Introduction: The Five-Layer AI Economy and the Rise of Compute Nationalism

On the morning of May 13, 2026, President Donald Trump boards Air Force One for Beijing, accompanied by more than a dozen of America’s most powerful corporate executives: Elon Musk of Tesla, Tim Cook of Apple, Larry Fink of BlackRock, Kelly Ortberg of Boeing, David Solomon of Goldman Sachs, Stephen Schwarzman of Blackstone, Jane Fraser of Citigroup, Cristiano Amon of Qualcomm, Chuck Robbins of Cisco, and others. The summit agenda covers trade, Taiwan, the Iran war, and — above all else — artificial intelligence. The centerpiece of that last item is a question neither side has been able to resolve for three years: who controls the hardware on which the world’s intelligence infrastructure runs?

The story of one executive and that flight captures the entire geopolitical drama in miniature. On May 11, 2026, Reuters first reported that Jensen Huang — founder and chief executive of NVIDIA, the company that more than any other has become the physical substrate of the global AI economy — had not been invited to join the delegation.1 The White House, according to sources, was focused on agriculture and commercial aviation for the trip; the chipmaker’s absence was read immediately, and correctly, as a signal that NVIDIA’s sales in China were unlikely to recover soon. Huang had already told CNBC’s Jim Cramer that it would be “a great honor to represent the United States” if invited — his reticence diplomatic, his exclusion pointed.

Then, on the afternoon of May 13, 2026, Reuters reported again: Huang would join after all.2 After news of his absence circulated widely in the press, Trump called Huang directly and asked him to come. Huang flew to Alaska to board Air Force One mid-journey. “Jensen is attending the summit at the invitation of President Trump to support America and the administration’s goals,” an NVIDIA spokesperson confirmed to CNBC.3 In the space of forty-eight hours, the most commercially and strategically consequential CEO in the AI economy had gone from uninvited, to conspicuously absent, to boarding the president’s plane on short notice — a sequence that is, in miniature, the entire story of America’s tortured relationship with NVIDIA’s role in the U.S.–China AI war.

The reason the sequence matters — commercially, geopolitically, and strategically — is the subject of this paper. And it cannot be understood without a framework adequate to the moment.

“AI is no longer a single breakthrough or application — it is essential infrastructure. Every company will use it. Every nation will build it. From energy and chips to infrastructure, models and applications, every layer of the stack is advancing at once.”

— Jensen Huang, NVIDIA GTC 2026 Keynote

Artificial intelligence is almost universally described as a software revolution. That description is convenient, familiar, and structurally incomplete. Software is the visible layer of AI — the part users experience through chat interfaces, copilots, synthetic media, autonomous agents, and embedded machine intelligence across enterprise platforms. But beneath every visible AI interaction sits a far larger physical infrastructure whose scale, capital intensity, and geopolitical significance are beginning to reshape global power in ways that software narratives alone cannot capture.

Modern frontier AI depends on electricity generation, semiconductor design, wafer fabrication, advanced packaging, extreme-ultraviolet lithography, cooling systems, high-bandwidth networking fabrics, datacenter construction, cloud orchestration, model engineering, and deployment architectures capable of operating at industrial scale. The economics of this stack increasingly resemble energy infrastructure, transportation backbones, telecommunications networks, and defense industrial bases — not software startups.

Economic historian Chris Miller of Tufts University’s Fletcher School, whose landmark study of the semiconductor industry became one of the defining texts of the geopolitical era, captured this structural shift with precision: “Semiconductors have defined the world we live in, determining the shape of international politics, the structure of the world economy, and the balance of military power.”4 In 2026, his observation applies with even greater force to the specific class of AI accelerators that have become the scarce industrial input at the center of the world’s most consequential technological competition. As Miller further observed: “You can’t understand the modern world without putting chips at the center of the story.”5

That competition has a precise center of gravity: NVIDIA’s GPU chips for AI. Huang has described AI as “a five-layer cake”6 — energy at the base, then chips, then infrastructure (datacenters and networking), then models, then applications at the top. The insight is not merely taxonomic. It is strategic. Each layer depends on the layers below it. Applications are the visible tip; energy is the invisible foundation. NVIDIA sits at the second layer — chips — but its influence radiates upward through every layer above it. Whoever controls that second layer shapes the structure of all others.

Harvard political scientist Graham Allison, whose concept of the Thucydides Trap has become the defining framework for understanding U.S.–China great-power dynamics, warned that the two countries’ relationship will constitute “a ruthless rivalry… across nearly every dimension — tech, trade, industry, military, and global influence.”7 Artificial intelligence has become the most contested terrain within that rivalry, and the contest over NVIDIA’s chips its most visible front.

This paper argues that the contemporary confrontation over AI hardware, semiconductor ecosystems, rare earth leverage, industrial subsidies, sovereign cloud infrastructure, and strategic capital flows should not be understood merely as a ‘chip war.’ That phrase is useful but too narrow. This is a struggle over control of the infrastructure required to generate intelligence at industrial scale — and it requires a new conceptual framework equal to its scope.

I call this framework Compute Nationalism.

Compute Nationalism is the doctrine under which nation-states treat computational infrastructure — including semiconductor design, advanced manufacturing ecosystems, datacenters, power systems, rare earth supply chains, AI accelerators, inference capacity, and sovereign cloud infrastructure — as strategic national assets to be protected, expanded, subsidized, weaponized, or denied in pursuit of geopolitical power. This paper proceeds through five sections: defining the doctrine, examining America’s containment architecture, analyzing China’s counteroffensive, surveying the global spread of the framework, and extracting strategic lessons for corporations, startups, investors, and governments. Throughout, the Beijing summit of May 2026 serves as both its immediate occasion and its living demonstration.

Section 1: Defining Compute Nationalism — State Sovereignty in the Age of Industrial Intelligence

Every geopolitical transition eventually exposes the inadequacy of inherited vocabulary. The digital era produced useful concepts: cybersecurity, platform governance, digital sovereignty, supply chain resilience, techno-nationalism. None of them fully captures what is emerging around artificial intelligence infrastructure, and the reason is structural.

Most earlier digital frameworks were built on an assumption that information moved faster than physical constraints — that software scaled cheaply, cloud infrastructure could be rented, capital moved efficiently across borders, and globalization distributed industrial capacity to wherever it was most economical. Those assumptions are weakening. Artificial intelligence at frontier scale is materially constrained in ways that earlier software revolutions were not. Compute Nationalism begins with that recognition.

Its central premise is straightforward: computational infrastructure has become a strategic national asset because the production of intelligence increasingly depends on scarce industrial capacity. A startup building a consumer application faces modest infrastructure requirements. A nation seeking sovereign frontier AI capability confronts a categorically different set of constraints: access to advanced semiconductor fabrication, packaging ecosystems, GPU allocation, grid-scale electricity, cooling infrastructure, high-bandwidth networking, sovereign cloud orchestration, and inference deployment economics. These are infrastructure questions. Infrastructure questions invite state intervention. Thus, Compute Nationalism.

Compute Nationalism vs. Digital Sovereignty

Digital sovereignty focuses on control over information systems: where data is stored, who governs privacy, which cloud providers host sensitive workloads, whether foreign governments can access national information infrastructure. These are important questions, but they are downstream of the more fundamental question that Compute Nationalism addresses: who controls the physical machinery required to train the models that make sense of that information, and deploy intelligence at the scale of an economy or a military? Data without compute cannot train frontier models. Privacy law does not create fabrication ecosystems. Cloud jurisdiction does not manufacture accelerators. Digital sovereignty addresses governance. Compute Nationalism addresses industrial capability.

Compute Nationalism vs. Techno-Nationalism

Techno-nationalism encompasses national efforts to strengthen domestic technological competitiveness across industries through industrial policy, localization mandates, strategic investment, or trade intervention. The category includes electric vehicle subsidies, battery manufacturing, quantum computing initiatives, and telecommunications competition. Compute Nationalism is more precise — it focuses specifically on intelligence-producing infrastructure, systems whose output is not a product but a capability: the capacity to generate, process, and deploy AI at scale. Compute infrastructure combines properties that are rare in combination: extreme capital intensity, supply constraint, energy dependence, military relevance, commercial indispensability, and physical scarcity. This combination makes compute exceptional and makes Compute Nationalism a distinct conceptual category.

Compute Nationalism vs. Economic Nationalism

Economic nationalism prioritizes domestic industry through tariffs, protectionism, local manufacturing incentives, and trade barriers. Its central concerns are employment, trade balances, and domestic production. Compute Nationalism extends beyond these economic concerns to something more fundamental: strategic intelligence capacity. A semiconductor fabrication plant is not merely a manufacturing asset generating jobs and export revenue. It is national intelligence infrastructure. A hyperscale datacenter is not merely commercial real estate. It is sovereign compute capability that shapes who can generate intelligence and at what speed. That difference — between economic asset and strategic capability — is what makes Compute Nationalism a distinct doctrine.

Compute Nationalism vs. AI Nationalism

AI nationalism, as sometimes used in policy discussions, tends to focus on the model layer — on whether a nation has its own large language models, AI systems reflecting its values, or domestically developed AI products. This is a legitimate concern. But it addresses only the top layer of Huang’s five-layer stack. Compute Nationalism concerns the entire stack — the energy, chips, and infrastructure without which no model layer can exist. A nation can build domestic AI models and remain entirely dependent on foreign compute infrastructure for their training and deployment. Compute Nationalism addresses this deeper dependency. The framework is not about which models exist; it is about who controls the industrial capacity to generate intelligence at scale.

Intelligence Has Become Industrial

For most of human history, intelligence was treated primarily as a function of human capital — education, institutions, research culture, labor quality. Artificial intelligence transforms that model. Intelligence now has industrial inputs: power, silicon, cooling, packaging, networking, manufacturing ecosystems, and datacenter deployment. This industrialization is not incidental. It determines who can produce intelligence, at what scale, under what political conditions, and at what cost.

Huang’s GTC 2026 formulation is worth quoting precisely because of its deliberate scope: “AI is no longer a single breakthrough or application — it is essential infrastructure. Every company will use it. Every nation will build it.”8 The phrase ‘every nation will build it’ is, in effect, a prediction of Compute Nationalism. When intelligence production requires industrial infrastructure that states can build, subsidize, or deny, states will inevitably behave as sovereign actors in that space. This is not ideology. It is institutional logic. CNAS observed in its 2025 analysis of global compute security: “Controls on semiconductor manufacturing equipment going to China were imposed as early as 2019, followed by AI chip export controls targeting U.S. adversaries in 2022, which were strengthened in 2023, 2024, and 2025.”9 The escalation pattern is not coincidental — it is the structural logic of Compute Nationalism playing out in real time.

Scarcity Creates Strategy

If frontier compute were universally abundant and freely tradeable, geopolitical competition over it would be substantially muted. But it is not. Scarcity exists simultaneously across multiple layers: accelerator supply, advanced packaging bottlenecks, EUV lithography concentration, grid-scale power limitations, permitting delays, rare earth processing dependencies, and the sheer capital intensity of fab construction. Each layer of scarcity creates a corresponding leverage point for the state that controls it and a corresponding vulnerability for the state that depends on it. The architecture of Compute Nationalism is, at bottom, an architecture of scarcity management. The central strategic question of our era is not ‘Who has the best AI software?’ but ‘Who controls the infrastructure required to generate intelligence at industrial scale?’

Section 2: America’s Architecture of Compute Containment

The most common mischaracterization of American AI export policy is the description of it as trade regulation. What Washington has constructed over the past four years is not a regulatory framework in any conventional sense. It is an attempt to shape the geography of intelligence production — to determine, through administrative instruments, which nations can build AI at frontier scale and which cannot. This is infrastructure strategy executed through licensing thresholds, entity lists, product classifications, and allied coordination. It deserves a more precise name: compute containment.

The Export Control Timeline: A Four-Year Escalation

The architecture of American compute containment was built in escalating layers over four years. October 2022: the Biden Commerce Department issued initial export controls restricting the A100 and H100 GPUs to China, setting performance thresholds based on total processing performance (TPP). October 2023: the rules were tightened further, adding restrictions on the A800 and H800 — chips NVIDIA had designed specifically to fall below the original thresholds. January 2025: the Biden administration issued the “AI Diffusion Rule,”10 establishing a global three-tier licensing framework: Tier 1 (eighteen allied nations, no restrictions), Tier 2 (most of the world, subject to quantity caps), and Tier 3 (approximately twenty adversarial countries, including China, Russia, Iran, and North Korea — full prohibition). Blackwell B200 and related chips were categorically denied to Tier 3.

April 2025: the Trump administration imposed a new requirement — a license for export of even NVIDIA’s downgraded H20, the chip designed specifically to comply with prior rules. NVIDIA had warned of the risk in SEC filings; the license requirement materialized without warning. The financial consequences were immediate and disclosed: NVIDIA recorded a $4.5 billion charge11 in Q1 FY2026 for H20 excess inventory and purchase obligations, and projected an $8 billion loss in H20 revenue for Q2 FY2026 — a total impact exceeding $12 billion within six months of the April 2025 restriction.

December 8, 2025: President Trump announced, via Truth Social, that the United States would allow NVIDIA to ship its more advanced H200 chip to China — one generation behind Blackwell — under new terms: strict licensing and a 25 percent government surcharge on each sale. January 13–15, 2026: the BIS issued a final rule shifting H200 and AMD MI325X export license review from ‘presumption of denial’ to “case-by-case review,”12 provided applicants demonstrated that exports would not reduce U.S. chip supply, that purchasers had adopted export compliance procedures, and that chips had undergone independent third-party testing in the United States before shipment to China. The Blackwell B200 and all next-generation architectures remained under categorical denial. As of May 2026, not a single H200 has been delivered to China, and NVIDIA’s AI chip market share in China stands at precisely zero.

NVIDIA as Strategic Chokepoint

NVIDIA’s trajectory from graphics processor manufacturer to strategic chokepoint is one of the more remarkable institutional transformations in modern corporate history. Its dominance over the AI accelerator market emerged through compounding advantages: the CUDA software ecosystem, which created extraordinary developer lock-in across millions of researchers and engineers; hardware performance leadership that consistently exceeded alternatives; software maturity built over fifteen years; hyperscaler standardization; and startup dependency on a single vendor for the compute that determines whether frontier model training is viable at all.

The financial consequence of this dominance — and its geopolitical implications — became visible in NVIDIA’s China exposure. China once represented 20 to 25 percent of NVIDIA’s data center revenue. CFO Colette Kress confirmed in Q1 FY2026 earnings that the Chinese AI accelerator market alone was expected to grow to nearly $50 billion annually13 — a market NVIDIA has now lost entirely. Huang’s assessment of the resulting policy was delivered with unusual bluntness in an interview with the Special Competitive Studies Project in May 2026: “Conceding an entire market the size of China probably does not make a lot of strategic sense, so I think that has already largely backfired.”14

Huang’s deeper argument is architectural, not merely commercial. In his April 2026 interview with Dwarkesh Patel, he framed it in terms of the five-layer stack: “Why are you causing one layer of the AI industry to lose an entire market so that you could benefit from another layer of the AI industry? There are five layers, and every single layer has to succeed.”15 His concern is that the chip layer — layer two — is being sacrificed in a way that accelerates China’s adoption of a non-American AI technology stack. Every Chinese AI developer who pivots to Huawei’s Ascend platform weakens CUDA’s global dominance and strengthens an alternative ecosystem that can be exported to other nations. “It would be extremely foolish to create two ecosystems,” he warned: one open-source on a foreign tech stack and one closed on the American stack. “That would be a horrible outcome for the United States.”16

Semiconductor Equipment Warfare: The Deeper Chokepoint

Finished chips dominate the public narrative of compute containment. Manufacturing tools may determine its long-term outcome. The semiconductor supply chain is not a single industry but a layered civilizational infrastructure, and the tools required to manufacture advanced chips are even more geopolitically concentrated than the chips themselves.

ASML, the Dutch company that holds a monopoly on extreme-ultraviolet lithography equipment, sits at the center of this architecture. The Netherlands has barred ASML from shipping EUV systems to China since 2019, under coordinated pressure from Washington. As of September 2023, the Netherlands extended licensing requirements to ASML’s advanced deep ultraviolet (DUV) immersion systems — the NXT:2000i, 1980i, and 1970i models — effectively closing China’s ability to stock up on the second-best lithography option. By early 2024, ASML confirmed it was unlikely to receive export licenses for advanced immersion DUV systems for Chinese customers.17 ASML CFO Roger Dassen acknowledged in October 2024 that China revenues were expected to fall to approximately 20 percent of total revenue in 2025 — a 48 percent year-over-year decline — driven primarily by U.S.-coordinated export restrictions.

Japan joined the coalition in 2023, restricting Tokyo Electron and Nikon from shipping certain advanced semiconductor manufacturing equipment to China. Applied Materials, Lam Research, and KLA Corporation — the three dominant American semiconductor equipment makers — were covered under U.S. export rules that restricted their most advanced deposition, etch, and inspection systems from Chinese advanced-node customers. The CNAS noted in December 2025 that ASML continues to service existing machines in China despite U.S. pressure to restrict servicing, “potentially extending machine lifespans to 30 years”18 — a gap in the containment architecture that Washington was actively pressing the Netherlands to close.

EDA software constitutes an additional invisible chokepoint. Electronic design automation tools from Synopsys and Cadence are essential for designing advanced semiconductors. U.S. export restrictions on EDA software for Chinese entities have progressively tightened, denying Chinese chipmakers access to the design environment required for next-generation chip architecture. Materials, gases, and specialty chemicals — the chemical layer of semiconductor manufacturing — add yet another control surface. The architecture of compute containment is not a single policy; it is a multi-layer chokepoint system spanning design software, lithography, deposition, etch, inspection, and materials.

The Remote Access Security Act and the Cloud Loophole

In January 2026, the House of Representatives passed the Remote Access Security Act by a 369-to-22 vote, closing a significant regulatory gap that had allowed Chinese companies to rent access to export-controlled AI chips through offshore data centers without technically violating existing law. INF Tech had rented 2,300 Blackwell GPUs through an Indonesian data center; Tencent had secured contracts for 15,000 Blackwell processors via a Japanese provider.19 The legislation extended export control logic to cloud computing for the first time, treating certain forms of remote access to controlled AI compute as legally equivalent to physical chip export.

The Chip-Smuggling Underground

A parallel dimension of the compute containment story is chip smuggling. Experts estimated that approximately 140,000 advanced chips worth $5 to $7 billion were smuggled into China in 2024 alone.20 On December 8, 2025 — the same day Trump announced the H200 export approval — federal prosecutors unsealed Operation Gatekeeper, revealing a massive China-linked smuggling network that had changed shipping labels on NVIDIA H100 and H200 chip cartons to bear the name of a fictional brand, ‘Sandkyan,’ to evade export controls. The investigation documented falsified paperwork and sophisticated evasion techniques across multiple logistics chains. The existence of an active smuggling market at this scale indicates both the intensity of Chinese demand for American AI chips and the limits of administrative containment.

The Strategic Logic and Its Paradox

Washington’s core thesis is straightforward: advanced AI infrastructure may accelerate Chinese military modernization, cyber capability, autonomous weapons development, intelligence automation, and industrial productivity in ways that threaten U.S. strategic superiority. The security logic exists and is not trivial. What is genuinely contested is proportionality: whether the strategic benefit of denial outweighs the strategic cost of accelerating China’s domestic semiconductor ecosystem and fracturing the global AI technology stack. “It’s in the best interest of America to serve that China market,” Huang argued at the APEC CEO Summit in South Korea in October 2025. “It’s in the best interest of China to have the American technology.”21 The Beijing summit of May 2026 is, in part, an attempt to navigate exactly this paradox.

Section 3: China’s Counteroffensive — Compute Sovereignty Through Industrial Retaliation

If America’s doctrine is compute containment, China’s doctrine is compute sovereignty. The distinction matters because Beijing’s response is frequently mischaracterized as reactive improvisation. It is more coherent and more structurally grounded than that characterization allows. Great powers do not absorb strategic pressure passively. They reinterpret pressure as structural threat and redesign national behavior accordingly.

From Beijing’s perspective, dependence on foreign computational infrastructure constitutes a strategic vulnerability of the first order. If accelerators, manufacturing ecosystems, EDA software, advanced packaging, cloud dependencies, and capital access can be politically restricted, they cannot serve as durable foundations of national AI capability. That recognition transforms the strategic question from ‘How do we acquire more advanced foreign infrastructure?’ to ‘How do we eliminate dependence on infrastructure we do not control?’ That is compute sovereignty — and it is the Chinese expression of Compute Nationalism.

Rare Earth Retaliation: Weaponizing the Material Layer

Beijing’s most immediate retaliatory instrument has been the material layer of the AI stack. On December 3, 2024, China’s Ministry of Commerce (MOFCOM) issued Notice 2024 No. 46, announcing an immediate ban on exports of gallium, germanium, antimony, and superhard materials to the United States — effective the same day Washington announced its latest round of semiconductor export restrictions targeting 140 Chinese entities.22 The timing was deliberate. Gallium is essential to compound semiconductors and advanced radar; germanium is critical for infrared technology, fiber optic cables, and certain semiconductor processes; antimony is used in military applications including flame retardants and ammunition primers.

The U.S. Geological Survey estimated the combined economic impact of the gallium and germanium export ban at a potential $3.4 billion GDP loss for the United States, with roughly half the decrease concentrated in the semiconductor sector.23 China processes approximately 80 percent of the world’s gallium and 60 percent of its germanium — processing dominance that cannot be quickly replicated through alternative supply chains. The ban signaled that Beijing was willing and able to weaponize the material layer of the AI stack as symmetrically as Washington had weaponized the chip and equipment layers. In November 2025, as part of a broader diplomatic de-escalation following a tariff truce, China suspended the ban until November 27, 2026 — framing the suspension as a revocable gesture rather than a permanent concession.

Huawei Ascend: Building the Alternative Stack

No initiative better illustrates Chinese compute sovereignty than Huawei’s Ascend AI chip program. Washington initially targeted Huawei as a strategic security threat — cutting its access to advanced chips, American software, and global supply chains. Beijing responded by treating Huawei as the primary vehicle of domestic technological sovereignty. The pressure intended to weaken Huawei as a commercial competitor instead elevated it as a national infrastructure institution.

Huawei’s Ascend 910C, released in 2025, combines two 910B chips into a single package delivering 800 TFLOP/s of computing power at FP16, with 3.2 TB/s memory bandwidth — performance on par with NVIDIA’s H100 GPU, which is no longer available in China. The chip is manufactured by SMIC using a 7nm DUV-based process, which suffers from yield rates estimated by Mizuho Securities at approximately 30 percent — significantly constraining production volume and increasing unit cost.24 Huawei shipped an estimated 700,000 Ascend chips across its product line in 2025 and plans to manufacture 600,000 units of the 910C alone in 2026 — roughly double 2025 output — with total Ascend line production of up to 1.6 million dies.25

The significance of the Ascend program extends well beyond raw chip benchmarks. Huawei is simultaneously building the software ecosystem — its CANN compute architecture framework and MindSpore AI development framework — required to make the Ascend platform a viable alternative to NVIDIA’s CUDA. Every Chinese AI developer who builds on Ascend rather than CUDA creates a dependency relationship that reduces future demand for American chips, and potentially enables export of Chinese AI applications on Chinese hardware to third countries — precisely the outcome Huang warned about.

SMIC and Hua Hong: The Domestic Manufacturing Foundation

China’s domestic semiconductor manufacturing ecosystem centers on SMIC (Semiconductor Manufacturing International Corporation) and Hua Hong, the country’s two largest foundries. SMIC’s 7nm N+2 process — achieved without EUV lithography — has demonstrated that Chinese manufacturing can produce chips competitive with older-generation international nodes, even at constrained yield rates. Morgan Stanley analysis estimated SMIC capacity of approximately 7,000 wafers per month for advanced AI chip production in 2025, with a capacity target of 18,000 wafers per month by 2027.26 SMIC’s trajectory is the directional signal that matters: each year brings higher yields, greater capacity, and deeper institutional learning in processes that previously required foreign equipment or expertise.

The strategic significance of domestic manufacturing is not immediate performance parity with TSMC. It is dependency reduction. A China that can produce advanced chips at 70 percent of international performance benchmarks, using entirely domestic equipment and processes, is strategically far less vulnerable than one that performs at 95 percent but depends on a supply chain that can be politically severed overnight.

ByteDance, Alibaba, Tencent: The Corporate Demand Signal

The corporate dimension of Chinese compute sovereignty is visible in the behavior of China’s largest technology companies. ByteDance, Alibaba, and Tencent were approved by China to import over 400,000 NVIDIA H200 units in early 2026 — representing the first major batch of conditional H200 import approvals.27 Yet as of May 2026, none of those chips have been delivered. China’s customs authorities blocked H200 imports despite the formal U.S. regulatory approval, signaling that Beijing is demanding terms beyond Washington’s licensing framework — including domestic chip co-purchasing requirements and approved customer designations. This behavior is not commercial negotiation. It is sovereign posturing: Beijing refuses to accept American-controlled access on Washington’s terms, and is using corporate demand as a lever to extract better strategic conditions.

The ByteDance-led orders for 2026 were reported to total upward of $14 billion28 — a figure that illustrates the scale of Chinese demand for American AI hardware and, simultaneously, the scale of the commercial opportunity NVIDIA continues to forgo. The gap between potential and realized revenue is not a market failure. It is the material cost of the compute war.

Investment Nationalism and Capital Controls

China’s compute sovereignty agenda extends to the capital layer. The U.S. Outbound Investment Security Program, effective January 2, 2025, restricts American persons from investing in Chinese entities engaged in certain activities related to semiconductors, quantum information technologies, and AI systems.29 This ‘reverse CFIUS’ mechanism operates alongside traditional CFIUS inbound review to create a bilateral capital firewall around AI infrastructure. Chinese AI startups that once accessed American venture capital now face a more restricted funding environment, particularly for companies involved in frontier AI training hardware, advanced semiconductor design, or dual-use AI models.

Capital nationalism extends to the structure of Chinese AI investment within China. The Chinese government has directed an estimated $150 billion annually into its semiconductor sector through policy banks, state-owned enterprise partnerships, and government guidance funds. This sovereign capital deployment is not subject to return-on-investment discipline in the way private capital is — it is strategic patience embodied in balance sheets. Chinese foundries, packaging companies, and AI chip designers can sustain longer development timelines and lower yields than their private-market international competitors because the capital behind them is politically patient.

The Paradox of Strategic Pressure

The most important thing to understand about China’s response to American export controls is that the pressure appears to have accelerated, rather than contained, China’s sovereign AI ambitions. Huang recognized this dynamic explicitly: “In China, we have now dropped to zero,” he said in May 2026. “Conceding an entire market the size of China probably does not make a lot of strategic sense, so I think that has already largely backfired.”30 Huawei’s AI chip revenue is projected to reach $12 billion in 2026, with a 60 percent market share in China’s AI accelerator market by year-end.31 NVIDIA’s market share stands at zero. This is not the outcome compute containment was designed to produce.

Section 4: Global Case Studies in Compute Nationalism

One of the most important tests of any new geopolitical framework is portability. If a doctrine explains only one bilateral rivalry, it may be descriptive shorthand. If it predicts behavior across multiple actors with different political systems, economic structures, and strategic contexts, it represents a genuine conceptual category. Compute Nationalism passes that test. Similar strategic behaviors are increasingly observable across the global order. Different countries express the doctrine differently — some through industrial subsidies, some through sovereignty mandates, some through strategic control, some through modernization incentives — but the structural logic remains consistent: computational infrastructure is increasingly treated as strategically consequential national capability.

United States: The CHIPS Act as Compute Nationalism

American criticism of Chinese industrial policy becomes more complicated when examined alongside the CHIPS and Science Act of 2022 — a $280 billion federal commitment to domestic semiconductor manufacturing expansion, research incentives, and supply chain reshoring. As of November 2025, the Commerce Department had allocated over $36 billion in awards to 40 projects across 19 companies.32 The largest recipients: Intel ($7.9 billion for fabs in Arizona, New Mexico, Ohio, and Oregon), TSMC ($6.6 billion for three advanced fabs in Phoenix), Samsung ($6.4 billion for advanced logic manufacturing in Taylor, Texas), and Micron ($6.165 billion for memory chip production).

TSMC’s first Arizona fab began mass production of 4nm chips in early 2025, reporting yields 4 percent higher than comparable Taiwan facilities. The investment trajectory behind that facility is striking in scale: TSMC’s original $12 billion commitment in 2020 grew to $65 billion across three fabs by 2024, and on March 3, 2025 — in an announcement made alongside President Trump at the White House — TSMC announced an additional $100 billion for three further fabs, two advanced packaging facilities, and a dedicated R&D center, bringing the total Arizona commitment to $165 billion — the largest single foreign direct investment in U.S. history.33 The GAO estimated in December 2025 that these investments would bring the U.S. share of global leading-edge logic chip manufacturing from 0 percent in 2022 to 20 percent by 2030.34 This is industrial policy for compute infrastructure executed at national scale — the American expression of Compute Nationalism, justified in the language of competitiveness, resilience, and national security.

China: Made in China 2025 and AI Intensification

China’s Made in China 2025 initiative reflected strategic concern over dependency in critical industrial sectors long before AI became the dominant policy frame. Targets included robotics, advanced manufacturing, semiconductors, AI-adjacent technologies, industrial automation, and new materials — the industrial foundations of what would later become AI infrastructure. The initiative’s structural logic maps directly onto Compute Nationalism: reduce dependency, build sovereign industrial capability, localize strategic ecosystems, enhance resilience. Artificial intelligence intensified the urgency without changing the doctrine. Chinese compute nationalism did not emerge as a sudden reaction to American pressure. It emerged from preexisting industrial sovereignty logic that predates the current chip war by nearly a decade.

Germany: Industrie 4.0 and Competitive Modernization

Germany presents a less militarized but structurally relevant case. Industrie 4.0 — the national initiative to modernize manufacturing through digital integration, automation, and smart factory technologies — reflects a form of Compute Nationalism that operates through competitive modernization rather than strategic denial. Germany’s concern is not that a foreign adversary will restrict its chip access, but that falling behind in the digitization of its manufacturing base will erode the industrial competitiveness on which its economic model depends. The framework is defensive and market-oriented rather than adversarial — but it produces similar sovereign infrastructure investments: national AI research centers, advanced manufacturing subsidies, and strategic compute infrastructure investment. Compute nationalism can exist without overt confrontation; it can emerge from competitive modernization logic alone.

Russia: Sovereign Control Without Compute Abundance

Russia presents a lower-capability but conceptually important case. Russian Compute Nationalism has historically emphasized informational sovereignty and internet control — the sovereign internet infrastructure, the RuNet architecture designed to operate independently of global internet protocols in the event of external disconnection — rather than frontier compute leadership. Russia lacks the domestic semiconductor ecosystem, the capital, or the allied manufacturing relationships to build frontier AI infrastructure. Its expression of Compute Nationalism is therefore control-first: limit foreign platform penetration, mandate domestic data storage and processing, and establish the technical architecture for digital autarky. This is Compute Nationalism operating at the governance layer in the absence of industrial capacity at the chip layer.

UAE and Saudi Arabia: Sovereign Capital Meets AI Infrastructure

Perhaps the most strategically interesting emerging expressions of Compute Nationalism appear in the Gulf. In May 2025, President Trump and UAE President Mohamed bin Zayed Al Nahyan unveiled a 5-gigawatt UAE-U.S. AI Campus in Abu Dhabi — a 10-square-mile facility that will be the largest AI infrastructure project outside the United States.35 Abu Dhabi’s AI conglomerate G42 will build and operate the campus in partnership with American hyperscalers. The first phase deploys 1 gigawatt of capacity using NVIDIA’s Grace Blackwell GB300 systems, the most advanced AI platform NVIDIA offers — chips approved for export to the UAE despite the broader Tier 2 restrictions, subject to strict security and reporting requirements.

Saudi Arabia’s Humain initiative, backed by the Public Investment Fund, is building AI factories with projected capacity of 500 megawatts powered by several hundred thousand NVIDIA GPUs over five years. In May 2025, Nvidia and Humain unveiled a strategic partnership that included an initial deployment of an 18,000-unit GB300 Blackwell supercomputer.36 AWS is building a new Saudi cloud region expected to become available in 2026, with a committed investment exceeding $5.3 billion; Google Cloud and the PIF announced a $10 billion AI-hub partnership the same month. These are not merely commercial transactions. They represent sovereign wealth converting hydrocarbon capital into compute influence — a pattern that may define Gulf geopolitics in the AI era as definitively as oil defined it in the twentieth century.

The CSIS observed that this Gulf strategy creates a structural tension for Washington: as the United States exports compute capacity abroad to maintain strategic relationships, it necessarily cedes some leverage over governance standards and deployment norms. The UAE is already a pioneer in smart city AI adoption, and Saudi Arabia’s Vision 2030 places AI at the center of its post-oil economic model. These states are not passive recipients of American AI exports — they are active architects of sovereign compute capacity, using American hardware to build strategic infrastructure they intend to own and control.

India: Middle-Power Strategic Balancing

India’s position is structurally distinctive: a large and rapidly digitizing population, exceptional talent depth, strategic balancing between Washington and Beijing, nascent semiconductor ambitions, and significant infrastructure constraints. India illustrates a middle-power version of Compute Nationalism — strategic recognition that computational sovereignty matters, combined with limited current capacity to achieve it unilaterally. New Delhi has pursued bilateral AI and semiconductor partnerships with both the United States (through the iCET initiative) and has maintained economic relationships with Russia and China, resisting lock-in to any single technological alignment. This balancing is itself a form of Compute Nationalism: resisting dependency while accumulating sovereign capability incrementally.

The Global Stratification Risk

Across these cases, expression varies but structural logic converges. States increasingly ask: Do we control enough compute to remain strategically relevant? Are we dangerously dependent? Should AI infrastructure be subsidized, protected, or owned by the state? Should foreign investment in AI ecosystems be scrutinized? These are compute nationalist questions, and they are being asked simultaneously in Washington, Beijing, Riyadh, Abu Dhabi, Berlin, New Delhi, Tokyo, and Seoul. The likely structural outcome is a stratified global order: a small number of sovereign compute powers capable of full-stack AI independence; a second tier of alliance-dependent participants achieving partial sovereignty through allied relationships; and a larger tier of infrastructure consumers whose AI capabilities are fundamentally shaped by choices made above them. Strategic dependence may become the defining geopolitical hierarchy of the AI era.

Section 5: Strategic Lessons from the Compute War

Geopolitical frameworks earn their value through the quality of strategic guidance they generate. Compute Nationalism is not merely descriptive — it is predictive. If the doctrine is correct, certain strategic consequences follow for every major actor type in the AI economy. What follows is a practical map of those consequences.

For Corporations: Compute Infrastructure as Geopolitical Risk

Supply Chain Resilience

For multinational corporations, the most immediate lesson is that compute infrastructure is no longer politically neutral. Historically, firms optimized AI infrastructure for cost, performance, and vendor reliability. Compute Nationalism introduces geopolitical variables that most corporate planning processes were not designed to handle. The questions that now demand boardroom attention include: Can export controls disrupt our AI supply chain? Can foreign subsidiaries lose access to models or compute we depend on? Can political escalation alter our cloud licensing agreements? Are our AI dependencies concentrated in ecosystems vulnerable to sudden restriction?

GPU Access Risk

GPU access is the most immediate operational risk. The H20 ban of April 2025 demonstrated that the U.S. government can restrict access to chips specifically designed for compliance without warning and with immediate financial consequence. Any corporation that has built AI infrastructure, cloud services, or AI-native products around a specific chip generation faces existential supply risk if that chip is reclassified. Vendor diversification — maintaining viable relationships with AMD, Intel, and domestic cloud providers alongside NVIDIA — is no longer merely a commercial best practice; it is strategic risk management.

Sovereign Compute Planning

Infrastructure regionalization is emerging as the necessary corporate response: separate compute footprints, compliance architectures, deployment stacks, and data governance frameworks for different geopolitical zones. This increases cost and organizational complexity. But Compute Nationalism makes such architecture increasingly rational, and the companies that invest in regional infrastructure resilience now will be better positioned when the next round of restrictions arrives — as it will. AI infrastructure can no longer be delegated solely to engineering teams; it has become board-level governance encompassing power constraints, vendor concentration, geopolitical dependencies, and cross-border deployment risk.

For Startups: The End of Infrastructure Abstraction

Hyperscaler Dependency

The startup economy has historically thrived on infrastructure abstraction: rent cloud resources, scale software cheaply, expand globally, raise capital efficiently. Frontier AI disrupts this model. Many startups now depend on hyperscalers not merely for hosting but for survival — which creates a structural asymmetry: the infrastructure landlord may also be a competitor. Compute Nationalism amplifies this because infrastructure access may become politically mediated. A startup whose core product depends on compute that is restricted, rationed, or geopolitically conditioned has a strategic vulnerability that no software innovation can fully offset.

Inference Scarcity

As model deployment scales from training to inference at commercial scale, inference cost and inference access become the binding constraints. Inference scarcity — the limited availability of GPU capacity for serving AI models at industrial throughput — is already reshaping startup economics. Startups that secure compute access through privileged hyperscaler relationships, strategic investor partnerships, or early infrastructure commitments gain structural advantage over competitors with superior models but limited serving capacity. Compute Nationalism intensifies inference scarcity by constraining the global supply of advanced AI accelerators.

Compute Concentration

Capital itself is becoming geopolitical in frontier AI sectors. Cross-border funding scrutiny, national security review of foreign investment, and strategic technology classification are intensifying. The U.S. Outbound Investment Security Program, effective January 2025, restricts American persons from investing in Chinese AI and semiconductor entities; analogous frameworks are emerging in Europe, Japan, and South Korea. Startups with significant foreign ownership stakes — particularly from countries of concern — face increasingly complex regulatory environments as compute and AI become classified as strategic infrastructure.

For M&A: National Security Revaluation of AI Assets

CFIUS and AI

The CFIUS framework has been progressively expanded to cover AI infrastructure as a strategic asset category. The White House updated its Critical and Emerging Technologies list in February 2024 to specifically include advanced AI systems. Any foreign stake in AI companies that process substantial data, develop dual-use models, or operate critical compute infrastructure now triggers heightened CFIUS scrutiny — even minority positions, if the investor receives governance rights or access to sensitive technical information.37 In February 2025, President Trump’s ‘America First Investment Policy’ directed CFIUS to intensify review of Chinese investments in semiconductor, AI, and quantum computing sectors while creating fast-track processes for investments from allied nations.

AI National Security Review in Practice

The Trump administration has already demonstrated willingness to block AI-adjacent transactions on national security grounds regardless of transaction size. In 2024, President Trump issued an executive order blocking the $2.9 million sale of EMCORE Corporation’s indium phosphide chip assets to HieFo Corporation — a U.S. company controlled by a Chinese citizen — on CFIUS national security grounds.38 The transaction’s value was commercially insignificant; its strategic significance was not. AI M&A teams must now conduct CFIUS analysis early in deal processes, treat any founder or investor with Chinese nationality in a controlling position as a potential trigger, and be prepared for reviews extending 120 to 180 days in complex technology deals.

Valuation Changes

Compute Nationalism is repricing AI assets in ways that standard software valuation frameworks do not capture. Companies with privileged access to advanced AI compute — through strategic hyperscaler partnerships, long-term GPU purchase commitments, or sovereign infrastructure relationships — command premiums that reflect infrastructure scarcity rather than just model performance. Conversely, AI companies with significant China revenue exposure, Chinese investor representation, or supply chain dependencies on restricted components face haircuts that reflect geopolitical risk. M&A professionals who apply pre-2022 software valuation frameworks to AI infrastructure deals risk systematically mispricing both opportunities and risks.

For Investors: Pricing Infrastructure Fundamentals

Compute Ownership

Investors have frequently approached AI through software economics: network effects, marginal cost compression, user growth, platform adoption. Compute Nationalism suggests that infrastructure variables may prove more durably determinative than model benchmarks. The companies that control GPU clusters, datacenter capacity, long-term chip purchase agreements, and sovereign compute partnerships are increasingly the companies that control the economic structure of the AI industry. Infrastructure ownership — or privileged access — acquires disproportionate economic value when infrastructure is scarce and politically filtered.

Energy Access

Power infrastructure is increasingly an AI investment variable that would have seemed unusual a decade ago. AI reconnects digital valuation with industrial fundamentals. Location matters. Climate matters. Regulatory friendliness toward datacenter construction matters. Grid relationships matter. The ability to secure 100-megawatt power commitments from utility companies, navigate permitting environments, and access reliable generation capacity is now a meaningful competitive differentiator in AI infrastructure. Investors who ignore energy variables in AI infrastructure analysis are leaving risk factors unpriced.

Sovereign Partnerships

Sovereign capital behavior deserves monitoring as geopolitical telemetry. When the UAE’s G42 secures NVIDIA Blackwell chips that remain restricted to most of the world, or when Saudi Arabia’s HUMAIN announces a $500 megawatt AI factory partnership with NVIDIA, these are not merely commercial transactions — they are strategic signals about which nations Washington is partnering with in the construction of the global compute order. Investors who read sovereign AI investments as signals of geopolitical alignment, and who position accordingly in AI infrastructure assets that benefit from those alignments, are engaging in the kind of macro-level compute nationalism analysis that this era demands.

For Governments: AI Policy Is Now Infrastructure Policy

Industrial Policy

For governments, the implication is most fundamental. Much AI policy discourse remains focused on governance: algorithmic transparency, bias mitigation, privacy, platform accountability, election integrity. These matters are real and important. But Compute Nationalism identifies a prior question that governance frameworks cannot address: Does the nation control enough compute to remain strategically relevant? A government that produces sophisticated AI ethics frameworks while lacking domestic compute capacity has optimized the wrong layer of the stack. The policy agenda that follows from Compute Nationalism requires industrial-scale intervention: grid modernization, nuclear and renewable energy expansion, datacenter permitting reform, semiconductor investment coordination, and public-private compute partnerships.

Compute Readiness

Compute readiness — the ability to train and deploy frontier AI models on sovereign or allied infrastructure — is becoming a prerequisite for meaningful participation in the AI era rather than a nice-to-have. CNAS has framed the choice clearly: “The United States faces a choice: leverage its current lead to promote U.S. AI infrastructure and applications globally, while preserving its edge at the frontier; or continue to primarily focus on protection, while other countries gradually narrow the gap.”39 The same choice confronts every nation. Governments that build compute readiness early will have meaningful strategic options. Those that recognize the importance late will find their AI sovereignty constrained by decisions made by others.

AI Infrastructure Security

Defense planners face the sharpest version of this challenge. How dependent are military AI ambitions on commercial hyperscalers? How resilient are compute supply chains under sanctions or conflict conditions? Can strategic workloads remain sovereign in the event of allied access disruption? What happens to military AI capability if export controls on critical chips tighten further? These questions are no longer hypothetical. The March 2026 Iranian drone strikes on AWS facilities in the UAE and Bahrain — the first time commercial hyperscale datacenters became explicit kinetic targets — demonstrated that AI infrastructure is no longer strategically distinct from other forms of critical national infrastructure. It must be planned, protected, and governed accordingly.

The Caution Against Over-Nationalization

Compute Nationalism explains state behavior. It does not automatically endorse every nationalist response. Excessive fragmentation carries genuine costs: duplicative infrastructure, reduced efficiency, innovation slowdown, alliance friction, capital misallocation, and the erosion of the global technology commons that has been the foundation of AI’s rapid advance. The correct equilibrium between sovereign resilience and global openness remains genuinely uncertain. Policymakers who resolve that tension prematurely in either direction risk either strategic vulnerability or innovation stagnation. The goal is not autarky. It is strategic resilience — the capacity to continue generating intelligence even under conditions of adversarial disruption.

Conclusion: The Beijing Summit and the Infrastructure War That Defines It

On May 13, 2026, Jensen Huang boards Air Force One in Alaska and joins President Trump’s delegation to Beijing — a late addition, summoned by a personal phone call from the president after media coverage of his initial absence prompted a rapid reversal. The sequence that produced that boarding call — uninvited on May 11, publicly absent on the morning of May 13, then airborne to Beijing by afternoon — is itself the most compressed and vivid illustration of the central paradox this paper has attempted to illuminate. The most important company in the global AI economy has a market share in China of precisely zero. Its CEO could not make the original delegation list for a summit whose agenda is dominated by the very technology his company controls. And yet the moment his absence became visible, it proved politically untenable — because everyone understood, at once, what it signaled. An extraordinary strategic outcome produced by policy choices whose internal logic is coherent on both sides, and whose aggregate effect has been to accelerate exactly the dynamics each side sought to prevent: Washington sought to constrain China’s access to frontier compute, and produced a Chinese compute sovereignty drive of unprecedented intensity; Beijing sought to resist dependency, and found itself building on infrastructure that is less capable and less interoperable than what it has excluded.

Graham Allison has argued that the defining feature of the U.S.–China relationship, “for as far ahead as I can see, will be a ruthless rivalry” — a competition in which each side’s strategic choices trigger adaptation in the other that neither side fully anticipated.40 The compute war is the most vivid current expression of that rivalry. It is structural, not incidental. And it will not be resolved by a single summit, a single trade agreement, or a single licensing decision about H200 chips. The summit may produce diplomatic accommodations and commercial announcements. It will not resolve the underlying structural competition over intelligence-producing infrastructure.

What this paper has argued is that understanding this rivalry requires a framework equal to its depth. ‘Chip war’ is too narrow. ‘Tech competition’ is too vague. Compute Nationalism — the doctrine under which states treat intelligence-producing infrastructure as sovereign strategic assets — provides the conceptual structure needed to explain not just the U.S.–China confrontation but the broader global convergence toward sovereign AI infrastructure that is reordering geopolitical relationships across the Gulf, South Asia, Europe, and beyond.

Why “Compute”

Because the contest is larger than software, and larger than chips. Because intelligence production at frontier scale depends on energy, silicon, cooling, packaging, networking, manufacturing ecosystems, and datacenter deployment — a full industrial stack that cannot be improvised under pressure. Because whoever controls that stack increasingly controls the capacity to generate strategic advantage, economic productivity, and military capability in the twenty-first century. ‘Compute’ captures the physical substrate of intelligence production.

Why “Nationalism”

Because states are behaving as sovereign actors rather than neutral market participants. Because access to the most critical AI infrastructure is becoming politically conditioned. Because infrastructure is being subsidized, protected, weaponized, and denied on strategic grounds. Because dependency is generating sovereignty responses across every major power. Because AI infrastructure is becoming national capability rather than global utility. ‘Nationalism’ captures the political behavior that scarcity and strategic importance inevitably produce.

Oil, Compute, and the Historical Analogy

Oil shaped twentieth-century geopolitics because industrial civilization depended on energy access. States secured supply, protected routes, built reserves, and — on occasion — fought over infrastructure. Artificial intelligence may generate analogous strategic dynamics — not because compute is identical to oil, but because strategic dependency on a scarce industrial input creates similar political behavior regardless of what that input is. Control matters. Access matters. Infrastructure matters. Scarcity matters. The analogy is imperfect but illuminating. It points toward a future in which computational capacity is treated with the same sovereign urgency that energy once commanded — and in which the geopolitics of compute shapes the global order as decisively as the geopolitics of oil shaped the last century.

Compute Nationalism began — as the Beijing summit of May 2026 makes unmistakably plain — not in some hypothetical future, but in the world we are already living in. Understanding it is not academic. It is the first step toward navigating it.

Footnotes and Sources

1.  The Huang-Beijing story unfolded in three distinct reports across two days. Act One — May 11, 2026: Reuters reported Huang was not going to Beijing and had not been invited, with the White House focused on agriculture and aviation matters. Reuters — Nvidia CEO Huang Not Going to China During Trump Visit; confirmed by Tom’s Hardware — Jensen Huang Snubbed by White House for Trump’s China State Visit; and CNBC — Nvidia CEO Jensen Huang Isn’t Part of Trump’s China Trip. Act Two — afternoon of May 13, 2026: Reuters reported Huang would join after all, per two sources familiar with the matter. Reuters via Yahoo Finance — Nvidia’s Jensen Huang to Join Trump’s China Visit, Sources Say; also Investing.com — Nvidia’s Jensen Huang to Join Trump’s China Visit. Act Three — CNBC confirmed: Trump called Huang personally after press coverage of his absence; Huang flew to Alaska to board Air Force One. NVIDIA statement: ‘Jensen is attending the summit at the invitation of President Trump to support America and the administration’s goals.’ CNBC — Nvidia Says CEO Jensen Huang Is Joining Trump’s China Trip. Huang’s earlier quote to Jim Cramer that it would be ‘a great honor to represent the United States’ if invited: CNBC — Trump Invites Musk, Cook, Fink to China Trip

2.  Reuters, May 13, 2026, 1:39 AM UTC, byline Laurie Chen and Karen Freifeld: Reuters via Yahoo Finance — Nvidia’s Jensen Huang to Join Trump’s China Visit, Sources Say. ‘Nvidia CEO Jensen Huang will join over a dozen U.S. CEOs on President Donald Trump’s visit to China this week, two sources familiar with the matter told Reuters on Wednesday. Huang did not appear on an initial list of executives provided by the White House earlier this week.’

3.  CNBC confirmed the mechanism of Huang’s addition: CNBC — Nvidia Says CEO Jensen Huang Is Joining Trump’s China Trip. ‘After seeing the media coverage of Huang’s absence from the delegation, Trump called the Nvidia executive and asked him to join, a source familiar with the situation told CNBC. Huang flew to Alaska to board Air Force One, the source said.’ NVIDIA statement: ‘Jensen is attending the summit at the invitation of President Trump to support America and the administration’s goals.’

4.  Chris Miller, Chip War: The Fight for the World’s Most Critical Technology (Scribner, 2022). Miller is Associate Professor of International History at the Fletcher School, Tufts University. Quoted passage appears in the Introduction; confirmed in: Global Policy Journal — Book Review: Chip War. Miller: ‘Semiconductors have defined the world we live in, determining the shape of international politics, the structure of the world economy, and the balance of military power.’

5.  Chris Miller, interview with McKinsey, October 2024: McKinsey — Author Chris Miller on the Global Influence of Semiconductors.

6.  Jensen Huang, Dwarkesh Patel Podcast, April 2026. Full transcript: Dwarkesh.com — Jensen Huang: TPU Competition, Why We Should Sell Chips to China. Direct quotation: ‘AI is a five-layer cake. The AI industry matters across every single layer.’

7.  Graham Allison, Harvard Kennedy School. Quotation from dialogue with Center for China and Globalization, 2021: China-US Focus — Can China and US Escape Thucydides Trap?; and Harvard Kennedy School Thucydides Trap resource: HKS — Thucydides Trap. See also Allison’s Destined for War: Can America and China Escape Thucydides’s Trap? (Houghton Mifflin Harcourt, 2017).

8.  Jensen Huang, NVIDIA GTC 2026 Keynote Address, March 2026. Analyzed in: Colorado AI News — Quote of Note: Jensen Huang; and YourStory — Jensen Huang’s 5-Layer AI Stack

9.  Center for a New American Security (CNAS), Global Compute and National Security, August 2025. CNAS — Global Compute and National Security

10.  U.S. Congress, AI Diffusion Rule, January 15, 2025. Congressional Research Service analysis: Congress.gov CRS — U.S. Export Controls and China: Advanced Semiconductors. Three-tier framework: Tier 1 (18 allied nations), Tier 2 (most of world, quantity caps), Tier 3 (~20 adversarial countries including China, Russia — full prohibition). Blackwell B200 categorically denied to Tier 3.

11.  NVIDIA Corporation, Form 8-K, Q1 FY2026, filed with the U.S. Securities and Exchange Commission: SEC — NVIDIA Q1 FY2026 Earnings Release. Filing: ‘On April 9, 2025, NVIDIA was informed by the U.S. government that a license is required for exports of its H20 products into the China market. NVIDIA incurred a $4.5 billion charge in the first quarter of fiscal 2026.’

12.  U.S. Department of Commerce, Bureau of Industry and Security, Final Rule, effective January 15, 2026: BIS — Revised Semiconductor License Review Policy for China. Legal analysis: Morgan Lewis — BIS Revises Export Review Policy for Advanced AI Chips; Mayer Brown — Administration Policies on Advanced AI Chips Codified

13.  NVIDIA CFO Colette Kress, Q1 FY2026 earnings call, February 25, 2026. Cited in: 247 Wall Street — Nvidia: Jensen Huang Says We Should Be Selling Chips to China

14.  Jensen Huang, interview with the Special Competitive Studies Project (SCSP), May 2026. Reported in: Tom’s Hardware — Jensen Says Nvidia Now Has ‘Zero Percent’ Market Share in China. Full quotation: ‘Conceding an entire market the size of China probably does not make a lot of strategic sense, so I think that has already largely backfired. Maybe it made sense at the time, but I think the policy really needs to be dynamic and needs to stay with the times.’

15.  Jensen Huang, Dwarkesh Patel Podcast, April 2026. See footnote 4. Full quotation: ‘Why are you causing one layer of the AI industry to lose an entire market so that you could benefit from another layer of the AI industry? There are five layers, and every single layer has to succeed.’

16.  Jensen Huang, Dwarkesh Patel Podcast, April 2026. See footnote 4. Full quotation: ‘We want to make sure that all the AI developers in the world are developing on the American tech stack… It would be extremely foolish to create two ecosystems.’ Also reported: Tom’s Hardware — Jensen Huang ‘Nearly Lost His Composure’

17.  ASML EUV ban from China since 2019; DUV restrictions effective September 1, 2023. CNBC coverage: CNBC — Netherlands Follows U.S. with Semiconductor Export Restrictions. ASML 2025 China revenue outlook — 48% year-over-year decline: CNBC — ASML 2025 Outlook Shows U.S. Chip Export Curbs Impacting China Sales. September 2024 Dutch expansion of restrictions: Yahoo Finance — China Hit Hard by New Dutch Export Controls on ASML

18.  CNAS, December 2025: CNAS — The Export Control Loophole Fueling China’s Chip Production. ‘ASML actively services machines in China, despite U.S. pressure on the Dutch government to prevent this, potentially extending machine lifespans to 30 years.’

19.  Remote Access Security Act, passed House 369-22, January 12, 2026. Analysis: Introl — Remote Access Security Act: Cloud Loophole Export Controls 2026. INF Tech rented 2,300 Blackwell GPUs through Indonesian data center; Tencent secured $1.2 billion in contracts for 15,000 Blackwell processors via Japanese provider Datasection.

20.  Chip smuggling estimates: Introl analysis, February 2026: Introl — BIS H200 Export Policy China: Case-by-Case Review 2026. Operation Gatekeeper, unsealed December 8, 2025, revealed NVIDIA H100 and H200 chips relabeled as ‘Sandkyan’ brand to evade export controls.

21.  Jensen Huang, remarks to reporters at APEC CEO Summit, Gyeongju, South Korea, October 31, 2025: CNBC — Nvidia’s Huang Doesn’t Buy the National Security Concerns Over Selling Chips to China. Full quotation: ‘It’s in the best interest of America to serve that China market. It’s in the best interest of China to have the American technology.’

22.  China Ministry of Commerce, Notice 2024 No. 46, December 3, 2024. Translation: CSET Georgetown — China Rare Earth Export Ban. Context: Optilogic — How China’s Rare Earth Metals Export Ban Will Impact Supply Chains in 2025. Ban issued December 3, 2024 — one day after U.S. announced new export controls on 140 Chinese entities.

23.  Stimson Center analysis, April 2025: Stimson Center — China’s Germanium and Gallium Export Restrictions: Consequences for the United States. Suspension of ban announced November 2025: CNBC — China Suspends Ban on Exports of Gallium, Germanium, Antimony. Suspension runs until November 27, 2026.

24.  Huawei Ascend 910C specifications and yield analysis: Bitrue — Huawei Ascend AI Chip Specs 2025; Mizuho yield estimate (~30%) reported in: Wccftech — Mizuho: Huawei Will Likely Sell Over 700,000 Units of Ascend 910 Series Chips in 2025

25.  Huawei 2026 Ascend production targets: Bloomberg — Huawei to Double Output of Its Advanced AI Chip Ascend. 600,000 units of 910C in 2026; up to 1.6 million dies total Ascend line. Also: RCR Wireless — Huawei to Double Output of Ascend AI Chips

26.  Morgan Stanley SMIC capacity analysis: Huawei Central — SMIC Expected to Produce 30% Huawei Ascend 910B AI Chips by End of 2025. SMIC target: 7,000 wafers per month 2025, scaling to 18,000 by 2027. SemiAnalysis deeper analysis: SemiAnalysis — Huawei Ascend Production Ramp

27.  China approved 400,000 H200 units for ByteDance, Alibaba, and Tencent. Built In — Trump Lifted the AI Chip Ban on China; Network World: Network World — Nvidia H200 Chips in China: US Says Yes, China Says No. Chinese customs blocked imports despite U.S. regulatory approval, January 2026.

28.  ByteDance-led 2026 H200 orders exceeding $14 billion; Chinese firms ordered over 2 million H200s post-December 2025 announcement: Introl — BIS H200 China Export Policy / AI OVERWATCH Act 2026

29.  U.S. Outbound Investment Security Program, effective January 2, 2025. Treasury Final Rule, October 28, 2024. Analysis: Skadden — US Treasury Creates the ‘Reverse CFIUS’ Program; Cleary Gottlieb — Long-Awaited U.S. Outbound Investment Regime Published

30.  Jensen Huang, SCSP interview, May 2026. See footnote 12.

31.  Huawei projected AI chip revenue of $12 billion and 60% China market share in 2026: MSN — Nvidia’s China AI Chip Market Share Falls to Zero

32.  CHIPS Act awards as of November 2025 — $36 billion across 40 projects, 19 companies: Manufacturing Dive — Tracking CHIPS and Science Act Awards; Part Locator — CHIPS Act 2025 Timeline

33.  TSMC Arizona 4nm production commenced early 2025; 4% higher yields than Taiwan announced October 2024. March 2025 $100 billion expansion announced: Wikipedia — CHIPS and Science Act; TSMC — Arizona Third Fab Announcement, April 8, 2024

34.  GAO, Semiconductors: Information on Projects Funded to Strengthen U.S. Supply Chain, GAO-26-107882, December 2025: GAO-26-107882. ‘Commerce estimates that these projects will bring the U.S. share of global leading-edge logic chip manufacturing from 0 percent in 2022 to 20 percent by 2030.’

35.  UAE-U.S. AI Campus announced May 2025 by Trump and UAE President Mohamed bin Zayed Al Nahyan: Introl — The Middle East’s Trillion-Dollar Bet on AI Infrastructure; SemiAnalysis — AI Arrives in the Middle East: US Strikes a Deal with UAE and KSA

36.  NVIDIA-HUMAIN partnership, Saudi Arabia, May 2025. 18,000-unit GB300 Blackwell supercomputer, 500 MW target over five years: Scoop Empire — Nvidia Deal Ignites the Gulf’s AI Ambitions. AWS Saudi Arabia $5.3 billion region investment; Google Cloud-PIF $10 billion partnership: World Economic Forum — It’s Time to Start Treating AI Infrastructure as Critical Infrastructure

37.  White House Critical and Emerging Technologies list update, February 2024, explicitly including advanced AI systems. CFIUS analysis: GovFacts — How CFIUS Decides Whether AI Investments Threaten National Security. Trump ‘America First Investment Policy,’ February 21, 2025: K&L Gates — Trump Administration Directs CFIUS to Tighten Restrictions

38.  Trump blocks HieFo Corporation acquisition of EMCORE chip assets, 2024. Analysis: Fenwick — Trump Blocks $2.9M Chip Deal Over National Security Concerns

39.  CNAS, Global Compute and National Security, August 2025. See footnote 7.

40.  Graham Allison, Center for China and Globalization dialogue, April 2021. See footnote 5. Full context available at: Springer — Escaping Thucydides’s Trap: Dialogue with Graham Allison