The New Cold War
The U.S. and China are involved in a cold and bitter technological arms race to shape the future.
In the 45 years following World War 2, the U.S. and the Soviet Union were engaged in a geopolitical chess match that shaped the latter half of the 20th century globally. This ideological battle for global influence came in various forms including a nuclear arms race, propaganda campaigns, espionage, sanctions and technological showcases like the Space Race. Our modern Cold War has a different set of actors, China and the U.S. Their “conflict” lies in the race to develop premier technologies that will shape our future.
It has been an adversarial last few years between the U.S. and China. The former is the incumbent world power hoping to maintain its economic and military lead over the rest of the globe, while upholding the ideological hegemony of capitalism that runs it. China, on the other hand, under Xi Jinping, is hoping to break this hegemony and emerge as the new world leader. The country kickstarted its growth in the 1980s under the presidency of Deng Xiaoping and after 40 years now finds itself as the world’s 2nd largest economy.
The Hard Tech Cold War
The U.S. and China’s technological relationship has been tumultuous for over a decade. Chinese authorities blocked Facebook, Twitter (now X) and Google in July 2009 following riots in Xinjiang, a special autonomous region in western China. The crackdown was aimed at curtailing communications among independence activists. Since then, the list of banned platforms and publications has grown, and now includes Google, X, Instagram, Snapchat, Yahoo, Slack, YouTube, The New York Times, and The Wall Street Journal, among others. The U.S., which prides itself as a bastion of freedom and free speech, has avoided doing the same to Chinese companies until recently, and with every passing month the list of competing sanctions, bans and restrictions from either side grows longer.
The Donald Trump administration kicked this technological Cold War into high gear with key laws and policies towards the end of his term. In 2019, Chinese tech companies were slapped with sanctions by the U.S. aimed at restricting their access to advanced technology. This came after Trump imposed a special 25 percent tax on Chinese-made autos a year prior, on top of the ordinary 2.5 percent tax on foreign-made cars. That has so far prevented electric vehicle giant BYD and its Chinese peers from trying to enter the US market. The Biden administration has not only upheld these laws, but added to them.
The U.S. has long sought to limit the development of China’s semiconductor industry by placing companies on the entity list, which is a U.S. government compilation of foreign individuals, companies, and organizations deemed a national security concern, subjecting them to export restrictions and licensing requirements for certain technologies and goods. These companies include telecom giant Huawei Technologies Co. and the country’s largest chip maker, Semiconductor Manufacturing International Corp. or SMIC.
Advanced chips are increasingly becoming a pillar of geopolitical power, underpinning both military systems and data-processing capabilities that drive modern economies. To give you an understanding of this, Apple’s latest smartphones run on Taiwan Semiconductor Company (TSMC)’s new 4nm chips. Apple is a trillion dollar company and the iPhone is its’ premier product. These are trillion dollar, economy-altering stakes. In this day an age, a technological deficit is effectively an economic one.
In 2022, the U.S. government imposed export restrictions on advanced semiconductors and chip-manufacturing equipment. The rules required U.S. chip makers to obtain a license from the Commerce Department to export certain chips used in advanced artificial-intelligence calculations and supercomputing. These rules were no doubt aimed at handicapping China’s ascent as a computing powerhouse.
These restrictions rocked the business of U.S. chip manufacturers such as Intel INTC 0.00%↑ and Nvidia NVDA 0.00%↑ . I have written about Nvidia’s unique position in the AI gold rush, and a large part of their business was China, which accounted for around 20-25% of yearly revenues. The company attempted to circumvent the restrictions by developing less powerful versions of their chips, namely the A800 and H800 to replace the A100 and H100, which could be exported to China. However, in 2023 a second round of more restrictive and expansive export control threatened to squeeze Nvidia’s entire China business, by not only banning the two alternatives, but also signalling further scrutiny of future similar alternatives, like the H20 model that Nvidia is currently developing. Nvidia is currently operating as if H20 exports can be banned at any moment.
Not to be outdone, this week, China has introduced guidelines to phase out U.S. microprocessors from Intel, and AMD AMD 0.00%↑, from government personal computers and servers. The procurement guidance also seeks to sideline Microsoft's MSFT 0.00%↑ Windows operating system and foreign-made database software in favour of domestic options. The country is trying to spur its domestic computing capabilities by limiting the reliance on U.S. chips with companies such as Huawei SCIM and Baidu expected to pick up the demand. This is not all bad, the embargo has helped boost revenues at China’s domestic chip equipment manufacturing firms. China’s top 10 equipment makers reported revenue rose 39% in the first half of 2023 as compared to a year ago, according to Shanghai-based CINNO Research.
In my opinion, these sanctions feel like a strategy miss from the U.S. With the initial restrictions, U.S. companies could drip-feed inferior chips to China and achieve two things. First, they would handicap China’s computing capabilities until Chinese companies developed domestic chips capable of outperforming their U.S. counterparts. Second, by controlling which chips can be exported to China, the U.S. gets free intelligence on the level of technical capability that Chinese companies have. If no chips go to China altogether, the U.S. can’t get a clear view of what level the Chinese are at.
The AI Wars
When California-based OpenAI launched ChatGPT in the fall of 2022, China was caught flat footed. It was confirmation that their AI models were at least years behind the cutting edge. The Chinese Communist Party (CCP) was quick to ban the app, amid concerns that generative AI would render many human functions obsolete, something that China cannot risk considering their youth unemployment numbers. The ban, however, also escalated the tech trade war between China and the U.S.
China’s largest companies Alibaba, an E-commerce Hub and Baidu, a search engine, have since released their own chat based large language models, namely Tongyi Qianwen and Ernie. These models, while phenomenal at producing generative Chinese text, still lag behind U.S. models such as OpenAI’s GPT 4 and Meta’s Llama 2. Industry experts have estimated that even the best Chinese models are still about 2 years behind the bleeding edge U.S. models.
This lead could be exacerbated even further by the U.S. government’s enforced trade restrictions of advanced computing technology. It takes thousands of processors to train large language models, and the H800 and A800 currently being stockpiled by Baidu and Alibaba will soon be old tech in this rapidly developing technological environment. Nvidia just announced the Blackwell series of GPUs, which are expected to develop the next generation of language models. These chips will not be going to China. Local manufacturers like Huawei will have to step up and help their domestic counterparts to overcome this deficit.
Despite this, it is increasingly looking like large language models may soon commoditize. With various competing options all with comparable performance, the value in the AI tech stack looks like it will shift from the platform layer to the application layer. In simple terms, companies may opt to use their preferred LLM among many, and build their own native applications on them, with majority of value being captured by the application. We are already seeing signs of this now, with enterprises choosing and churning their AI platform of choice at will. If this shift happens, the Chinese stand to benefit, as their underdevelopment may be rendered irrelevant. Their tech giants; Baidu, Tencent and Alibaba have significant application expertise and will look to capture value in that way. I have written about the commoditization in AI here.
For now, the technology Cold War rages on. Technological progress looks to shape a larger part of our future by each passing day, and these two world powers want to be at the centre of it.
The last Cold War led to an arms race for nuclear weapons, a technology with the power to wipe out humanity. One has to wonder whether this Cold War is leading us to develop something just as dangerous.