Chips and geopolitics: the unexpected rise of Huawei in AI technology
Jun 3, 2024
In 2023, Nvidia held a 90% share of China’s AI chip market, with sales of $7 billion. Now, less than a year later, Nvidia is cutting prices to compete with Huawei in China and move its “Made for China” H20 AI chipset off the shelves. What went so wrong, so fast?
In two previous articles, I reviewed the development and deployment of photovoltaic (PV) technology and Pebble Bed Fission Reactors by China. The current commercial rivalry between Nvidia and Huawei mirrors these earlier competitions in many ways.
This battle is centred on the technical foundations of a revolutionary new technology: Generalised Pretrained Transformers (GPTs). It represents a struggle over the commanding heights of AI, where modified “graphics cards” have demonstrated capabilities for high-order, lightning-fast, and uncannily human-like thought when running large language models (LLMs).
Nvidia, a titan in the field of AI chips, has long been the provider of choice for cutting-edge AI applications. Its chips power everything from data centres to the latest game graphics, positioning Nvidia as a pivotal player in the AI revolution.
However, recent developments have seen Nvidia’s dominance challenged by Huawei, a company that has rapidly ascended in the tech world despite significant hurdles due to U.S. sanctions.
The core issue is the semiconductor bans the U.S. placed on Nvidia, initially aimed at curbing China’s access to advanced AI technologies. These restrictions, intended to slow China’s technological progress and maintain U.S. supremacy in critical technologies, have proven counterproductive.
Last year, Nvidia launched three new AI chips for the Chinese market, aiming to sustain its presence in China, where the company had earned 77% of its revenues over the previous year. The H20 chip was expected to be highly sought after, featuring nearly all of Nvidia’s best technologies. However, to comply with the Biden Administration’s directives to limit the performance of key semiconductor exports, Nvidia scaled back some of the export-qualified chip’s computing power.
In 2023, Nvidia held a 90% share of China’s AI chip market, with sales of $7 billion. Now, less than a year later, Nvidia is cutting prices to compete with Huawei in China and move its “Made for China” H20 AI chipset off the shelves.
What went so wrong, so fast? Seemingly out of the blue, Huawei introduced their own AI chip, the Ascend 910-B. Their new chip hit the Chinese market around the same time as the H20, and experts quickly benchmarked Huawei’s chip as better, both in performance and power consumption.
Worse for Nvidia, the 910-B not only fetched a premium—being priced 20% more than the Nvidia H20—it was highly preferred. The 910-B is currently selling at over 120,000 RMB per card compared to Nvidia’s 100,000 RMB.
Clearly, in retrospect, the restrictions intended to secure Nvidia’s share of the market have instead incentivised business decisions that have seen Nvidia’s share of the Chinese market drop from over sixty to below twenty percent, with the US rulings also impacting the company’s reputation in China and fanning speculation about the commercial risks of possible future restrictions on US technology exports.
The measures aimed to curb China’s progress and maintain U.S. dominance in a critical sector have unfolded in a quite unexpected way, demonstrating the complexity and interconnectedness of the global tech trade.
China’s rapid progress underscores its long-stated determination to achieve self-reliance in semiconductor technology and reduce its dependence on foreign suppliers. Huawei has not only persisted but managed to leapfrog technological barriers at a remarkable pace.
Initially projected to be at least a half a decade behind, Huawei’s swift catch-up in the AI chip realm forced Nvidia to reconsider its pricing strategy to retain competitiveness, not within five years, but within one.
This scenario reveals several broader truths about the current state of global technology competition. First, it highlights a resolute drive for technological self-reliance in China. Secondly, it demonstrates that China’s burgeoning capabilities pose a direct challenge to Western technological hegemony.
In response to the US challenges, Chinese policy focused on reinforcing domestic semiconductor innovation through substantial investments and comprehensive strategic plans. The “Made in China 2025” initiative, for example, long flagged Chinese ambitions. Additionally, the 14th Five-Year Plan included specific goals to increase R&D spending and foster innovation within the fabrication and chip design industry, further demonstrating a long-held Chinese determination to achieve technological self-sufficiency in chip design and fabrication.
In response to Western sanctions, the Chinese government implemented measures to bolster Huawei. These included financial assistance, policy incentives, and fostering a new non-proprietary Open Source Instruction set based on the RISC V protocol, further supporting supply chains and production ecosystems to mitigate the impact of U.S. sanctions.
In 2020, the PRC’s National Development and Reform Commission (NDRC) and Cyberspace Affairs Commission (CAC) highlighted the importance of open source communities for industrial digitalisation in an Implementation Plan for Digital Economy Development in the 14th Five-Year Plan (2021-2025). NDRC further encouraged the development of open source algorithms in its March 2021 AI Innovation Work Plan, with RISC V among the open source approaches gaining more traction.
By prioritising the development of homegrown technologies and reducing reliance on foreign suppliers, China has strategically positioned itself to continue to dominate the global stage and to further reshape the dynamics of international technology competition.
Kari McKern, who lives in Sydney, is a retired career public servant and librarian and IT specialist. She has maintained a life time interest in Asian affairs and had visited Asia often, and writes here in a private capacity.
Contributions to Pearls and Irritations
Your contribution will fund quality independent comment and analysis in Pearls and Irritations
Contribute
