Pick Your Sovereign Stack

Malaysia Bets on Huawei—What Are the Long-Term Consequences?

𝗣𝗶𝗰𝗸 𝘆𝗼𝘂𝗿 𝘀𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻 𝗔𝗜 𝘀𝘁𝗮𝗰𝗸.

This week Malaysia announced a “sovereign, full-stack AI ecosystem powered by Huawei GPUs” and plans to deploy a localized version of the DeepSeek opensource LLM.

That got me thinking: As countries consider their sovereign AI strategies, 𝙬𝙝𝙖𝙩 𝙖𝙧𝙚 𝙩𝙝𝙚 𝙞𝙢𝙥𝙡𝙞𝙘𝙖𝙩𝙞𝙤𝙣𝙨 𝙤𝙛 𝙩𝙝𝙚 𝙩𝙚𝙘𝙝 𝙨𝙩𝙖𝙘𝙠 𝙩𝙝𝙚𝙮 𝙘𝙝𝙤𝙤𝙨𝙚?

One is the durability of these choices. 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗹𝗼𝗰𝗸-𝗶𝗻, the idea that the more a society adopts a certain technology the more unlikely users are to switch, suggests that these early choices are likely to be sticky over time. If that’s right, 𝗲𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺 𝗰𝗵𝗼𝗶𝗰𝗲, 𝗮𝗻𝗱 𝘁𝗵𝗲 𝘀𝗽𝗼𝗻𝘀𝗼𝗿 𝗮𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝘁𝗵𝗮𝘁 𝗰𝗼𝗺𝗲𝘀 𝘄𝗶𝘁𝗵 𝗶𝘁, 𝗺𝗮𝘆 𝗵𝗮𝘃𝗲 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝘄𝗲 𝗰𝗮𝗻𝗻𝗼𝘁 𝘆𝗲𝘁 𝗶𝗺𝗮𝗴𝗶𝗻𝗲.

Another is what drives the choice? Leaving questions about geopolitical alignment aside, some of the key strengths of the Huawei stack include cost competitiveness and system-level scalability, while Nvidia currently boasts better per-chip performance, power efficiency and a stronger software system.

𝗜𝗳 𝘆𝗼𝘂 𝗮𝗿𝗲 𝗰𝗼𝘀𝘁 𝗰𝗼𝗻𝘀𝗰𝗶𝗼𝘂𝘀 𝗮𝗻𝗱 𝗵𝗮𝘃𝗲 𝗽𝗹𝗲𝗻𝘁𝗶𝗳𝘂𝗹 𝗮𝗰𝗰𝗲𝘀𝘀 𝘁𝗼 𝗲𝗻𝗲𝗿𝗴𝘆, 𝘁𝗵𝗲 𝗛𝘂𝗮𝘄𝗲𝗶 𝘀𝘁𝗮𝗰𝗸 𝗵𝗮𝘀 𝗮 𝗹𝗼𝘁 𝘁𝗼 𝗼𝗳𝗳𝗲𝗿. A quick look at the Malaysian power stack reveals steady growth with about about 80% of it’s power coming from coal and gas (43% from coal). It’s a good reminder that low-carbon power for data centers is not a hot topic every region.

Last, 𝗠𝗮𝗹𝗮𝘆𝘀𝗶𝗮 𝗶𝘀 𝗴𝗿𝗼𝘄𝗶𝗻𝗴 𝗮𝘀 𝗮 𝗱𝗮𝘁𝗮 𝗰𝗲𝗻𝘁𝗲𝗿 𝗵𝘂𝗯 𝘄𝗶𝘁𝗵 𝗺𝗮𝗷𝗼𝗿 𝗶𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁𝘀 𝗳𝗿𝗼𝗺 𝗔𝗪𝗦, 𝗚𝗼𝗼𝗴𝗹𝗲, 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝗮𝗻𝗱 𝗼𝘁𝗵𝗲𝗿𝘀 (the WSJ wrote on this last fall). Many of these facilities will host completing tech stacks (like Nvidia), so while the “sovereign AI” choice might be Huawei, there will be plenty competitors present at scale in the country as well.

𝗪𝗵𝗮𝘁 𝗱𝗼 𝘆𝗼𝘂 𝘀𝗲𝗲 𝗮𝘀 𝘁𝗵𝗲 𝗹𝗼𝗻𝗴-𝘁𝗲𝗿𝗺 𝗰𝗼𝗻𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲𝘀 𝗼𝗳 𝘀𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻 𝘁𝗲𝗰𝗵 𝘀𝘁𝗮𝗰𝗸 𝗰𝗵𝗼𝗶𝗰𝗲? 𝗜𝘀 𝗶𝘁 𝗮 𝗯𝗶𝗴 𝗱𝗲𝗮𝗹?