Introduction

On May 15, US President Trump concluded his state visit to China, marking the end of a highly anticipated summit. The accompanying business delegation from the US, which included executives from major tech companies such as Elon Musk, Tim Cook, Qualcomm, and Micron, drew significant market attention. Nvidia’s CEO Jensen Huang also confirmed his attendance, highlighting the tech sector as a focal point of this meeting.

Reflecting on Trump’s first visit to China in 2017, the US-China relationship was then characterized by a honeymoon phase of pragmatic cooperation, resulting in 34 cooperative projects worth a staggering $253.5 billion. Beyond energy, manufacturing, and agriculture, the tech industry emerged as a vital pillar of bilateral cooperation, achieving a mutually beneficial balance through interlinked supply chains and complementary technological advantages.

Fast forward nine years, the global landscape and major power relations have shifted, with competition increasingly centering on technology. From high-end chip embargoes and semiconductor supply chain blockades to the race in artificial intelligence and the competition for computational power and talent, US-China tech relations have entered a new phase of comprehensive strategic rivalry. The struggle for discourse power over tech regulations and industrial ecosystems has become crucial in shaping the future global order.

Reports indicate that both sides reached key consensus on practical cooperation and risk management in the tech sector during this meeting. This not only serves as a “cooling button” for the ongoing US-China tech rivalry but also injects strong stability expectations into global supply chains, capital markets, and innovation ecosystems.

Differences in AI Development Paths

The development paths and focal points of artificial intelligence (AI) in the US and China present stark contrasts, reminiscent of the difference between internal and external martial arts training.

The US approach to AI resembles a quest for enlightenment, emphasizing the breakthrough of general large models and anticipating a moment of “sudden realization.” Many in the US believe that a pivotal “singularity” must exist in AI development, where once achieved, large AI models could reach or even surpass the expertise of top global specialists across various fields. Coupled with interdisciplinary integration, such superintelligent entities could evolve rapidly, posing a significant challenge to humanity. Consequently, Americans are concerned about China potentially achieving superintelligent AI first, while also hoping to outpace their rivals in reaching this “singularity.”

In contrast, China has historically adhered to a pragmatic spirit, following a “AI +” development path. This is a bottom-up approach, akin to mastering various weapons through hard work and practice. Numerous Chinese AI companies focus on developing specific AI applications for various vertical scenarios, striving for profitability from inception rather than relying on venture capital. For instance, AI is applied to enhance processes and tools in diverse settings such as roads, mines, construction sites, hospitals, hotels, and even battlefields, improving efficiency and controlling costs. China’s AI is not an omniscient “deity” but rather a reliable and continuously improving “master craftsman” in specific niches.

In 2025, the author visited a rising AI and robotics company, Megatech, which had already achieved considerable sales and quarterly profits. Its clients include major pharmaceutical, semiconductor, and renewable energy firms, leveraging AI and automation to significantly accelerate research and production processes. For example, processes that would take millions of years in nature can now be iteratively optimized in a smart lab, drastically improving efficiency and reducing costs.

One of the most impressive aspects was the café in the factory, where two agile robotic arms prepared coffee with skills comparable to world-class baristas, efficiently crafting various drinks based on customer preferences. This café serves as both a relaxation spot for employees and a hub for innovation. The new Luckin Coffee store opened in New York in June 2025 utilized these embodied intelligent technologies.

Of course, some Chinese companies are also venturing into general AI, with DeepSeek being a notable example. The core difference lies in the operational models: US AI giants often adopt closed-source and paid models, while Chinese firms actively embrace open-source ecosystems and continuously iterate on large model performance. This approach undermines the monopolistic ambitions of their US counterparts, encouraging global users to adopt Chinese free models, thus facilitating a competitive strategy of “surrounding the city from the countryside.”

AI and the Rise and Fall of Bubbles

Bubbles are a normal part of the emergence of new phenomena. According to Gartner’s technology maturity curve, important new technologies typically experience a hype phase followed by a cooling period before entering a stable development stage. Initially, a significant influx of capital and talent is drawn to the promising new industry, but early cash flow and profitability often fall short of supporting inflated valuations, leading to a bubble that eventually bursts.

The current excitement in the AI sector is not immune to bubbles, reminiscent of the internet bubble from 1999 to 2001. At that time, the internet bubble spread from California and Wall Street to Beijing’s Zhongguancun, where it was reported that $7 billion was chasing projects that could tell compelling stories to the capital market. As a graduate student, the author once planned to start a company selling effective click-through rates to internet companies, pitching the idea to various entrepreneurs and venture capitalists.

The internet represented a paradigm shift in technology and business models, bringing immense advantages and dividends to the US economy. The growth during those years was genuinely driven by technology, leading to a fiscal surplus under the Clinton administration. However, the exuberance of capital often exceeded rational limits. By 2001, the internet bubble burst, with the Nasdaq index dropping over 70% from its peak.

Today, the current tech bubble is primarily evident in the AI and cryptocurrency sectors, bolstered by significant political support. Tech right-wing forces, represented by figures like Peter Thiel, have overshadowed Wall Street and the Federal Reserve, becoming key financial backers and officials in the Trump administration. The technology capital of the US West Coast is leveraging presidential power to overshadow the financial capital of the East Coast, as seen in the passage of the “Stablecoin Innovation Act” in 2025.

The current bubble may be nearing the levels of the 1999 internet bubble. Apart from Nvidia, which profits from selling “shovels” to the AI gold rush, many well-known AI companies struggle to demonstrate genuine profitability. Some companies have found that using AI to write code is less efficient than traditional programmers. Recently, many US companies have opted to invest in cryptocurrencies rather than their core businesses, and Nvidia has created a financial loop supported by debt, valuations, and orders with its major clients—classic signs of a bubble.

The anticipated “singularity” moment in the US tech sector may never arrive, and the actual growth trajectory of AI could be a curve that approaches ideal goals but never quite reaches them. The dream of superintelligent AI may simply serve as a placebo for the US tech sector in the face of fierce competition from China. Should the market consensus one day recognize this reality, the current bubble could face a reckoning.

Not all bubbles will burst immediately, however. Given that the Federal Reserve’s monetary policy is set to enter a loosening cycle, the current AI and cryptocurrency bubble may remain robust for some time. The author’s prediction is that during China’s “15th Five-Year Plan” period, the AI and cryptocurrency boom will peak before entering a correction phase. If this assessment holds true, investors in AI and cryptocurrency should consider reducing their positions at high points and shift towards holding cash until the market stabilizes and opportunities arise at reasonable prices.

AI and Institutional Evolution

Marxism posits that production relations must adapt to productivity. The advancement of AI significantly propels productivity, yet it simultaneously strains existing production relations and distribution systems. Institutions across the US, China, and other nations face systemic restructuring pressures.

The progress of AI is accelerating the overall productivity towards the communist stage envisioned by classical Marxist theorists. The persistent scarcity and inflation in human economic history are being replaced by abundance and deflation, with labor increasingly supplanted by capital and technology. This bifurcation is leading humanity to split into two classes: those who control AI and those who are controlled by it. A nation’s institutional authenticity—whether it is genuine socialism or blatant capitalism—can be evaluated by the degree of differentiation and distance between these two groups.

The arrival of industrialization brought about a significant employment reset effect: many traditional jobs disappeared, yet industrial equipment required more trained and knowledgeable individuals to operate. For instance, the advent of automobiles replaced the need for carriage drivers, yet the proliferation of cars created more driving jobs than existed for carriage drivers, along with numerous repair and highway-related positions. However, AI differs from previous technological revolutions; it obliterates numerous jobs rather than merely transferring them. Many positions across the humanities and sciences are at risk of being replaced by AI and robots, and while AI may not create new jobs, those involved in manufacturing and maintaining robots will also be robots.

Returning to the US-China AI competition, US general AI can write, compose poetry, create art, film, and code, primarily replacing the jobs of humanities graduates. Meanwhile, Chinese AI, while also capable of creative tasks, tends to replace jobs for engineering graduates. Recently, a popular narrative in China suggests that humanities graduates struggle to find jobs, emphasizing the need for engineering and technical skills. The author’s understanding of the AI industry suggests this perception may be superficial. It is hypothesized that a global higher education bubble may soon burst, making it increasingly difficult to find jobs regardless of one’s field of study, as the value derived from knowledge accumulation will be countered by AI.

In the AI era, employment may not represent a fundamental right or obligation but rather a privilege, as it provides individuals with a sense of purpose and control over their lives and societal affairs. However, genuine job opportunities will become increasingly scarce. The majority of those unable to find real employment will contribute primarily as consumers. Many will possess skills primarily in providing emotional value to others, bestowing significance upon mundane tasks. Most people’s so-called innovation will consist of creatively consuming others’ products, finding new ways to engage with them.

The social order in the AI era may witness more conflict rather than harmony. As the labor of most becomes unnecessary while human nature remains unchanged, a scenario of “distribution by noise” may emerge: those who can instill fear and anxiety in rulers may receive more subsidies, benefits, and privileges. This trend may represent the evolving strategies of internal competition among nations.

Thus, in the AI era, competition between countries, especially major powers, may hinge on who possesses a more resilient and flexible institutional framework, capable of adapting to the potential and shocks brought by technology through innovative and rapid iterations. As the AI era dawns, a nearly eternal debate in world history re-emerges: is a bottom-up system of free expression and competition more rational, or is a top-down mechanism of effective control more suitable? Or, how can a balance be struck between the two?

Winning the US-China AI Competition

The US-China AI competition primarily revolves around computational power, algorithms, and data.

Currently, the focus of competition lies in computational power, which is underpinned by high-end chips and power supply. In recent years, the US has attempted to leverage its advantages in the chip supply chain to constrain China. However, due to ineffective sanctions, rather than crippling China, this has accelerated the development of a complete domestic supply chain in China. Since 2018, the US has launched a strategic offensive in the chip sector while China has had to engage in a strategic contraction. China’s rejection of Nvidia’s restricted version of chips indicates that both sides have entered a phase of strategic stalemate, with China’s domestic supply chain gaining strength.

In the future, leveraging its robust talent base and a new national system, China is expected to transition to a strategic counter-offensive in the chip sector within five years. However, the US’s development of computational power also faces bottlenecks, particularly in power supply. The growth of electricity supply in the US has been slow for years, and the approaches favored by the Democratic and Republican parties differ significantly. Trump’s stance is “Drill, baby, drill,” while the Democrats advocate for the development of clean energy sources like solar and wind power. Should the Democrats’ energy policies prevail, the US’s computational power may soon be constrained by China’s renewable energy supply chain.

The foundation of algorithms lies in talent, and the competition in algorithms is fundamentally a competition for talent. The weakness in computational power has prompted China to leverage its mathematical “genetic talent” to innovate in algorithms, exemplified by DeepSeek’s resurgence during the Spring Festival of 2025. Subsequent mergers and acquisitions illustrate that top AI developers can achieve astonishing valuations in a thriving capital market.

A potential challenge lies in the migration of Chinese mathematical and AI talent to the US, which could determine the outcome of the US-China AI competition. Due to significant income disparities, top talent from China may easily flow to the US, undermining China’s competitive edge. In the near term, the US may have fewer but higher-quality talents, while China may have a larger pool of somewhat lower-quality talent.

Ultimately, the decisive factors may lie in data and cash flow. Whichever AI ecosystem can integrate data from over 6 billion people outside the US and China will gain a significant advantage in cash flow, accelerating the iteration and evolution of its AI technology. Therefore, capturing the intermediate markets will be crucial in the long-term AI competition between the two nations.

We must strategize from this overarching perspective. For instance, in the early stages of competition, China should utilize fiscal subsidies, financial support, and diplomatic pressure to encourage as many countries and regions as possible to adopt Chinese AI solutions, including autonomous vehicles, home robots, smart home systems, educational aids, intelligent diagnostics, and military systems. This approach should adopt a land-grabbing mentality, aiming to capture intermediate markets at the lowest possible prices, even at a loss. We should continuously encourage Chinese AI model teams to promote open-source and free solutions globally, providing comprehensive, high-performance, and affordable options. This could be a highly competitive strategy for China in the US-China AI competition.

If China’s manufacturing sector has brought about a price shock in industrial goods over the past few decades, then in the coming decades, we aim to initiate a new price revolution in the AI sector.