Wrong Yard, Shaky Fence

Washington has framed its China containment strategy around the concept of a "small yard, high fence." Huawei and DeepSeek suggest the policy is an abject failure.

Technology containment was one of the key elements of the Biden Administration’s tough China strategy. Couched in terms of the “small yard, high fence,” a series of increasingly tighter export controls and sanctions took dead aim on limiting China’s ascendancy as a technology superpower. The US put enormous pressures on western allies to join this anti-China campaign. Alas, the policy is an abject failure.

At least that’s the lesson that can be gleaned from the extraordinary resilience of Huawei and the stunning emergence of DeepSeek. Huawei, long China’s national technology champion, was Washington’s early target in Trump 1.0, with tough follow-on actions by the Biden team. Squeezed initially by the blacklisting of the dreaded US entity list, reinforced by a series of additional sanctions and restraints on its supply chain, the Chinese company was reeling. With its smartphone business on the verge of near collapse, Huawei’s overall revenues contracted by nearly 30% in 2021 and its profits plunged by almost 70% in 2022. The demise of Huawei was widely thought to be at hand. Yet nothing could have been further from the truth. It’s Phoenix-like recovery of 2022-24 — led by indigenous semiconductor chips, new 5G smartphones running on proprietary (non-Android) operating systems, electric vehicles, and stunning penetration into non-US global markets — has vaulted Huawei back into the upper echelon of global technology powerhouses.

And now there is DeepSeek, an extraordinary testament to China’s stated goal of becoming a global leader in artificial intelligence. In July 2017, China’s State Council approved a “New Generation AI Development Plan” as a follow-on to earlier industrial policy initiatives such as “Made in China 2025” (approved in 2015) and the “Internet-Plus Action Plan” (2017). China threw down the gauntlet with the AI plan, which set a target of 2025 for AI to become “the main driving force for China’s industrial upgrading and economic transformation.” By 2030, China’s AI Plan aspired for global leadership in “AI theories, technologies, and applications,” with China seeking by then to attain a dominant status as “the world’s primary AI innovation center.”

Kai-Fu Lee: AI is Entering the Era of Implementation

When the AI plan was unveiled six and a half years ago, the reactions in the West ranged from skepticism to fear. With DeepSeek bursting on the scene this week, skepticism quickly gave way to shock. It wasn’t just the subsequent rout in global equity markets, which had become a levered play on the established Western AI players such as OpenAI, Meta, Google, Microsoft, Amazon, and, of course, Nvidia, it was the very idea that America could actually lose the battle of “AI Superpowers” as Kai-Fu Lee presciently warned in his 2018 book. At first blush, DeepSeek has all the capabilities of its western large-language model competitors but with a training-cost structure that is roughly 5% to 10% of that which underpins the onerous machine learning requirements of the big capital- and energy-intensive players in the West. Three key lessons come immediately to mind:

• The hyper-speed of Chinese development has long been driven by powerful competitive leapfrogging. To a large extent, this reflects the accelerating pace of globalization driven by the development and adaptation of new information technologies. But is also reflects the unique Chinese characteristics of a new globaization. Three decades of 10% economic growth were an outgrowth of rapid industrial and technological upgrading — from low-value-added shoes and toys in the early 1980s to increasingly sophisticated and higher-value-added consumer electronics, capital equipment, state-of-the- art infrastructure, high-speed rail, green technologies, and electric vehicles. In all these areas, China’s competitive prowess matched or surpassed historical development accomplishments of other nations in terms of scale, cost, and quality. Huawei and DeepSeek, with the latter emblematic of stunning success in AI, are but the latest in a long line of Chinese competitive leapfrogging examples.

• Containment is an oxymoron in open-architecture technology businesses. Try as it might to contain Chinese access to new technology hardware, the United States has failed to curtail the software-driven applications focus that is now driving AI. This is an important point stressed by Kai-Fu Lee who made the compelling argument that China was well positioned to benefit from a shift from AI theory to applications. Thanks to the “open science” of mathematical and computational models of deep-learning algorithms developed in the US, the UK, and Canada, but readily available on open-architecture Internet-based research platforms, Lee saw China, with its competitive advantage in Big Data, poised for leadership in what he dubbed the “Age of Implementation.” The emergence of DeepSeek hints that this new age could well be at hand.

• China’s successes on the technology front have not arisen out of a vacuum but, instead, reflect the strong political and ideological commitment of Chinese Communist Party leaders. Dating back to 2017, when the AI development plan was first ratified by the State Council, Xi Jinping was quick to endorse the harnessing of Big Data, stressing the “integrated development of the digital economy and the real economy.” A 2020 RAND study concluded that this was a “whole of government” approach to big data analytics and applications. More recently, Xi has broadened this effort to encompass an emphasis on “new quality productive forces” aimed at modernizing the industrial system around a “new national system” under the central purview of the CCP. The stunning successes of Huawei and DeepSeek are emblematic of the importance that China’s political and ideological commitment play its progression as an emerging techno- and AI-superpower.

Having said all that, some big political questions remain on China’s ultimate AI potential. That is true from the standpoint of domestic Chinese politics. China’s 2021 regulatory crackdown underscored internal “red lines” both for Internet platform companies and social “bad habits” pertaining to gaming, video streaming, fan culture, and private tutoring. I have argued that this reflected the CCP’s growing discomfort with animal spirits, long a key force underpinning new technologies and innovation. Will AI-enabled innovation challenge these same red lines that seemed so threatening just four years ago? The same can be said for the CCP’s increased emphasis on discourse power and censorship. There are already signs that China’s large-language AI models are confronting censorship restraints, consistent with a warning I issued a couple of years ago. AI ultimately poses a challenge to all political systems. China is not exempt from that potential clash.

China’s AI ascendence also faces geopolitical pressures. DeepSeek’s success has been accompanied by the predictable allegations of intellectual property theft that always seem to occur on the heels of Chinese innovations. I have written about this for years, featured in the “China Gripe” chapter of my 2014 book, Unbalanced, which was a precursor to the false narratives I stressed in my latest book, Accidental Conflict. Currently, DeepSeek is being charged with “distillation” — the training of smaller models by drawing on output generated by larger models. This cost-saving practice is fairly common in AI academic circles, so common that it even has a name — “distillation.” I am hardly an expert in the science of AI, but I am struck by the irony of the aggrieved victim, a company with the name of OpenAI insisting that its proprietary technology — a rather curious concept, in and of itself, in the open-architecture world it celebrates — be barred to competitors. Maybe it should change its name to “ClosedAI”?

History is replete with tensions between politics and innovation. These tensions play out internally in the battle between labor and capital. And they shape competitive struggles between nations. China’s techno-ascendancy bears on both dimensions of this struggle. This has major implications for the US-China conflict. The United States is reeling as China ups the ante on AI. In the end, it had the wrong yard, and a very shaky fence.

Next Week: Two Bubbles?