPU servers, Chinese vendors offer similar services for as little as $6 per hour.

This significant cost difference is due to the abundance of smuggled GPUs in China, which drives down rental prices.

SMIC to Build Four New 12-Inch Fabs Across China, Aiming for 28nm and Beyond by 2025 – techovedas

The Growing Role of Chinese Cloud Providers

Major Chinese companies like Alibaba and ByteDance adhere to U.S. export rules and charge higher rates.

Their prices are often double or quadruple those of smaller, less regulated providers who exploit the smuggled GPU market. These larger firms avoid using smuggled GPUs to stay compliant, which is reflected in their higher prices.

Why Nvidia AI GPUs Are Cheaper in China

Several factors contribute to the lower rental costs in China:

  • Smuggling and Resale Markets: GPUs are acquired through a black market, bypassing export restrictions.
  • Competitive Rates from Smaller Providers: These vendors operate outside major regulatory oversight.
  • Lower Operational Costs: Chinese providers have lower overhead costs than U.S. giants.
  • Abundant GPU Supply: High availability of GPUs in China, thanks to resale platforms.
  • High Demand for AI Power: Providers leverage low-cost rentals to attract more customers.

U.S. Restrictions and Nvidia’s Adaptation

To counteract U.S. restrictions, Nvidia introduced the A800 and H800 models with limited external connections.

These models were also restricted in October 2023. Despite these measures, Nvidia GPUs remain available in China through unofficial channels.

For instance, H100 GPUs are sold in Shenzhen’s Huaqiangbei market for $23,000 to $30,000, slightly above official prices.

The Future of AI GPU Access in China

Chinese companies are adapting to tighter U.S. controls by setting up entities in other countries to purchase GPUs, which are then shipped to China. Smugglers have even removed serial numbers to avoid detection.

This ongoing black market activity challenges U.S. efforts to control advanced technology distribution.

Conclusion

The availability of Nvidia AI GPUs in China remains robust despite U.S. restrictions. Smaller Chinese cloud providers continue to offer these GPUs at significantly lower prices than their U.S. counterparts.

This price disparity highlights the challenges in regulating the flow of cutting-edge technology into China. in China, despite U.S. restrictions, has allowed smaller cloud providers to offer highly competitive rates, undercutting major U.S. players like AWS.

By leveraging smuggled and resold Nvidia GPUs, these Chinese providers can offer 8-way Nvidia A100 servers at just $6 per hour, making AI computing more affordable and accessible in China.

As the global demand for AI infrastructure continues to rise, this pricing disparity highlights the ongoing challenges in regulating access to cutting-edge technology.

Kumar Priyadarshi
Kumar Priyadarshi

Kumar Priyadarshi is a prominent figure in the world of technology and semiconductors. With a deep passion for innovation and a keen understanding of the intricacies of the semiconductor industry, Kumar has established himself as a thought leader and expert in the field. He is the founder of Techovedas, India’s first semiconductor and AI tech media company, where he shares insights, analysis, and trends related to the semiconductor and AI industries.

Kumar Joined IISER Pune after qualifying IIT-JEE in 2012. In his 5th year, he travelled to Singapore for his master’s thesis which yielded a Research Paper in ACS Nano. Kumar Joined Global Foundries as a process Engineer in Singapore working at 40 nm Process node. He couldn’t find joy working in the fab and moved to India. Working as a scientist at IIT Bombay as Senior Scientist, Kumar Led the team which built India’s 1st Memory Chip with Semiconductor Lab (SCL)

Articles: 2237