Weekly: Beyond export controls; Deep-dive on CXMT; On Nvidia
10 min read.
Highlights
Beyond export controls. A couple great analyses from CSIS on American semiconductor strategy. Chip controls by themselves are not enough and need to be supplemented by domestic investment, innovation, and education. I remember the call for domestic investment was very popular in Washington around 2020, 2021 or so, until it culminated in the CHIPS Act. The CSIS recommendations are all good and common sense, though I suspect the Trump admin likely does not have a grand strategy that ties together different elements of their tech and China strategies.
Deep-dive on CXMT. Korea’s JoongAng Daily does a deep-dive into CXMT, which I have often said does not get enough attention. They are China’s top DRAM company. They were founded in 2016, had roughly 0% market share in 2020, had about 5% market share in 2024, and is expected to have 10-15% market share by the end of 2025. They have the backing of the Chinese state. I have written about their pricing strategies before here. I recommend reading the full JoongAng Daily article.
On Nvidia. FT reviews two new books on Nvidia: The Nvidia Way by Tae Kim and The Thinking Machine by Stephen Witt. Founder and CEO Jensen Huang features prominently as the protagonist, bearing many of the same hallmarks as fellow tech billionaires: intelligence, long-term thinking, conviction, etc.
Thanks for reading.
Table of Contents
Sujai Shivakumar, Charles Wessner, and Thomas Howell, “The Limits of Chip Export Controls in Meeting the China Challenge,” CSIS, 04/14/2025.
Benjamin Jensen, “Winning the Tech Race with China Requires More than Restrictions,” CSIS, 04/17/2025.
Jennifer Lee, “The US Flip-flop Over H20 Chip Restrictions,” The Diplomat, 04/17/2025.
Lee Jae-im, “China's CXMT emerges as silent threat to Samsung, Micron,” JoongAng Daily, 04/14/2025.
Asa Fitch, “The Chip Industry Has Too Many Eggs in the AI Basket,” WSJ, 04/17/2025.
Tim Bradshaw, “How Nvidia became the driving force behind the AI revolution,” FT, 04/12/2025.
Steven Rosenbush, “Nvidia’s Dominance Reflects the CEO’s Unique Business Logic,” WSJ, 04/12/2025.
1.
Sujai Shivakumar, Charles Wessner, and Thomas Howell, “The Limits of Chip Export Controls in Meeting the China Challenge,” CSIS, 04/14/2025.
The U.S. government and those of its allies have imposed and progressively tightened controls on the export of semiconductor technology, devices, and tools to China in an effort to maintain U.S. leadership in this critical sector. China has responded with an all-out effort to stimulate domestic chip innovation capabilities and eliminate dependency on foreign sources. The outcome of the chip innovation race will determine which country leads in the development and application of AI, with major strategic and economic security implications.
Preventing China from acquiring the most advanced chip technology makes sense from a national security perspective, but export restrictions alone cannot substitute for comprehensive industrial and research policy measures necessary to ensure U.S. leadership in semiconductor design, production, and infrastructure. Moreover, the imposition of export controls has resulted in China doubling down on its existing deeply subsidized development efforts that could well produce breakthrough technologies capable of leapfrogging the current state of the art, potentially destabilizing the U.S. semiconductor ecosystem.
Moving from “Protect” to “Promote”
Western chip export controls have had some success in that they briefly set back China’s developmental efforts in semiconductors, albeit at some cost to the United States and allied firms. But current trends underscore that they are at best a short-term palliative for the long-term strategic challenge posed by China in this key sector. The United States and its allies can only address that challenge successfully through comprehensive support for their own chip industries, support that may well exceed the normal parameters of U.S. policy. The unprecedented chips legislation recently enacted in the United States and European Union is part of the response, but far more is needed, particularly expanded public investments in relevant semiconductor research to keep and develop a strong pipeline of well-trained engineers while supporting Western chip firms’ efforts to remain at the technological cutting edge. In the United States, a positive step is the newly formed National Center for the Advancement of Semiconductor Technology, which will be crucial to such an effort, promoting industry-government research and development via the National Semiconductor Technology Center.
2.
Benjamin Jensen, “Winning the Tech Race with China Requires More than Restrictions,” CSIS, 04/17/2025.
If America wants to maintain its edge over China, the answer isn’t more barriers—it’s building more bridges. The United States should seek to retain global talent, expand educational opportunity, develop research infrastructure, and enhance public understanding of how artificial intelligence (AI) actually works to maintain its lead in the ongoing technology competition with the Chinese Communist Party.
There are five immediate steps the Trump administration could take to sustain America’s advantage in the AI race.
1. Fix immigration to retain the world’s best minds. AI’s global talent pool wants to come to the United States—but these individuals increasingly can’t stay in the country.
2. Invest in STEM education and AI literacy—from K–12 to PhDs. Education is central to national security and economic prosperity.
3. Democratize access to high-performance computing. High-performance computing (HPC) is a mix of art and science for applying computational power to perform complex calculations, simulations, and modeling tasks that eclipse the capabilities of standalone desktops.
4. Execute the federal AI strategy. The White House’s new Office of Management and Budget guidance is a great start toward aligning the federal government on the implementation of AI. It requires agencies to appoint chief AI officers, audit their AI systems, and improve transparency. But guidance is not execution.
5. Double down on public innovation funding. From DARPA’s early investments in the internet to the National Science Foundation’s funding of early AI research, federal science funding has built the foundation for every major tech wave in the past 50 years. This legacy must continue.
No amount of export controls will make up for a broken education and immigration system and cuts to science funding.
3.
Jennifer Lee, “The US Flip-flop Over H20 Chip Restrictions,” The Diplomat, 04/17/2025.
The ultimate decision to restrict China’s access to the H20 chips should not come as a surprise. Since returning to the White House, the Trump administration had indicated that it would deepen tech restrictions in the context of China-U.S. competition. This included a move to restrict H20 exports. H20 chips, along with China’s access to a variety of other advanced chips purchased through third parties, may have contributed to Chinese companies’ ability to innovate and work around U.S. limitations.
Two developments were particularly eye-opening. First, Huawei’s Mate60 phone in 2023 included an advanced 7nm chip produced by China’s state-owned SMIC. Second, fast forwarding to this year, DeepSeek’s R1 AI model stunned AI stakeholders as it is competitive with (or even surpasses on some benchmarks) leading American AI models.
The back-and-forth on the H20 chips reflects Washington’s at times improvised approach to tech policy amid competition with China. The Trump administration is moving to cut the budget for BIS and dismissed a senior director on the National Security Council (NSC) staff responsible for technology and national security (David Feith, who also had a background in Asia policy). These actions – which have faced little scrutiny amid the dismissal of several other National Security Council officials and the general trend of changing regulations without public comment or notice – are at odds with Washington’s aims to protect national security and outcompete China in strategic high tech.
Flip-flopping on matters ranging from the expected H20 controls to the tech exemptions for tariffs (which the administration now maintains are just temporary) exacerbates uncertainty for companies trying to make long-term investment decisions while staying competitive with Chinese and other foreign firms.
4.
Lee Jae-im, “China's CXMT emerges as silent threat to Samsung, Micron,” JoongAng Daily, 04/14/2025.
A lesser-known market shaker is ChangXin Memory Technologies (CXMT), a Hefei-based producer that manufactures advanced DRAM such as Double Data Rate 5 (DDR5) without relying on cutting-edge lithography technologies.
Although its market share remains modest at 5 percent, CXMT has the potential to affect market shares previously held by Samsung, Micron and SK hynix with the strong backing of the Chinese government.
CXMT's growth could be particularly detrimental to Samsung, since it already trails behind SK hynix in the high-end HBM segment. CXMT poses a challenge in the commercial lines of memory chips because of its significantly cheaper pricing, with commodity DRAM products at around $1 — half the market price offered by major suppliers Samsung, SK hynix and Micron, according to industry reports.
CXMT has emerged as China’s national champion in memory chip localization, succeeding where others like Fujian Jinhua, Wuhan Hongxin Semiconductor and Tsinghua Unigroup fell short despite heavy state backing — demonstrating that its technological competitiveness is strong enough to drive meaningful gains in market share.
The Chinese company has been pushing to drive up chip production through funding, having secured a funding round of $1.48 billion from national investors in March 2024, which was followed by another $2.4 billion capital infusion from its parent company, Innotron, to build an advanced packaging plant in Shanghai. Samsung Electronics, on the other hand, said it spent 46.3 trillion won ($32 billion) in semiconductor operations in 2024.
Korea’s total memory semiconductor exports in 2024 were $88.29 billion, with China and Hong Kong accounting for 55.4 percent, or $48.93 billion, according to a report from the Korea Institute for International Economic Policy.
5.
Asa Fitch, “The Chip Industry Has Too Many Eggs in the AI Basket,” WSJ, 04/17/2025.
The chip industry has cashed in on the artificial-intelligence bonanza, sending the profits and stocks of companies like Nvidia and Taiwan Semiconductor Manufacturing Co. to unprecedented heights.
With President Trump’s tariff threats and rising potential for a global economic slowdown, it is becoming harder to ignore how everything other than AI isn’t doing all that well.
TSMC, the world’s largest chip maker and an industry bellwether, on Thursday reinforced the sense that the industry’s eggs are largely in the AI basket. It gave strong revenue guidance for the current quarter and maintained capital spending plans despite the potential disruption from tariffs. AI-chip revenue is expected to double this year, and Chief Executive C.C. Wei said in a call with analysts that it was destined for a compound annual growth rate of around 45% in the coming years.
In one way, the optimism is understandable given that the biggest spenders on AI and the chips that drive it—Meta, Alphabet’s Google, Microsoft and Amazon.com—haven’t backed off from huge capital-spending plans.
Chip makers would be in much better shape if rising demand was spread across a broader base, like it was during the Covid pandemic. Back then, no one could get enough chips, whether for cars or computers or smartphones. (AI chips weren’t yet an enormous market.) It was a well-diversified upswing.
But with tariffs and export restrictions in play, investors and chip makers shouldn’t be so sanguine about AI’s ability to keep filling a revenue cup left emptier by other, more sluggish areas.
6.
Tim Bradshaw, “How Nvidia became the driving force behind the AI revolution,” FT, 04/12/2025.
Jensen Huang has often tried to explain what his business does. Yet, even after Nvidia became the world’s most valuable company, many people can barely pronounce its name right, let alone understand how an outfit that started out making graphics chips for video games ended up powering a revolution in artificial intelligence and writing a new chapter in computing history.
Most of the stock market gains were made following ChatGPT’s launch in 2022. That may explain why nobody had written a book about Nvidia before — and why two corporate biographies have now been published within five months.
Tae Kim’s diligent and thorough The Nvidia Way arrived first. But it is Stephen Witt’s The Thinking Machine, out this month, that provides the richer and more accessible account of Nvidia’s 30-year journey from Silicon Valley also-ran to AI behemoth some of whose hardware, he writes, ranks alongside Turing’s Colossus and the Apple II as “one of the most important computers ever built”.
It was the capability of Nvidia’s GPUs to enable the AI revolution that sent Silicon Valley giants scrambling to spend billions on its chips and investors piling into its stock
Both books centre their narrative on Nvidia’s co-founder and chief executive Huang, dubbed “Professor Jensen” by Kim. In a tech industry that venerates founders, Huang is the last of Silicon Valley’s pre-dotcom chief executives still running his own company. His tech icon status is only burnished by his long battle against doubters and, now, ungodly wealth.
Portrayed as a benevolent dictator with a penchant for aphorisms and yelling at underlings, Huang has lived every entrepreneurial trope: the rags to riches immigrant; several near-death experiences; the David vs Goliath battle; and the unlikely gamble that is now finally paying off.
Yet Huang — who now argues Nvidia makes not merely chips but “AI factories” — shares many attributes with Silicon Valley’s best-known names. He has the raw intelligence of Google’s Larry Page and Sergey Brin; Elon Musk’s long-term thinking; Steve Jobs’ knack for predicting what people want before they know it themselves; Mark Zuckerberg’s ability to pivot a multibillion-dollar company on a dime; Jeff Bezos’ underdog mentality. Beyond such Olympian tech-bro traits, both books also highlight Huang’s more everyday human qualities — focus, workaholism, luck and a dry sense of humour.
7.
Steven Rosenbush, “Nvidia’s Dominance Reflects the CEO’s Unique Business Logic,” WSJ, 04/12/2025.
While the scope of the company has expanded over the years from making chips for videogames to producing all manner of AI infrastructure, that principle of restraint has remained.
Focusing on what is both essential and foundational has powered Nvidia’s growth. It’s able to do so many things because it knows what it can’t or won’t do.
“Obviously we did a very poor job explaining what we do,” Huang responded. He acknowledged that Nvidia provides technology across AI infrastructure, networking, switches, storage and of course computing of every size, shape and form. “However—this is the however—we build everything, but we offer it to the world however they would like to take it,” he said. “And the reason for that is we are not a solutions company.”
-