Bytedance secures access to Nvidia Blackwell cluster in Malaysia, circumventing US export ban on China

2026-03-13 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Fundamental Awareness, short

Summary

ByteDance, the parent company of TikTok, is reportedly planning to establish a significant AI computing cluster in Malaysia, circumventing US export controls that prevent direct access to advanced Nvidia chips in China. The plan involves approximately 500 Nvidia Blackwell systems, equipped with around 36,000 B200 chips, with hardware costs exceeding $2.5 billion. This initiative, in collaboration with Southeast Asian firm Aolani Cloud, aims to support ByteDance's global AI research and development, as well as its growing portfolio of over a dozen AI applications. Malaysia, previously a hub for chip smuggling, has implemented a permit requirement for US high-performance chips since July 2025, ensuring the chips remain in the country and are not physically transferred to China. ByteDance is also exploring additional chip capacities in Indonesia to further its global expansion and compete with companies like Google and OpenAI.

Key takeaway

For Directors of AI/ML evaluating global infrastructure strategies, ByteDance's approach highlights a viable model for accessing advanced AI chips despite export restrictions. Consider establishing or expanding compute operations in compliant Southeast Asian nations like Malaysia or Indonesia through trusted cloud partners. This strategy allows for the utilization of high-performance Nvidia Blackwell chips for global AI development, mitigating geopolitical risks while adhering to international regulations.

Key insights

Chinese tech firms are establishing offshore AI infrastructure to bypass US export restrictions on advanced chips.

Principles

• Export controls drive offshore compute strategies.
• Cloud partnerships enable access to restricted hardware.

Method

Companies like ByteDance partner with tier-1 cloud providers in compliant nations (e.g., Malaysia) to lease computing power from US-restricted chips, ensuring hardware remains local.

In practice

• Explore cloud partnerships for restricted hardware access.
• Establish R&D centers in compliant jurisdictions.

Topics

• NVIDIA Blackwell
• US Export Controls
• ByteDance AI Strategy
• Malaysia Data Centers
• AI Chip Geopolitics

Best for: Director of AI/ML, CTO, VP of Engineering/Data, Executive, Investor, Policy Maker

Related on AIssential

• China’s ByteDance Developing New AI Chips Like Those from Nvidia Partner Groq — ByteDance is developing AI chips similar to Groq's, signaling strategic hardware independence.
• The US says ASML’s top chip tool may be in China. ASML says it isn’t — U.S. export controls on advanced chip technology face scrutiny amid allegations of an ASML EUV machine reaching China.
• Meituan's LongCat-2.0 shows China can train massive AI models without Nvidia — China has demonstrated the ability to train trillion-parameter AI models using only domestic hardware, challenging US export controls.
• NVIDIA Secures Six Deals with AI Tech Giants in South Korea — NVIDIA is strategically deepening its supply chain and application partnerships in South Korea to sustain global AI growth.
• Jensen Huang just put Nvidia’s Blackwell and Vera Rubin sales projections into the $1 trillion stratosphere — Nvidia projects $1 trillion in Blackwell and Rubin chip orders by 2027, reflecting booming AI hardware demand.
• How the Global AI Race Has Changed — The global AI landscape is rapidly diversifying with new powers and technological fronts emerging beyond the traditional US-China rivalry.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.