Opposing Data Centers: An Anti-American, Anti-Democratic Threat
Be a skeptic, always.
Howard Marks book The Most Important Thing is a wonderful exploration of critical thinking. Chapter 1 centers around the crucial first step in any thought exercise… second-level thinking.
The recent, and truly rapid, rise of data center development opposition may have far greater consequences than the people opposing realize. This is my attempt to think beyond the current gospel speak at every township’s city council meetings - however, I do understand their frustration, albeit quite misplaced.
The AI race is real, and all American citizens should strongly prefer to not lose.
The AI Superpower Rivalry
The United States still leads China in artificial intelligence — but the gap is shrinking, and the picture is a lot more nuanced than most headlines suggest.
By the numbers, U.S. private AI investment outpaces China’s by roughly 12-to-1, and American companies control about 75% of global AI computing power. Those are massive advantages. But China has closed the frontier model gap from about three years to just six-to-twelve months since 2023. Chinese open-source models now dominate global adoption. China produces 47% of the world’s top AI researchers. And it has structural advantages in energy infrastructure and industrial deployment speed that only get stronger over time.
Whether any of this rises to Cold War levels of threat depends on which “race” you’re actually measuring — and a growing number of serious analysts think the whole “race” framing might be doing more harm than good. The picture that emerges from CSIS, RAND, Stanford HAI, Carnegie, Georgetown CSET, CNAS, and recent government reports is genuinely complicated. Neither complacency nor alarm quite fits.
The Military AI Balance: America Leads, but China Is Building Something Different
The PLA has made “intelligentized warfare” the third phase of its military modernization, following mechanization and informatization. China’s 2019 defense white paper built AI, quantum computing, and big data directly into its warfighting doctrine, and by 2020, intelligentization was a core initiative under Xi Jinping’s goal of a modernized military by 2035. In 2025, private Chinese companies began winning a majority of contracts to build DeepSeek-integrated tools for the PLA — covering everything from battlefield simulation to intelligence analysis to autonomous decision support.
On the autonomous weapons front, things have moved fast. The Atlas drone swarm system, shown on CCTV in March 2026, runs from a single ground vehicle carrying 48 fixed-wing drones. One command vehicle can control up to 96 drones that autonomously avoid collisions, hold formations, and carry out reconnaissance or strike missions. The Jiu Tian drone mothership, flight-tested in June 2025, is a 25-meter wingspan platform that can release 100–150 loitering munitions from internal bays. And PLA-linked research from Beijing Institute of Technology has detailed fully autonomous kill-chain execution in urban environments — no communications required.
The U.S. isn’t sitting on its hands. Project Maven is evolving from an experiment into a proper program of record, with the NGA claiming it will deliver “100 percent machine-generated” intelligence to combatant commanders by June 2026. The Army gave Palantir a $10 billion enterprise agreement in July 2025. The Collaborative Combat Aircraft program committed $8.9 billion over FY2025–2029. DARPA’s RACER program wrapped up final tests in January 2026, producing an autonomy stack that can go on any ground vehicle. And the Air Combat Evolution program pulled off the first AI-versus-human F-16 dogfight — with AI winning 5–0 in simulated engagements.
The real asymmetry here isn’t capability. It’s experience. As CNAS’s Jacob Stokes put it, the PLA “possesses plenty of lethal military power, but right now none of it appears to have meaningful levels of autonomy enabled by AI.” A Defense News analysis from April 2026 pointed out that China simply doesn’t have the volume of real operational data the U.S. has built up over decades — and there’s still an unresolved tension between the decentralized decision-making AI needs and the PLA’s deeply centralized command culture. The Modern War Institute at West Point concluded in March 2026 that the U.S. holds a “commanding” AI lead in military applications.
That said, China may already be ahead in specific areas. It’s taken the lead in drone swarm applications, especially when you pair those swarms with drone carriers already in service. Meanwhile, the U.S. Replicator initiative — launched in 2023 to field “multiple thousands” of unmanned systems — had only produced hundreds by mid-2025. Georgetown CSET analyzed 2,857 PLA procurement documents and found $535.5 million in AI contract values from January 2023 through December 2024.
The Taiwan Question
CSIS ran 24 iterations of a Chinese amphibious invasion of Taiwan set in 2026. In most scenarios, the U.S., Taiwan, and Japan defeated the invasion — but at a staggering cost: 10–20 U.S. warships including 2 aircraft carriers, 200–400 aircraft, and roughly 3,000 troops killed in just three weeks. A separate CSIS blockade wargame (26 iterations) showed that even lower-escalation scenarios produced thousands of casualties, with two games spiraling into general war. PLA drone swarms are being developed specifically for amphibious and “island-blocking” scenarios — targeting areas that analysts note are “suspiciously shaped like Taiwan.”
China’s export of AI-enabled surveillance technology is a separate but significant strategic play. Chinese firms sell facial recognition AI to roughly twice as many countries as the U.S. (83 versus 57), with a clear tilt toward autocracies. Over 80 countries have received Chinese surveillance infrastructure through “safe city” and Digital Silk Road projects, creating network effects that extend Beijing’s influence over how digital governance works worldwide.
Compute and Infrastructure: A Huge U.S. Lead with Some Real Asterisks
The physical foundation of AI power is compute, and by most standard measures, the U.S. advantage here is enormous. The U.S. hosts about 5,500 data centers versus China’s roughly 550. American firms control an estimated 75% of global AI computing power. In terms of H100-equivalent GPUs, one analysis puts the U.S. at approximately 39.7 million versus China’s 400,000. Export controls shifted things dramatically: before October 2022, the U.S. held about 51% of global AI compute versus China’s 33%. After controls, that moved to roughly 74% versus 14%.
U.S. Big Tech is spending at jaw-dropping levels. Alphabet, Amazon, Meta, and Microsoft collectively committed over $650 billion in AI infrastructure spending for 2026. xAI operates what appears to be the world’s largest training cluster — 200,000 GPUs in Memphis. China’s largest known clusters top out around 30,000 GPUs.
But two factors complicate this picture in ways the U.S. data center industry should pay close attention to.
Energy. Nvidia CEO Jensen Huang noted that China has roughly twice the energy generation capacity of the United States — and China’s keeps growing while U.S. capacity has been essentially flat for about 20 years. Morgan Stanley forecasts U.S. data centers could face a 44-gigawatt electricity shortfall within three years. About half of planned U.S. data center builds have already been delayed or canceled because of power infrastructure issues and electrical component shortages. As Brookings put it: while the U.S. has the edge in cutting-edge AI chips, China has a significant advantage in energy.
Construction speed. China’s centralized decision-making means no public opposition to siting, no permitting delays, and surplus solar and battery manufacturing capacity. Huang’s comparison was blunt: U.S. data centers take about three years from groundbreaking, while China “can build a hospital in a weekend.” China announced over 500 new “smart computing center” projects in 2023–2024, though many sit idle due to speculative overbuilding. China’s total AI capital expenditure could hit $98 billion in 2025 — a 48% jump year-over-year. The U.S. commitment is still about 6x larger, but the gap in energy and deployment speed could become the constraint that actually matters.
Semiconductors: America’s Strongest Card and Its Limits
The semiconductor supply chain is the single most powerful lever the U.S. has in this competition. Taiwan produces about 92% of the world’s most advanced logic chips (5nm and below), with TSMC holding 70.2% of global foundry market share. TSMC started mass production of 2nm chips in Q4 2025, with wafer prices approaching $30,000.
ASML’s near-monopoly on EUV lithography might be the most critical chokepoint of all. Every chip at 7nm and below needs EUV for competitive yields, and China can’t get those machines. China is working on alternatives — Reuters reported in December 2025 that a prototype EUV machine was secretly built in Shenzhen with help from former ASML engineers — but experts don’t expect it to produce working chips until 2028–2030.
China’s domestic chip champion, SMIC, is stuck at 7nm for volume production. Their approach uses DUV multi-patterning, which requires 34 lithography steps versus 9 with EUV — a brute-force workaround that comes with serious cost and yield penalties. SMIC’s “5.5nm” process is in pilot testing with yields below 20%, and mass production is unlikely before late 2026. That leaves SMIC two to three full generations behind TSMC, Samsung, and Intel.
On the chip design side, Huawei’s Ascend 910C — China’s best AI accelerator — delivers roughly 60% of an Nvidia H100’s inference performance, according to DeepSeek’s own researchers. And Nvidia is now shipping Blackwell B200 chips that are 4–5x more powerful. The interconnect gap may matter even more: Nvidia’s NVLink is over 10x faster than Huawei’s Unified Bus, which gives the U.S. a decisive edge in scaling massive training clusters. Carnegie’s Alasdair Phillips-Robins estimated that U.S. firms produce 20x more chips (performance-adjusted) than China in 2025, rising to 50x in 2026.
Huawei isn’t sitting still, though. Its CloudMatrix 384 system connects 384 Ascend 910C chips via all-optical mesh, hitting 300 petaFLOPS in BF16. Huawei reportedly shipped about 1 million 910C units in 2025. And there’s an entire ecosystem of Chinese AI chip designers — Cambricon, Biren, Moore Threads, MetaX, Baidu Kunlunxin — competing for SMIC’s limited advanced-node capacity.
Are export controls working?
The case that they’re working looks strong on structural metrics. U.S. share of global AI compute jumped from 51% to 74% after controls went into effect. China’s best chips remain a generation behind, and EUV restrictions prevent advancement beyond 7nm at competitive yields. As CSIS’s Gregory Allen testified, without controls, “it is possible, perhaps even likely, that the first million-chip AI cluster would be built in China.”
The case that they’re not working is just as compelling. Chip smuggling has happened at a scale that seriously undermines U.S. objectives. The March 2026 Supermicro case involved $2.5 billion worth of Nvidia-equipped servers rerouted to China via Taiwan. Operation Gatekeeper dismantled a $160 million smuggling network. CNAS has called chip smuggling a “national security priority.”
The CHIPS Act has gotten real traction, with $33.7 billion awarded across 20 projects in 21 states. TSMC’s Arizona Fab 1 began 4nm production on schedule in the first half of 2025, and the U.S. is projected to produce 28% of the world’s advanced logic chips by 2032 — up from essentially zero in 2022. But most fabs won’t reach volume production until 2026–2028, and U.S. construction costs run 4–5x what they are in Taiwan.
DeepSeek and the Efficiency Revolution
The January 2025 release of DeepSeek-R1 was probably the single most disruptive moment in the AI competition since export controls began. The model matched OpenAI’s o1 on reasoning benchmarks while reportedly training for just $5.6 million on 2,048 H800 GPUs. The market’s reaction was immediate and dramatic: Nvidia lost roughly $589 billion in market cap in a single day — the largest one-day loss in U.S. stock market history.
The technical innovations were real: a fine-grained Mixture of Experts with 256 experts per layer (activating only 37 billion of 671 billion total parameters per token), Multi-Head Latent Attention, FP8 mixed-precision training, and PTX-level reprogramming of restricted H800 chips. Stanford HAI faculty concluded that DeepSeek “challenged preconceived notions regarding capital and computational resources necessary for serious advancements in AI.”
That said, the true costs were almost certainly much higher than the headline figure. SemiAnalysis estimated DeepSeek used up to 50,000 H-series chips across all research stages, putting total investment above $1.3 billion. CSIS’s Gregory Allen noted the $5.6 million figure excluded costs for prior research and ablation experiments.
The broader Chinese open-source ecosystem has become a strategic force in its own right. Alibaba’s Qwen passed Meta’s Llama as the most-downloaded model family on Hugging Face in September 2025, with over 700 million cumulative downloads and 100,000+ derivative models. Chinese open-source models went from 1.2% of global usage in late 2024 to roughly 30% by late 2025. One Andreessen Horowitz partner estimated that 80% of U.S. startups use Chinese base models for their derivative work.
Investment and Talent: Two Very Different Stories
Stanford HAI’s 2025 AI Index put U.S. private AI investment at $109.1 billion versus China’s $9.3 billion — nearly 12-to-1. Five U.S. companies alone planned over $450 billion in AI capex for 2026. But government spending tells a different story. China launched a National AI Industry Investment Fund with $8.2 billion in initial capital in January 2025, then announced $138 billion over 20 years for AI and quantum technology in March 2025. Recorded Future assessed that China’s overall government-led funding likely exceeds what U.S. federal and state governments are spending combined.
The talent pipeline is where the argument that “China is closer than people think” really holds up. China now produces 47% of the world’s top AI researchers by undergraduate origin, up from 29% in 2019. Nearly 40% of top AI talent at U.S. institutions originally came from Chinese universities. Tsinghua University ranked first globally for NeurIPS papers accepted from 2021–2024, with 643 papers. At ICLR, Chinese papers went from a 5-to-1 disadvantage in 2021 to near parity by 2025.
A Hoover Institution study found that over half of DeepSeek’s researchers were trained entirely in China, and of those with U.S. affiliations, about 70% now work in China. At least 85 scientists left the U.S. for Chinese institutions since early 2024, with more than half departing in 2025. The U.S. still has a structural advantage in retaining elite talent — roughly 87% of top Chinese AI researchers in the U.S. stay here, drawn by compensation premiums ($185K versus $67K). But the trend line is moving in the wrong direction.
China’s Digital Silk Road: Building AI Influence Worldwide
While the frontier model race gets the headlines, China’s strategy to build AI infrastructure across the developing world may end up mattering more. Digital Silk Road investments reached an estimated $79 billion by 2018 and have grown significantly since, spanning over 80 countries. Huawei supplies 5G infrastructure, cloud services, and data centers across Africa, the Middle East, Southeast Asia, Central Asia, and Latin America.
In cloud market share, the big three U.S. providers (AWS at 29%, Azure at 20%, Google Cloud at 13%) control about 62–63% of the $107 billion quarterly global market. Alibaba Cloud holds roughly 4% globally but dominates within China at 36% domestic share, with Huawei Cloud at 19%.
The strategic concern is straightforward: if Chinese AI infrastructure, open-source models, and digital standards become the default across the developing world, that creates network effects and vendor lock-in through Chinese-controlled systems. DeepSeek is already being used by HSBC, Standard Chartered, and Saudi Aramco. Singapore’s national AI program chose Qwen. Indonesia’s Indosat partnered with a DeepSeek-based AI firm.
How Does This Play Out?
Scenarios where China overtakes the U.S. require some specific things to go wrong. The most credible is a scaling plateau combined with Chinese efficiency innovation — if frontier model performance flatlines despite massive U.S. spending, China’s algorithmic efficiency advantages become decisive. Georgetown CSET’s Helen Toner put it bluntly: “If performance plateaus despite all the spending by OpenAI and others — a growing concern in Silicon Valley — China has a chance to compete.” The second scenario is export control erosion: if millions of advanced chips reach China through smuggling or policy changes, the compute advantage shrinks from 10x or more to just 1.2–5x.
Scenarios where the U.S. holds its lead rest on several pillars: export controls stay effective, allied cooperation continues, the 12-to-1 private investment ratio holds, ASML’s EUV monopoly keeps constraining Chinese chip fabrication, the 18-nation Tier 1 alliance network stays intact, and Nvidia’s CUDA software ecosystem maintains its lock-in.
The most likely outcome, according to the weight of serious analysis, is neither overtaking nor permanent U.S. dominance. As Recorded Future assessed: the competition will likely tighten, with China as a close second globally and its models possibly outperforming the U.S. at times or in specific sectors.
China’s Large-Scale Robotics Fleet Deserves Greater Attention
The deployment gap is real and potentially decisive. China deploys 5x more industrial robots, has integrated AI into over 60% of large manufacturers, and runs “dark factories” with minimal human involvement. The Stimson Center warned that “China may already be years ahead” in AI deployment at scale.
The energy constraint could end up being what actually decides this. China’s energy surplus and construction speed may become the bottleneck for AI infrastructure — not chip performance. U.S. data centers face years of delays from grid limitations and permitting challenges.
There’s a trust gap working against the U.S. 83% of Chinese respondents see AI as beneficial, versus 39% in the United States. If Western publics resist AI adoption, deployment stalls no matter how good the technology is.
And open-source proliferation creates real dependencies. With Chinese models making up 30% of global AI workloads and 80% of U.S. startups building on Chinese base models, the global supply chain for AI software is getting harder and harder to audit.
So What Does It All Mean?
The evidence doesn’t fully support calling this “the greatest global threat since the Cold War”- yet. But it is almost certainly the most consequential technology competition since the nuclear era — it just works through fundamentally different mechanisms.
The U.S. has decisive advantages in compute infrastructure (75% of global AI computing power), private investment (12-to-1), frontier model capability (6–12 months ahead), and semiconductor chokepoints. Those are real.
But the U.S. also has three genuine vulnerabilities. First, talent — China produces nearly half the world’s top AI researchers, the reverse brain drain is picking up speed, and U.S. immigration policies are actively pushing skilled scientists toward Chinese institutions. Second, deployment speed — China’s advantages in energy, construction, regulatory flexibility, and public trust could make it the dominant deployer of AI in the real economy even if it never builds the most powerful model. Third, the open-source ecosystem — Chinese models now dominate global open-weight AI, creating dependencies that extend Beijing’s influence through software rather than hardware.
For the data center industry in particular, the binding constraint is shifting from chips to power. China’s roughly 2x energy generation advantage, combined with a projected 44-gigawatt U.S. electricity shortfall and a 50% cancellation rate on planned builds, suggests the U.S. AI infrastructure lead could erode — not because China builds better chips, but because America can’t power enough data centers fast enough. That’s the part of this competition that needs the most urgent attention from industry, and it’s the part that gets the least discussion in national security circles.
References
All sources accessed March–April 2026 unless otherwise noted.
[1] Stanford HAI. “The 2025 AI Index Report: Economy.” Stanford University, 2025. https://hai.stanford.edu/ai-index/2025-ai-index-report/economy
[2] ETF.com / Nasdaq. “Beyond the Headlines — The AI Race Explained: U.S. vs China.” March 2026. https://www.etf.com/sections/features/beyond-headlines-ai-race-explained-us-vs-china
[3] AI HIVE. “US and China AI Infrastructure: 2025 Perspective.” Comparative Analysis, 2025. https://www.ai-hive.net/post/comparative-analysis-of-us-and-china-ai-infrastructure-and-development-a-2025-perspective
[4] The Wire China. “Scoring the AI Race, a Year After the DeepSeek Shock.” March 22, 2026. https://www.thewirechina.com/2026/03/22/scoring-the-ai-race-a-year-after-the-deepseek-shock/
[5] TrendForce. “Chinese AI Models Reportedly Hit ~15% Global Share in Nov. 2025, Fueled by DeepSeek Open-Source Push.” January 26, 2026. https://www.trendforce.com/news/2026/01/26/
[6] Sovereign Magazine. “China’s AI Rise: Innovation Overcomes Chipmaking and Investment Gaps.” 2025. https://www.sovereignmagazine.com/science-tech/artificial-intelligence/chinas-ai-rise-innovation-overcomes-chipmaking-investment-gaps/
[7] Brookings Institution. “How Will the United States and China Power the AI Race?” 2025. https://www.brookings.edu/articles/how-will-the-united-states-and-china-power-the-ai-race/
[8] Foreign Affairs. Multiple authors. “The Myth of the AI Race” and related analyses, 2025–2026.
[9] LSE United States Politics and Policy. “Rather Than Framing AI Competition as a ‘Race’ with China, the US Should Promote Greater Regulation.” April 2, 2026. https://blogs.lse.ac.uk/usappblog/2026/04/02/
[10] Nunosempere Forum. “Enforcement Without Experience: Military AI and China — Responsible AI in Military Contexts, Part 2 of 5.” 2025. https://forum.nunosempere.com/posts/dj4guht9a4mXu4ijG/
[11] Asia Times. “New PLA Unit Underscores Intelligentized Warfare Shift.” April 2024. https://asiatimes.com/2024/04/new-pla-unit-underscores-intelligentized-warfare-shift/
[12] RealClearDefense. “DeepSeek: PLA’s ‘Intelligentized Warfare.’” November 18, 2025. https://www.realcleardefense.com/articles/2025/11/18/
[13] DefenseScoop. “New Pentagon Report on China’s Military Notes Beijing’s Progress on LLMs.” December 26, 2025. https://defensescoop.com/2025/12/26/
[14] Army Recognition. “China’s New Atlas Drone Swarm System Demonstrates How Algorithm-Driven Warfare Becomes Operational.” March 2026. https://www.armyrecognition.com/news/aerospace-news/2026/
[15] CNA. “China Readies Drone Swarms for Future War.” September 2025. https://www.cna.org/our-media/indepth/2025/09/china-readies-drone-swarms-for-future-war
[16] The Diplomat. “Machines in the Alleyways: China’s Bet on Autonomous Urban Warfare.” February 2026. https://thediplomat.com/2026/02/
[17] Wikipedia. “Project Maven.” Updated 2026. https://en.wikipedia.org/wiki/Project_Maven
[18] GovCon Wire. “Pentagon to Designate Palantir Maven AI as Program of Record.” 2025. https://www.govconwire.com/articles/pentagon-palantir-maven-ai-program-of-record
[19] CSIS. “The Department of Defense’s Collaborative Combat Aircraft Program: Good News, Bad News, and Unanswered Questions.” 2025. https://www.csis.org/analysis/department-defenses-collaborative-combat-aircraft-program
[20] Airforce Technology. “Collaborative Combat Aircraft (CCA), US.” Updated 2025. https://www.airforce-technology.com/projects/collaborative-combat-aircraft-cca-usa/
[21] Defence Blog. “DARPA Completes RACER Autonomous Vehicle Testing.” January 2026. https://defence-blog.com/darpa-completes-racer-autonomous-vehicle-testing/
[22] The Debrief. “DARPA’s Groundbreaking ACE Program and X-62A Becomes First AI-Controlled Jet to Dogfight Against Manned F-16.” 2024. https://thedebrief.org/darpas-groundbreaking-ace-program/
[23] CNAS. “How AI Is Changing Warfare.” Congressional Testimony, 2025. https://www.cnas.org/publications/congressional-testimony/how-ai-is-changing-warfare
[24] Defense News. “Outpaced by the US, China’s Military Places Selective Bets on Artificial Intelligence.” April 7, 2026. https://www.defensenews.com/global/asia-pacific/2026/04/07/
[25] Congress.gov / CRS. “DOD Replicator Initiative: Background and Issues for Congress.” IF12611. https://www.congress.gov/crs-product/IF12611
[26] Small Wars Journal / Naval News. CSIS Wargame: China’s Invasion of Taiwan and CSIS Blockade Wargame. 2023–2025. https://smallwarsjournal.com/2025/08/02/ and https://www.navalnews.com/naval-news/2023/01/
[27] CNA. “PRC Concepts for UAV Swarms in Future Warfare.” October 2025. https://www.cna.org/analyses/2025/10/prc-concepts-for-uav-swarms-in-future-warfare
[28] Brookings. “Exporting the Surveillance State via Trade in AI.” 2023. https://www.brookings.edu/articles/exporting-the-surveillance-state-via-trade-in-ai/
[29] Biometric Update. “Report Finds US Technology Still Flowing into China’s Surveillance System.” November 2025. https://www.biometricupdate.com/202511/
[30] NextBigFuture. “China AI Chip and AI Data Centers Versus US AI Data Centers.” October 2025. https://www.nextbigfuture.com/2025/10/
[31] Fortune. “Nvidia CEO Says Data Centers Take About 3 Years to Construct in the U.S., While in China ‘They Can Build a Hospital in a Weekend.’” December 6, 2025. https://fortune.com/2025/12/06/nvidia-ceo-jensen-huang-ai-race-china-data-centers/
[32] Tom’s Hardware. “Half of Planned US Data Center Builds Have Been Delayed or Canceled.” 2026. https://www.tomshardware.com/tech-industry/artificial-intelligence/
[33] Brookings. “How Will the United States and China Power the AI Race?” 2025. https://www.brookings.edu/articles/how-will-the-united-states-and-china-power-the-ai-race/
[34] MIT Technology Review. “China Built Hundreds of AI Data Centers to Catch the AI Boom. Now Many Stand Unused.” March 2025. https://www.technologyreview.com/2025/03/26/
[35] Economy Insights / EE Times Asia. “Taiwan’s Chip Leadership” and “Inside Taiwan’s Semiconductor Supremacy.” 2025. https://www.economyinsights.com/p/taiwans-chip-leadership
[36] Semiecosystem (Substack). “TSMC Dominates Q2 Foundry Rankings.” 2025. https://marklapedus.substack.com/p/tsmc-dominates-q2-foundry-rankings
[37] SemiWiki. “TSMC’s 2nm Chips: The Results Are Out.” 2026. https://semiwiki.com/forum/threads/tsmcs-2nm-chips-the-results-are-out.24329/
[38] Wikipedia. “ASML.” Updated 2026. https://en.wikipedia.org/wiki/ASML_Holding
[39] Medium / Marc Bara. “The Advanced Semiconductor Supply Chain: Why Money Is Not Enough.” January 2026. https://medium.com/@marc.bara.iniesta/
[40] Granitefirm. “How Is SMIC After US Embargo?” March 2025. https://www.granitefirm.com/blog/us/2025/03/08/smic-after-us-embargo/
[41] Semiecosystem (Substack). “Can China Make 5nm Chips?” 2025. https://marklapedus.substack.com/p/can-china-make-5nm-chips
[42] Tom’s Hardware. “DeepSeek Research Suggests Huawei’s Ascend 910C Delivers 60% of Nvidia H100 Inference Performance.” 2025. https://www.tomshardware.com/tech-industry/artificial-intelligence/
[43] Carnegie Endowment for International Peace. “Don’t Panic Yet Over AI Chip Sales to China.” December 2025. https://carnegieendowment.org/emissary/2025/12/china-ai-chip-sales-nvidia-trump
[44] Tech Startups. “Huawei’s Ascend 910C-Powered System Reportedly Outperforms Nvidia’s H100 on Key Metrics.” April 2025. https://techstartups.com/2025/04/28/
[45] TrendForce. “Cambricon Remains China’s Top AI Chip Startup; Rumored 2026 Triple Output Faces SMIC Limits.” December 2025. https://www.trendforce.com/news/2025/12/15/
[46] HelloChinaTech. “China’s Compute Bet: Can Efficiency Replace Scale?” 2025. https://hellochinatech.com/p/china-compute-efficiency-vs-scale
[47] CSIS. “Countering China’s Challenge to American AI Leadership.” Gregory C. Allen testimony. 2026. https://www.csis.org/analysis/countering-chinas-challenge-american-ai-leadership
[48] Tech Insider. “Super Micro $2.5B Chip Smuggling Case.” Updated April 2026. https://tech-insider.org/super-micro-nvidia-chip-smuggling-china-2026/
[49] Arnold & Porter. “DOJ Announces Shutdown of Major China-Linked AI Tech Smuggling Network Through Operation Gatekeeper.” December 2025. https://www.arnoldporter.com/en/perspectives/blogs/enforcement-edge/2025/12/
[50] CNAS. “Countering AI Chip Smuggling Has Become a National Security Priority.” 2025. https://www.cnas.org/publications/reports/countering-ai-chip-smuggling
[51] Smart Incentives. “Tracking CHIPS Act Incentives.” Updated 2026. https://smartincentives.org/tracking-chips-act-incentives/
[52] Part Locator. “CHIPS Act 2025 Timeline: When Will US Chip Shortages Actually End?” 2025. https://partlocator.com/blog/chips-act-2025-semiconductor-supply-chain-impact
[53] Centre for International Governance Innovation. “DeepSeek and China’s AI Innovation in US-China Tech Competition.” 2025. https://www.cigionline.org/articles/deepseek-and-chinas-ai-innovation/
[54] arXiv. “Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures.” 2025. https://arxiv.org/html/2505.09343v1
[55] Intuition. “DeepSeek’s Market Shock: What You Need to Know.” 2025. https://www.intuition.com/deepseeks-market-shock-what-you-need-to-know/
[56] Graphcore Research Blog / Vitalab. “DeepSeek-V3 & DeepSeek-R1 Technical Reports.” 2025. https://graphcore-research.github.io/deepseek/
[57] Stanford HAI. “How Disruptive Is DeepSeek? Stanford HAI Faculty Discuss China’s New Model.” 2025. https://hai.stanford.edu/news/how-disruptive-deepseek
[58] Digital in Asia. “Inside China’s AI Machine: Models, Chips, and Strategy.” April 6, 2026. https://digitalinasia.com/2026/04/06/china-ai-models-chips-strategy/
[59] South China Morning Post. “China’s Open-Source Models Make Up 30% of Global AI Usage, Led by Qwen and DeepSeek.” 2025. https://www.scmp.com/tech/tech-trends/article/3335602/
[60] Recorded Future. “US-China AI Gap: 2025 Analysis of Model Performance, Investment, and Innovation.” 2025. https://www.recordedfuture.com/research/measuring-the-us-china-ai-gap
[61] Second Talent. “USA vs China in AI & LLM: Statistics & Market Analysis [2025].” 2025. https://www.secondtalent.com/resources/usa-vs-china-ai-llm-statistics/
[62] Federal Reserve. “The State of AI Competition in Advanced Economies.” FEDS Notes, October 6, 2025. https://www.federalreserve.gov/econres/notes/feds-notes/
[63] Hoover Institution. “A Deep Peek Into DeepSeek AI’s Talent and Implications for US Innovation.” 2025. https://www.hoover.org/research/deep-peek-deepseek-ais-talent
[64] Council on Foreign Relations. “China’s Digital Silk Road Initiative.” Interactive tracker. https://www.cfr.org/china-digital-silk-road/
[65] China-Global South Project. “Digital Silk Road 2.0: How China Is Shaping Central Asia’s AI Ecosystem.” 2026. https://chinaglobalsouth.com/analysis/china-central-asia-ai-cooperation/
[66] Synergy Research Group. “Cloud Market Share Trends — Big Three Together Hold 63%.” Q3 2025. https://www.srgresearch.com/articles/cloud-market-share-trends
[67] AAG IT Support. “The Latest Cloud Computing Statistics.” Updated October 2025. https://aag-it.com/the-latest-cloud-computing-statistics/
[68] The Asia Live. “Digital Silk Road: How China’s Quiet AI Revolution Is Rewiring Global Influence.” April 12, 2025. https://theasialive.com/digital-silk-road/
[69] Atlantic Council. “It’s Time to Reckon with the Geopolitics of Artificial Intelligence.” 2025. https://www.atlanticcouncil.org/content-series/inflection-points/
[70] Georgetown CSET. “The AI Cold War That Will Redefine Everything.” February 2026. https://cset.georgetown.edu/article/the-ai-cold-war-that-will-redefine-everything/
[71] Carnegie Endowment. “Peril and Promise in the U.S.-China AI Race.” Pivotal States Podcast. 2025. https://carnegieendowment.org/podcasts/pivotal-states-podcast/peril-and-promise-in-the-us-china-ai-race
[72] Epoch AI (Substack). “Why China Isn’t About to Leap Ahead of the West on Compute.” 2025. https://epochai.substack.com/p/why-china-isnt-about-to-leap-ahead
Additional sources consulted: RAND Corporation, AEI, Stanford FSI, Contrary Research, CFR, Newsweek, Yole Group, TechTarget, CSIS wargame reports, U.S. Army Military Law Review, Belfer Center (Harvard), and various Congressional Research Service reports.
Mr. Fischer. Your articles and insights are genuinely appreciated.