Weekly: Global tech race; On export controls; AMD vs Nvidia
7 min read.
Highlights
Global tech race. Ever since Trump revoked Biden’s AI Diffusion Rule, there’s been a slew of think pieces opining on what Washington should do next. In that vein, there’s two pieces today: one from CFR experts writing in Foreign Affairs, and another from a CNAS expert writing in Foreign Policy.
The FP piece lays out a strategic framework for “how the United States can win the global tech race,” arguing for greater US investments, and more commercial and technological diplomacy.
The FA piece asks “what if China wins the AI race?” which I think is a refreshing perspective in a climate where the ‘winner-takes-all’ mentality (rightly or wrongly) spirals into a hands-off-the-wheel posture as we advance rapidly towards new technologies. The authors challenges the ‘winner-takes-all’ assumption baked into DC debates and outlines de-risking strategies that allow the US to benefit even if it finishes second.
On export controls. CSIS’ Navin Girishankar interviewed former Deputy Assistant Secretary at the Commerce Dept Bureau of Industry and Security, the body in charge of export controls, in a podcast about economic security and innovation.
AMD vs Nvidia. SemiAnalysis offers analysis on AMD’s latest chips, which were in the news a lot this past week as a potential competitor to Nvidia. The bottom line is that AMD is in a “wartime stance” to compete with Nvidia and though their latest seems roughly on par with Nvidia’s Blackwell chips for small to medium models inferencing, Nvidia’s networking capabilities mean their racks (chips strung together) have much stronger performance over AMD’s.
Thanks for reading.
Table of Contents
Sebastian Elbaum and Adam Segal, “What If China Wins the AI Race?,” Foreign Affairs, 06/13/2025.
Vivek Chilukuri, “How the United States Can Win the Global Tech Race,” Foreign Policy, 06/09/2025.
Navin Girishankar and Matthew S. Borman, “Export Controls to Secure U.S. Innovation featuring Matthew Borman,” CSIS, 06/18/2025.
Teng Hung Chen, Dylan Patel, Daniel Nishball, Wega Chu, Ivan Chiam, Patrick Zhou and Gerald Wong, “AMD Advancing AI: MI350X and MI400 UALoE72, MI500 UAL256,” SemiAnalysis, 06/13/2025.
1.
Sebastian Elbaum and Adam Segal, “What If China Wins the AI Race?,” Foreign Affairs, 06/13/2025.
Technology executives, national security analysts, and U.S. officials all seem to agree that the United States must win the AI competition over China. Washington has pursued a two-pronged strategy in its bid for supremacy: constraining China by restricting the export of key technological components and accelerating domestic innovation on foundational AI models. To achieve the latter goal, both administrations have pursued a program of relatively light regulatory oversight of industry leaders, targeted investment in semiconductors and energy infrastructure, and encouraging the adoption of AI across the federal government, especially defense and intelligence agencies, for uses as varied as investigating outbreaks of foodborne diseases and detecting financial fraud.
By the end of 2023, the leading U.S. models were overperforming their Chinese counterparts by double-digit percentage points in response accuracy. But China has quickly closed the gap with a savvy combination of government initiatives such as the Next Generation AI Development Plan, an emphasis on AI education and workforce training, mighty research investment, close coordination between Beijing and the tech industry, and massive public investment in data centers, energy transmission, and semiconductor manufacturing. Meanwhile, China has taken the lead in integrating AI into high-tech manufacturing.
Thankfully, Washington can find alternative strategies to ensure that the United States benefits from AI progress even if it does not win the innovation competition outright.
First, the United States should look for new ways to demonstrate the merits of its models to global markets. The U.S. National Institute of Standards and Technology, through the AI Safety Institute and industry partners, could promote new evaluation frameworks for foundational AI models.
U.S. companies must build systems and applications that run on foundational models but mitigate the risks of relying on any one specific model. Incorporating another software layer between applications and foundation models, known as an intermediate abstraction layer, can isolate the downstream systems from the foundational model, making them more independent and resilient.
Lastly, the United States should regulate the types of data U.S. developers and companies share with foreign model builders without defaulting to blanket bans on data transfers. Washington may understandably be wary of sharing U.S. data with China for privacy and national security reasons, but there may be instances where the economic or societal benefits of fine-tuning a Chinese model with U.S. data outweigh the risks.
Finishing second is not a death knell for U.S. AI, but refusing to adapt to compete would be.
2.
Vivek Chilukuri, “How the United States Can Win the Global Tech Race,” Foreign Policy, 06/09/2025.
The Trump administration has scrapped its predecessor’s sweeping export controls for advanced artificial intelligence chips, known as the AI diffusion rule.
As the administration decides what comes next, it should raise its sights from merely proposing a “simpler” rule to manage the diffusion of AI chips. Instead, it should seize the opportunity to offer an ambitious vision to promote the broader diffusion of U.S. technology.
To start, Washington must finally learn from its failure in the transition to 4G and 5G telecommunications networks, where Beijing’s state-backed model—and the absence of a compelling U.S.-led alternative—enabled Huawei and ZTE to all but corner emerging markets. Huawei now operates in more than 170 countries worldwide and is the top global provider of telecommunications equipment. But if there is broad consensus among U.S. policymakers that Beijing won that global technology transition, there is little agreement about how to win the next.
Washington can start with reforms in three broad areas. First, unleash the United States’ strategic investment tools. One of Washington’s most promising but underused tools is the International Development Finance Corporation (DFC). Created during the first Trump administration, the DFC makes market-driven investments to advance both humanitarian and national security goals, and it has several tools to attract private capital from equity investments to political risk insurance.
Second, Washington should turbocharge its commercial diplomacy for technology. Between 2016 and 2020, an average of just 900 U.S. personnel from the State and Commerce departments were deployed abroad for commercial diplomacy, and just a fraction focused on technology.
Finally, the United States should embrace a newly ambitious vision for technology partnerships. Too often, U.S. and allied firms lose one-off bids to subsidized, politically backed Chinese competitors, even if the firms might prefer to align with the high-tech U.S. ecosystem. Washington should explore how to make such an offer without simply imitating Beijing’s state-led model.
Washington can also do more to align with technology-leading allies on joint investments in strategic emerging markets. For example, Washington could better coordinate with Japan’s Overseas Development Assistance program to boost Open RAN networks across the Indo-Pacific, tap the European Union’s Global Gateway to connect subsea cables to Africa, and support India’s Digital Public Infrastructure to counter China’s “smart city” offerings.
3.
Navin Girishankar and Matthew S. Borman, “Export Controls to Secure U.S. Innovation featuring Matthew Borman,” CSIS, 06/18/2025.
Matthew Borman joins Navin Girishankar to discuss the evolution of U.S. export controls as a key economic security tool and their effectiveness at securing innovative and sensitive technologies. Matt is a non-resident Senior Technical Expert at CSIS and served as principal deputy assistant secretary for strategic trade and technology security and deputy assistant secretary for export administration at the Department of Commerce’s Bureau of Industry and Security for more than 20 years.
4.
Teng Hung Chen, Dylan Patel, Daniel Nishball, Wega Chu, Ivan Chiam, Patrick Zhou and Gerald Wong, “AMD Advancing AI: MI350X and MI400 UALoE72, MI500 UAL256,” SemiAnalysis, 06/13/2025.
For the past six months, AMD has been in a Wartime stance. They have been working hard and working smart towards their goal of being competitive with Nvidia. At its Advancing AI 2025 event, AMD launched the MI350X/MI355X GPUs which could be competitive to Nvidia’s HGX B200 solutions for inference of small to medium LLMs on a performance per TCO basis. Notwithstanding the reality distortion field projected by AMD, the MI355X is not a rack scale product, and it is not competitive against Nvidia’s GB200 NVL72 at frontier model inference or training.
Executive Summary
The MI355X is competitive with the HGX B200 for small to medium model inferencing but it will not be competitive against the GB200 NVL72
Despite AMD’s marketing RDF, the MI355 128 GPU rack is not a “rack scale solution” – it only has a scale up world size of 8 GPUs versus the GB200 NVL72 which has a world size of 72 GPUs. The GB200 NVL72 will beat the MI355X on Perf per TCO for large frontier reasoning model inference
The MI355X will have similar collective performance as the HGX B200 but MI355X collectives will run at least 18x slower than on the GB200 NVL72, if not even slower
AMD announced its Developer Cloud, which will bring on demand pricing to $1.99/hr/GPU for the MI300 vs $3.00/hr/GPU in the current AMD Neocloud market in a move that could potentially make renting AMD GPUs competitive vs renting Nvidia GPUs
Nvidia’s DGX Lepton Marketplace has upset a lot of Neoclouds, potentially giving AMD an opening to convince Neoclouds to support both Nvidia and AMD
AMD is finally adopting a similar strategy to Nvidia and using its strong balance sheet to support the Neocloud and hyperscale ecosystem in adopting AMD by renting a portion of GPUs back from the clouds. This will help drive accelerate end user adoption of AMD’s systems.
The MI400 Series will be a rack scale solution that could potentially be competitive with Nvidia’s VR200 NVL144 in H2 2026
-