NARRATIVE: $295 billion for AI — China goes all in, America watches

Introduction: a plan that changes the rules of the global technological game

June 9, 2026: Beijing codes its ambition into law

On June 9, 2026, China's National Development and Reform Commission finalised the largest artificial intelligence infrastructure plan ever conceived by a single state. Its estimated cost: $295 billion — or 2,000 billion yuan — spread over five years. Its central constraint: at least 80% of underlying technologies, including AI chips, must be of domestic Chinese origin. This is not merely an economic plan. It is a declaration of technological war on the West — a decision by Beijing to permanently free itself from dependence on American semiconductors and build sovereign technological dominance.

To grasp the scale of the thing, some comparisons are in order. The $295 billion plan represents roughly the annual GDP of Finland. It exceeds the total NASA budget for the next fifteen years. And it will be deployed in a country of 1.4 billion people, with an industrial execution capacity that has no global equivalent in the construction of physical infrastructure. The electrical transformers, the data centres, the cooling systems — everything that constitutes the material substrate of AI — will be built in China, by Chinese companies, with predominantly Chinese materials. This narrative began in 2025. In 2026, it accelerates.

Bessent sounds the alarm at the Economic Club of New York

On June 29, 2026, US Treasury Secretary Scott Bessent delivered a speech at the Economic Club of New York that should haunt the boardrooms of major American technology companies. "The greatest risk to AI is China getting ahead." This sentence — direct, unambiguous, coming from a Treasury Secretary rather than a Pentagon official — says something important about growing awareness in Washington of what is at stake. Bessent added: "The reason China is willing to have discussions with us on AI is because we are ahead. We need to stay there."

This statement was made in a context of technological competition intensifying at every level. The United States has maintained strict export controls on the most advanced AI chips to China — with a notable exception for Nvidia's H200 chips, which Beijing has actively discouraged its companies from adopting. China, for its part, has built its own alternatives: Huawei with its Ascend chips, Cambricon, Biren Technology — less powerful than the best American chips, but catching up rapidly. The $295 billion plan is precisely designed to accelerate this catch-up and turn the current American advantage into a future structural vulnerability.

The plan in detail: infrastructure, chips, zero dependency

2,000 billion yuan: what it buys

The Chinese $295 billion plan goes far beyond funding data centres. It covers the entire AI supply chain: semiconductor chip production, energy infrastructure to power computing centres, cooling systems, talent training, and the development of language and vision models. The 80% domestic content constraint obliges every link in this chain to find Chinese solutions — creating an artificial but massive demand for Chinese semiconductor manufacturers, energy providers, and engineers trained in Chinese universities.

Bank of America anticipates that China's AI capex spending will triple by 2030, according to a report published on June 30, 2026. Globally, the firm estimates that AI capital investment will quintuple by the same date — a third of this spending now going toward non-computing infrastructure: energy, cooling, building. Electrical transformers already have three-year delivery lead times. The race for computing power is also a race for electricity. And in this domain, China — with its massive investments in nuclear, hydroelectric, and renewable energy — starts with a structural advantage.

80% domestic: the accelerated decoupling strategy

The 80% domestic technology content requirement is the plan's most radical provision. It obliges Chinese companies that want access to public funding tied to the plan to use Chinese chips, Chinese software, and Chinese networks. This provision creates enormous pressure on semiconductor manufacturers like SMIC and chip designers like Huawei HiSilicon to close the technological gap with TSMC and Korean foundries. It also creates a captive market for Chinese AI — progressively eliminating American alternatives from the decision-making circuits of companies that want to remain in the state's good graces.

This technological decoupling strategy is symmetrical to that of the United States, which used export controls to cut Beijing off from the most advanced chips. China responds by building its own supply chain, with the advantage of a domestic market of 1.4 billion consumers and companies that can absorb initially less powerful products. It is a race toward self-sufficiency — and China has the financial, human, and institutional resources to compete seriously.

Military AI: from absent transparency to documented threat

Autonomous drones, joint operations, Taiwan simulation

Defence analyst Rick Fisher, in an opinion published by The Epoch Times on June 27, 2026, bluntly exposes the military dimension of the Chinese AI plan: "China is rapidly exploiting AI to enhance future joint operations based on unmanned systems — in the aerial, space, naval, and land domains. It is close enough to this objective that it likely figures prominently in the early phases of a Taiwan invasion." This direct formulation — the invasion of Taiwan as a planning scenario, not a fringe hypothesis — reflects the evolution of American national security analysis.

The $295 billion plan is not presented by Beijing as a military plan. It is presented as an economic investment in the country's digital infrastructure. But the boundary between civilian AI and military AI is in China — by design — intentionally blurred. The military-civil fusion policy (军民融合) requires private Chinese companies to share their technologies with the military on demand. The $295 billion invested in civilian AI infrastructure is therefore, de facto, an investment in the military capability of the People's Liberation Army. This link is documented, institutionalised, and publicly denied by Beijing.

The Pentagon also acts: the 188 entities of the 1260H list

The American response to this reality is visible in another June 2026 decision: the addition of 65 new entities to the Pentagon's 1260H list, bringing the total to 188 Chinese military companies. Alibaba, Baidu, BYD, NIO, Unitree Robotics, WuXi AppTec: major civilian technology companies designated as contributing to the Chinese defence industrial base. This designation is not symbolic: from June 30, 2026, the Department of Defense is prohibited from contracting with these companies. And in 2027, restrictions will extend to their subcontractors.

These designations concretely illustrate why the Chinese AI plan worries Washington: the companies on which Beijing is counting to build its AI infrastructure are precisely those the Pentagon considers military actors. Alibaba Cloud, Baidu AI Cloud, BYD with its battery and autonomous vehicle technologies — all these entities contribute to a technological ecosystem with real military applications. And the $295 billion plan will accelerate their development.

The American response: export controls, a lead strategy, and its limits

Export controls: a leaking dam

American policy in the face of the Chinese AI plan rests primarily on export controls on advanced chips. The Trump administration maintained and extended controls introduced under Biden, prohibiting the export to China of the most powerful AI chips — notably Nvidia's H100 and H200 — with occasional exceptions. Bessent warned that any attempt to "weaponise supply chains, steal technology, circumvent sanctions, manipulate markets or coerce our partners will not go unanswered."

These controls have a real effect — they have delayed the Chinese AI programme and forced massive investments in domestic alternatives. But they also have documented limits. China accesses chips via third countries. It is developing its own alternatives, less powerful but improving rapidly. And Beijing has actively discouraged its companies from adopting the H200 chips that Washington had exceptionally authorised — preferring to develop its own solutions and reduce any residual dependency. The dam holds, but it leaks. And the $295 billion plan is designed to make it irrelevant in five years.

Quantum, critical minerals, shipbuilding: the other fronts

Bessent identified in his speech several areas where the United States needed to maintain its leadership: AI, quantum computing, shipbuilding, critical minerals. The last point is particularly crucial: China controls a majority share of global production of rare earth elements and other minerals essential for manufacturing chips, batteries, and weapons systems. Without access to these minerals, the American technological supply chain is vulnerable — regardless of the level of investment in semiconductor factories. This is why the United States and its allies have launched initiatives to diversify critical mineral supply sources — with still modest results.

Shipbuilding illustrates another aspect of the problem. The American navy needs ships — American shipyards have insufficient capacity. China today builds more ships than the entire rest of the world combined. This industrial capacity mirrors its ability to deploy its AI plan: it can build the physical infrastructure of technological dominance at a speed and scale that Western democracies, with their more complex planning processes, cannot easily match.

Implications for Ukraine and the world order

China, Russia, and the technological war in Ukraine

The Chinese AI plan is not directly linked to the war in Ukraine — but the connections are real. Beijing remains Russia's principal diplomatic and economic support in the conflict. It buys its oil, supplies it with dual-use technologies, and keeps it from complete diplomatic isolation. If China succeeds in building an autonomous military AI infrastructure over the next five years, the implications for aid to Russia — notably for drone systems, military communications, and electronic surveillance — will be direct. A technologically self-sufficient China is a China that can support its strategic allies without fearing American sanctions on technologies.

The strategic China-Russia-Iran triangle would directly benefit from Chinese technological superiority in AI. Iranian drones striking Ukraine, North Korean missiles that Russia uses on Ukrainian cities — these systems are the sketch of what is coming if China can integrate its AI capabilities into the military doctrines of its partners. That is why the $295 billion plan is not only an American issue. It is an issue for every democracy seeking to resist the advance of authoritarian regimes in the coming decades.

What Europe should understand — and has not yet grasped

Europe is watching this technological race with a certain detachment. It is investing in AI — the AI Act, Horizon Europe initiatives, national projects in France, Germany, Sweden. But its investment level is incomparable to the Chinese $295 billion. And its awareness of the geopolitical implications of this race remains insufficient. If China wins the military AI race, the consequences will not be limited to the Pacific: they will directly affect the West's capacity to support Ukraine, to resist Russian pressure, and to defend democratic values against authoritarian regimes.

The link between technology, democracy, and security is not abstract. It materialises in surveillance drones, in electronic warfare systems, in autonomous weapons, in algorithmic disinformation. All these technologies are being developed in China, financed in part by the $295 billion plan, and intended to reinforce not only Beijing's military power but also that of its authoritarian partners. Europe cannot remain a spectator of this race. It does not have the luxury of indifference.

Chinese companies under pressure: between state plan and market realities

Alibaba, Baidu, Huawei: pawns in a state game

Alibaba filed a lawsuit against the Pentagon on June 23, 2026 to be removed from the 1260H list, arguing that its designation "has no basis in fact or law." Baidu, BYD, and NIO have similar reactions — these companies do not want to be perceived as military entities, because this perception harms their international business. But in the economic and institutional reality of Xi Jinping's China, the line between private company and state actor exists only nominally. The military-civil fusion policy demands cooperation, not consent.

These companies find themselves in an uncomfortable position: they must navigate between the requirements of the Chinese state — participating in the AI plan, sharing data with the military, respecting technological self-sufficiency targets — and international market constraints — avoiding Western designations, maintaining partnerships with American and European companies. This tension is irresolvable in a world of accelerated technological decoupling. And the $295 billion plan will accentuate this tension: it pushes Chinese companies even deeper into the orbit of the state, making their formal independence even more fictional.

The lobbyists abandon ship

On June 30, 2026, lobbying firm BGR became the latest major Washington firm to cut its ties with Alibaba, ending a contract worth $140,000. In the preceding two weeks, M/O Strategies, Story Partners, Mercury Public Affairs, Sidley Austin, Greenberg Traurig, and Brownstein Hyatt Farber Schreck had done the same. The same firms had also dropped Tencent. This flight of lobbyists says something eloquent about the perception of the future: major Washington firms consider that the legal and reputational risks associated with representing entities on the 1260H list outweigh the revenues these contracts generate. Decoupling is no longer only economic and technological — it is institutional.

This lobbyist flight is also a signal to Chinese companies themselves: the American influence system is progressively closing to their interests. The channels through which they could advocate their case in Washington — lobbying firms, professional associations, academic partnerships — are shrinking. And this narrowing of influence channels reinforces the logic of complete decoupling. If you cannot influence American policy, the most rational strategy is to no longer depend on it. That is exactly what the $295 billion plan seeks to achieve.

The talent battle: China's human advantage in the AI race

Millions of engineers trained for a single objective

China produces approximately 1.5 million graduates in science, technology, engineering, and mathematics every year — more than triple the American output. This talent pool is not an abstract figure: it is the human fuel that makes the deployment of the $295 billion plan possible. Training teams of engineers capable of designing AI chips, building data centres at industrial scale, and training competitive language models takes years. China began building this human capital a decade ago — and the results are now visible in the quality of models like DeepSeek, Alibaba's Qwen, and Baidu's Ernie.

In the United States, a significant portion of AI talent comes from immigration — notably from China and India. The debate over H-1B visas, growing restrictions on foreign students in sensitive research programmes, and political pressure to reduce skilled immigration create a difficult tension: the United States is limiting access to foreign talent for security reasons, but risks depriving itself of a crucial competitive advantage. China, for its part, does not have this problem: its talent stays home, attracted by competitive salaries and a sense of national mission in the AI race.

The brain return: diaspora and national ambition

China's Thousand Talents Program and its successors have drawn researchers trained in the best American and European universities back to build Beijing's AI ecosystem. This brain repatriation strategy has raised legitimate concerns in Washington — several technology espionage cases have resulted — but its primary effect is more mundane and more powerful: it created a bridge between the world's best scientific practices and China's national programme. Researchers who worked in the laboratories of Google, Microsoft Research, or MIT before returning to China brought with them invaluable knowledge of the methods, organisational cultures, and research directions that underpin current American superiority in AI.

This transfer of knowledge — legal and illegal, deliberate and involuntary — is one of the least discussed but most significant factors in Sino-American technological competition. The $295 billion plan creates new incentives for the best Chinese talent to remain, or return, to build in China. And in a field where individual talent has a multiplier effect — one exceptional researcher can advance an entire laboratory — this human dynamic could prove as decisive as infrastructure investments.

The energy challenge: who will power the race to billions of calculations

Nuclear, hydroelectric, solar: China is also building the grid that powers AI

Modern artificial intelligence consumes quantities of electricity that defy imagination. Training a large language model like GPT-4 consumes as much electricity as several thousand households use in a year. Multiply this by the ambition of the $295 billion plan — dozens of new data centres across China, running continuously — and you obtain a colossal energy demand. China responds to this with an equally colossal strategy: it is today the world's leading investor in nuclear, solar, and wind energy, and its hydroelectric projects remain the largest in the world.

Electrical transformers — the essential equipment that distributes power to data centres — have three-year delivery lead times worldwide, as Bank of America highlighted in its June 2026 report. But China is also the world's leading manufacturer of this equipment. It can prioritise its own infrastructure — and it does. While the United States and Europe wait for their transformers in a global queue, China builds its own and installs them with priority for its own data centres. This vertical mastery of the energy supply chain is a strategic advantage that Western democracies cannot quickly replicate.

Energy sovereignty as a geopolitical weapon

Dependence on energy resources is a lesson Europe learned painfully with Russia and natural gas. China has drawn the opposite lesson: it is investing massively to ensure that its AI race will never be limited by access to energy. This strategy is consistent with the 80% domestic content objective: create an AI economy that functions independently of any foreign decision. If the United States and its allies impose technology sanctions, Beijing wants to be in the position of absorbing them without major disruption — because energy, talent, chips, and software will all be produced domestically or by partners that escape Western control.

This convergence — massive AI plan, growing energy independence, abundant domestic talent, substitute chips in rapid development — is not a coincidence. It is a coherent, deliberate strategy for transforming China into a sovereign technological economy. And the window for the West to respond effectively to this strategy narrows with every billion yuan spent. The $295 billion plan is being executed. The question is no longer whether China is trying — the question is whether the West responds with the same determination.

Conclusion: a race whose outcome will define the century

What is really at stake behind the $295 billion

The AI race between the United States and China is not a simple economic rivalry between two powers. It is a conflict to determine who will be in a position to dictate the rules of the technological, economic, and military game of the second half of the 21st century. If China succeeds in its $295 billion plan and achieves its 80% technological self-sufficiency objective, it will have built an AI industrial base that Western sanctions can no longer reach. This is not an alarmist projection — it is the explicitly stated objective in Chinese planning documents.

Bessent is right: the race must be won. But winning the race is not limited to maintaining export controls. It requires massive investment in American and allied AI infrastructure, coordinated industrial policy, and a shared understanding among democracies that this race is a collective project, not only an American one. Europe, India, Japan, South Korea, Australia — all technological democracies have a direct interest in the West staying ahead. And all must invest accordingly.

The world we need to avoid

The scenario to avoid is not Chinese superiority in commercial AI — it is superiority in military AI. A world where China has superior autonomous drones, more effective electronic warfare systems, and invulnerable cyberdefence capabilities is a world where conventional deterrence no longer works as it does today. That is the world Beijing is seeking to build — for Taiwan first, but also for the entire global strategic balance. And the $295 billion of the Chinese AI plan is the road map to that world.

This narrative does not end well if democracies continue to treat the AI race as an economic question rather than a security one. Bessent told the truth at the Economic Club of New York. It remains to be seen whether democracies — American, European, and beyond — hear the message with the same urgency as the one who delivered it.

By Maxime Marquette, columnist

Columnist's transparency note

Bias and positioning

I consider China's growing technological power to be a real threat to the global democratic order, and military AI to be its most concerning vector in the long term. This bias is declared. In this narrative, I sought to document the facts with precision while clearly expressing my reading of their significance. All factual claims — figures from the Chinese plan, Bessent's statements, Pentagon designations, lobbyist departures — are grounded in public sources cited in the Sources section.

What I do not know

The actual execution of the $295 billion plan remains uncertain. Chinese economic plans sometimes show gaps between official ambition and execution reality. The figures for actual spending, the technical progress of Chinese chip manufacturers, and the real military effectiveness of developed AI systems are not fully known to the public. I make no claim to certainty on these points.

Sources

Primary sources

The Star Malaysia — America's biggest risk on AI is China getting ahead, Bessent says — June 29, 2026

CNBC — China's AI capex to triple to 2030 amid aggressive power infrastructure spending: BofA — June 30, 2026

The Epoch Times — In AI Race With China, America Must Put Dominance Before Cooperation — June 27, 2026

Secondary sources

19FortyFive — Defence and geopolitics analysis (China AI threat, June 2026)

Global Data Center Hub — China funded the AI grid: the chips — June 2026

Foreign Policy — China technology race analysis (July 2026)

Axios — China AI infrastructure coverage (June 2026)