China, by contrast, has long had a chequered track record on the innovation front. The Center for Strategic and International Studies (CSIS) once dubbed the country the “Fat Tech Dragon”, in reference to its low innovation “metabolism” in possessing an R&D engine that was large but inefficient. In a similar vein, Chinese innovation was often tagged with the derisive label of “good enough innovation”. Under the heavy hand of party-state direction, China’s top-down innovation model has long faced persistent problems of wasteful spending, allegations of intellectual property (IP) theft and a perceived lack of meritocracy, among other issues.
Yet, recent moves by the US — starting with sanctions on Huawei in 2019 and expanding into sweeping export controls and a broad push for tech competitiveness — show a growing sense of urgency in Washington. The tone of public discourse has shifted in tandem, with headlines telling a story of mounting anxiety: from Politico’s measured “Artificial Intelligence Cold War on the horizon” in 2020, to the Information Technology and Innovation Foundation’s (ITIF) urgent “Wake up, America” report in 2023. Most recently, Foreign Affairs last month even entertained what was once an unthinkable question: “What if China wins the AI race?”
On a more strategic level, the US’ willingness to impose export controls at the expense of hurting its own firms is telling. Nvidia CEO Jensen Huang, for instance, has been a vocal critic of these restrictions, the most recent of which have led to a multibillion-dollar write-down on the company’s H20 chips designed specifically for the Chinese market. With China’s AI sector projected to reach US$50 billion within the next two to three years, the loss of access to this lucrative, growing and thus-far captive market is a massive blow for the US tech sector. Perhaps even more critically, Huang is concerned that cutting China off its chips would only force the country to innovate and improve, and might even help it produce a future chip superior to Nvidia’s own. After all, necessity is the mother of invention. What is clear is that Washington no longer appears to see its lead as untouchable. The question, then, is: What do the facts show?
The US’ edge in chips
See also: Trump wins the trade war — and the advantage of future US relative productivity gains
In terms of raw compute power, the US is undeniably ahead of China. In 2023, OpenAI’s GPT-4 model reportedly required around 25,000 graphics processing units (GPUs) and nearly 100 days to train. While improvements have since been made in streamlining model architecture and algorithmic efficiency (with several key breakthroughs emerging from China), the sheer complexity of frontier AI models means that significant AI performance continues to rely on raw computing power. On this metric, CSIS projects that by year-end, the US will have three times as many AI accelerators as China (see Chart 1). This gap is even more pronounced when accounting for the superior performance of US chips.
Just as “artificial intelligence” is itself an umbrella term denoting a wide range of advanced technologies, the semiconductor industry, which serves as the infrastructure backbone to AI, is a similarly complex ecosystem split across numerous subsectors that span the globe (see Chart 2). Notable chokepoints along geographical lines exist in Taiwan (Taiwan Semiconductor Manufacturing Co), the Netherlands (ASML) and the US (Nvidia). Within this landscape, the US holds the strategic upper hand by controlling critical upstream components of this supply chain. For example, more than 90% of chips worldwide are designed using software from just three firms: Cadence Design, Synopsys and Mentor Graphics — all of which are US-based (Mentor Graphics, now the EDA [electronic design automation] software arm of Germany’s Siemens, is based and operates outside the US). Similarly, ASML relies heavily on its US-based subsidiary Cymer as the sole global supplier of the light sources essential for extreme ultraviolet (EUV) lithography — the process that underpins all advanced chip production.
See also: Trumpianomics will benefit the US greatly for years to come
Advanced lithography remains the greatest obstacle to China’s ambitions for semiconductor self-sufficiency. To its credit, the country has moved swiftly to address this shortfall. Following its blacklisting in 2019, Huawei — now the focal point of China’s semiconductor efforts — launched the Tashan Project in 2020, with the goal of building an indigenous chip production line.
In parallel, the government rolled out a series of large-scale funding initiatives. For example, government-affiliated Tsinghua Unigroup and other state-backed funds invested more than US$24 billion ($30.6 billion) into a single firm: Yangtze Memory Technologies (YMTC). Notably, it is in this subsector — memory chips — that China has come closest to achieving global parity. Still, the gap at the leading edge remains stark. SMEE, China’s sole commercial lithography equipment maker, is limited to 28nm technology — far behind ASML, which plans to deliver sub-2nm systems this year.
Indeed, the US has long established a lead in the semiconductor industry. Lest we forget, long before Silicon Valley became a shorthand for software and start-ups, it stood for semiconductors — etched in silicon, quite literally. The founding of Shockley Semiconductor Laboratory in 1955 is often cited as the pivotal moment that anchored this region as the beating heart of global tech innovation. That legacy continues. According to the World Intellectual Property Organisation’s 2024 Global Innovation Index, which measures a region’s innovation output per capita, the San Jose-San Francisco cluster ranks second only to Cambridge, the UK. China’s leading cluster, Beijing, ranks 11th.
Silicon Valley playbook
Beyond semiconductors, Silicon Valley offers an instructive lens into the broader mechanisms of the US’ market-driven innovation model. In a society shaped by intense competition and individual ambition, commercial success is often seen as the highest form of validation. As serial entrepreneur and founder of the “lean start-up” movement Steve Blank puts it, “we have a special word for a failed entrepreneur; it’s called experience”. This entrepreneurial mindset is further supported by a legal and regulatory framework that actively encourages risk-taking. The US’ debtor-friendly bankruptcy regime, underpinned by the “fresh start” principle, enables entrepreneurs to try again without being penalised for past failures. Meanwhile, high labour mobility (bolstered by the Federal Trade Commission’s nationwide ban on non-compete agreements last year) fuels talent spillovers and accelerates the diffusion of ideas across firms and sectors.
Strong industry-academic linkages also play a pivotal role. The Stanford-Silicon Valley template, pioneered by Frederick Terman through the establishment of the Stanford Industrial Park in 1951, created a permanent and self-sustaining collaborative research loop between industry and academic institutions that has since seeded more than 5,000 firms. The Bayh-Dole Act of 1980 further incentivised academic research by allowing universities to commercialise, and thus profit from, government-funded research. Meanwhile, the National Science Foundation’s Industry-University Cooperative Research Centers (IUCRC) programme continues to foster academia-industry collaboration by providing shared funding for research that aligns with industry needs. All in all, these structures have made universities key nodes in the US’ innovation network. By contrast, China has struggled to replicate such linkages at scale, with university-industry collaboration remaining a persistent pain point.
For more stories about where money flows, click here for Capital Section
Separately, the US’ deep capital markets have served as a critical engine of innovation. As early as 1946, the establishment of the American Research and Development Corp helped lay the groundwork for the modern venture capital industry, marking the country as a pioneer in this space. In the AI sector, specifically, investment momentum has surged in recent years. The Information Technology and Innovation Foundation reports that, from 2014 to 2024, US-based investors backed more than 9,500 AI companies with total private funding to the tune of around US$600 billion. This compares to China’s US$85 billion for the same period. The unmatched appetite of the US’ private capital ecosystem in funding high-risk, high-reward ventures creates a powerful flywheel: Subsequent commercial successes, most notably represented by the Magnificent 7 companies, are later able to reinvest aggressively in R&D, thus further entrenching US technological leadership.
Together, these advantages have helped the US attract the best and brightest from around the world. While China understandably produces more STEM (science, technology, engineering and mathematics) graduates because of its population size, the US maintains a consistent lead in talent quality. As at 2024, seven of the world’s top 10 universities publishing AI-related research were based in the US. A broader indicator is NeurIPS, widely considered the world’s most prestigious and competitive AI conference: US institutions were involved in roughly half of all accepted papers in 2024 (see Chart 3). Having said that, at the elite level, the landscape is becoming more balanced. In the most competitive category (Oral Presentations, which had an acceptance rate of under 1%), US and Chinese researchers were nearly evenly represented. This points to a narrowing gap in top-tier talent — and raises broader questions about the durability of the US’ advantage.
Wild ‘Trump’-card
Without question, any serious analysis of the US’ position in the AI race must also acknowledge the self-inflicted damage of recent years. The increasingly polarised rhetoric of the US-China rivalry, combined with President Donald Trump’s hard-line immigration and education policies, has contributed to a hostile environment for international talent. This is an especially troubling trend, given that US unicorn companies are predominantly led by immigrants, underscoring just how critical global talent is to the US’ maintaining an edge in cutting-edge innovation.
In contrast, China has been actively dismantling barriers in a bid to close the talent gap. Since launching AI-focused undergraduate tracks in 2018, the number of Chinese universities offering such programmes had grown from 35 to 626 as at 2025. At the same time, a persistent domestic talent shortage in China has eroded the salary premium once granted to overseas returnees, reducing the incentive to study or work abroad. Combined with the aforementioned political headwinds in the US, this has contributed to a growing trend of Chinese AI postgraduates choosing to remain at home. And as MIT Technology Review notes, since AI researchers overwhelmingly prefer to stay in the country where they complete their graduate studies, today’s shift in student flows will almost certainly reshape tomorrow’s talent map. The impact is already visible in the growing “reverse brain drain”, as evidenced by the increasing number of high-profile returns of leading academics, including Guggenheim Fellow Gao Huajian and acclaimed mathematician Sun Song.
Yet, for all its missteps on the talent front, the US appears to be taking decisive steps in other areas that reinforce its position at the forefront of innovation. In January, the White House unveiled Project Stargate, a private-sector initiative backed by OpenAI, SoftBank Group, Oracle and MGX to invest up to US$500 billion in US AI infrastructure. Nvidia pledged an equal sum in April to establish a fully domestic semiconductor production line. Other tech giants, including Meta Platforms, Google and Amazon.com continue to pour capital into developing AI capabilities. At the federal level, the signals appear more mixed, but progress is visible. At present, the Trump administration is working on a comprehensive national AI framework (the “AI Action Plan”) due later this year. The repeal of Biden-era diffusion rules, which proposed a complex three-tier licensing system and would have come into effect in May this year, was welcomed by industry players who were overwhelmingly critical of its overly bureaucratic and restrictive rules. More controversially, Trump’s “One Big Beautiful Bill” included a 10-year moratorium on federal AI regulation. While intended to spur innovation, the proposal raised alarm over the lack of safeguards around safety, bias and misuse. In early July, the “AI moratorium” provision was dropped from the bill, following near-unanimous opposition in the Senate.
Disruption, a two-way street
The core foundations of the US innovation model appear intact. In 2024, US-based institutions led the global charge by developing 40 frontier AI models, more than twice China’s contribution. And while China has visibly narrowed some gaps, it still faces steep challenges in mastering the hardest parts of the AI stack, such as high-end chip production. In short, the American recipe for innovation — anchored by deep capital markets, a risk-tolerant culture and strong academia-industry linkages — remains robust. The US’ capitalist system is dynamic and resilient — with ruthless but fair competition that fosters entrepreneurship. And as long as that system continues to produce results across the metrics that matter, it stands to reason that the US should remain in the lead, albeit by a thinner margin than before — right?
Well, it is worth recalling that for much of the 19th century, the US was viewed as a technological backwater, while Europe held the mantle of global innovation leadership. It was not until the US harnessed the paradigm-shifting forces of the Industrial Revolution that firms such as DuPont, Ford Motor Co and General Motors propelled the nation to global economic prominence. Today, with AI, we are once again living through a generation-defining technological wave. And China, the US’ most formidable rival in this race, stands as a worthy contender. Just as it once engineered an economic miracle, some argue it may now be poised to deliver another — this time, in AI. Could the next era of technological leadership, like the age of mass manufacturing and globalisation before it, also be made in China? Next week, we examine the other half of the AI race: the China paradox.
The Malaysian portfolio gained 0.1% for the week ended July 23. Insas Bhd – Warrants C was up 1 sen (3 cents), to 3 sen per warrant, while United Plantations and Malayan Banking were up 0.8% and 0.6% respectively. On the other hand, Hong Leong Industries (-1.6%), Kim Loong Resources (-0.4%) and LPI Capital (-0.4%) closed in the red last week. Total portfolio returns now stand at 183.2% since inception. This portfolio is outperforming the benchmark FBM KLCI, which is down 16.4% over the same period, by a long, long way.
The Absolute Returns Portfolio fared better, gaining 2% last week, and boosting total portfolio returns since inception to 30.6%. The top three gainers were Tencent Holdings (+6.9%), Alibaba Group Holding (+6.2%) and ChinaAMC Hang Seng Biotech ETF (+4.7%). The sole loser was CrowdStrike Holdings (-1.9%).
The AI portfolio outperformed both the Malaysian and Absolute Returns portfolios last week, up by 2.8%. The gains lifted total portfolio returns since inception to 2.6%. The biggest gainers were Twilio (+8.6%), Workday (+6.3%) and Alibaba (+6.2%). SAP (-5.0%) and ServiceNow (-1.1%) were the only losing stocks.
Disclaimer: This is a personal portfolio for information purposes only and does not constitute a recommendation or solicitation or expression of views to influence readers to buy/sell stocks, including the particular stocks mentioned herein. It does not take into account an individual investor’s particular financial situation, investment objectives, investment horizon, risk profile and/or risk preference. Our shareholders, directors and employees may have positions in or may be materially interested in any of the stocks. We may also have or have had dealings with or may provide or have provided content services to the companies mentioned in the reports.
