FACT-CHECK: China's $295 billion vs America's $725 billion in AI — the real numbers behind the race

The formula making the rounds in tech media

Since the start of 2026, a formula has been circulating in specialized technology media, geopolitical circles, and government publications: China is investing $295 billion in artificial intelligence and data infrastructure, while American tech giants are mobilizing more than $725 billion. These two figures are often placed side by side as if they were comparable — two countries, two investments, an AI race being run on roughly equal terms. That reading is inaccurate. And that inaccuracy has real consequences for how we understand global technological competition.

This fact-check aims to clarify these numbers: where do they come from, what exactly do they measure, over what time frame, by which actors, and what they actually reveal about the gap between the two powers in the race for artificial intelligence. The truth is more nuanced and, in some respects, more concerning than the surface comparison suggests — not for the reasons that proponents of the "China is catching up with America" narrative advance, but because the actual figures reveal structural dynamics that headlines fail to capture.

Why this fact-check is necessary now

The comparison of AI investments between China and the United States is not merely a statistical question. It informs government policy decisions, technology company strategies, military budget allocations, and trade negotiations between major powers. Framing errors in this comparison can lead to suboptimal policies — either underestimating Chinese capabilities, or overestimating them and generating a disproportionate panic response. Both types of error carry real costs. Hence the importance of a precise factual look at what these numbers actually say.

The sources used in this fact-check are the ones that first reported or analyzed these investments: Bloomberg on the Chinese plan, the annual reports of American tech companies, and analyses from specialized firms like TechStrong AI and Techforward. Every verified claim will be clearly attributed to its source. Uncertainties and gaps will be explicitly flagged.

Numbers in technology have a structural problem: they are used to impress as much as to inform. When someone says "China is investing $295 billion," there is a rhetorical intent behind the figure — sometimes to warn of a threat, sometimes to reassure about China's lag, sometimes to justify American spending. This fact-check has a single objective: to understand what these numbers actually measure. The rest — the political conclusion drawn from them — belongs to the reader.

The origin of the figure: Bloomberg, June 2026

The figure of $295 billion for Chinese AI investment comes from a Bloomberg report published on June 9, 2026. According to this source, the Chinese government — via the National Development and Reform Commission (NDRC) — is preparing a national financing plan for the construction of a network of artificial intelligence infrastructure and data centers distributed across the entire territory. The total amount would be 2 trillion yuan, approximately $295 billion at the June 2026 exchange rate.

Three essential points to retain about this figure. First: it is a total over five years (2026–2030), not a single year. Second: the plan is still in draft form — the NDRC is "preparing" this plan; it had not been officially adopted at the time of publication. Third: the funding would come primarily from the issuance of Chinese sovereign bonds — a public debt mechanism, not private investment or operating revenues. These three clarifications radically change the nature of the figure.

The operators: China Mobile and China Telecom in the lead

According to available information, the operators designated to deploy this infrastructure network would primarily be the two state telecommunications giants: China Mobile and China Telecom. These companies would be charged with building and operating the data centers, the associated energy distribution networks, and the high-speed connections between network nodes. The stated goal: have the national AI infrastructure fully operational by 2028.

The distinctive feature of this model — state operators deploying infrastructure financed by sovereign bonds — reflects China's economic structure and also its constraints. China's private technology companies (such as Alibaba, Tencent, ByteDance, Baidu) are not the primary actors in this state plan. They have their own investments — estimated at an additional $200 billion in 2026 — that are added on top of the government plan but are not included in the $295 billion. This distinction matters for the comparison with the United States.

China's $295 billion plan is state infrastructure over five years. It is significant. It is ambitious. It is not the same thing as $725 billion spent by private companies in a single year because their market demands it and their revenues allow it. Comparing these two figures without specifying these structural differences is like comparing the budget of a public airport built over ten years with the annual revenue of all American airlines combined. Both are large numbers. They do not measure the same thing.

Who is spending what: the four giants and their numbers

The figure of $725 billion for American AI and infrastructure investments in 2026 comes from a compilation of capital investment budgets announced by the four largest American technology groups. Amazon (AWS): $200 billion in capex investments announced for 2026, the vast majority in data centers and cloud infrastructure. Microsoft (Azure + OpenAI): $190 billion. Alphabet (Google Cloud + DeepMind): between $175 and $185 billion. Meta (AI infrastructure): between $115 and $145 billion.

Total: between $680 and $720 billion, rounded to $725 billion in certain analyses. These figures represent a 77% increase compared to 2025, when the same four groups invested approximately $410 billion. This explosion in capex spending in a single year reflects the race to next-generation AI models, competition for computing capacity, and growing enterprise customer demand for cloud-AI services.

What this figure includes and what it excludes

The figure of $725 billion represents the capital expenditure (capex) of the four largest American technology players across their entire infrastructure — data centers, servers, undersea cables, energy distribution networks, and real estate. It is not strictly limited to artificial intelligence — it covers the full cloud infrastructure of which AI is a central but not exclusive component. Some of this spending goes to standard storage, web services, and enterprise productivity tools that do not directly involve AI.

Another important clarification: these figures represent only a portion of the American technology ecosystem. Hundreds of other actors — NVIDIA, AMD, AI startups, independent data center operators, semiconductor companies — are investing massively in the same infrastructure. The full scope of American AI and associated infrastructure investment in 2026 far exceeds the $725 billion from the four Big Tech players if these additional actors are included. The figure is therefore actually an underestimate of the total American effort.

When Amazon alone spends $200 billion on infrastructure in a single year — that is two-thirds of the entire Chinese five-year plan — you understand what the dominance of the American tech ecosystem actually means. That is not a reason to lower our guard against Chinese ambitions. But it is a reason not to fall into the trap of the catastrophist narrative presenting China as already catching up with America on this file. The numbers tell a different story.

The fundamental ratio: 1 American year vs 5 Chinese years

The most revealing comparison is the simplest. $295 billion: the Chinese plan over five years (2026–2030). $725 billion: the investment of the four American Big Tech companies in a single year (2026). In other words, the four largest American technology companies spend, in a single fiscal year, two and a half times more than what China plans to spend in five years in its national AI plan.

If you project American investments over the same five-year period — even maintaining zero growth, which is unrealistic given the current trajectory — the four American Big Tech companies alone would invest more than $3.6 trillion, more than twelve times the Chinese plan. That proportion is staggering. It does not mean China is not a serious technological power. It means the usual media comparison — presenting the two as in relative parity — is fundamentally inaccurate.

Quality vs quantity: the chip differential

But raw investment figures do not tell the whole story. There is a crucial qualitative dimension in this comparison: access to cutting-edge semiconductors. American data centers are built around NVIDIA H100 and H200 chips — the most advanced graphics processors available for AI applications. These chips are subject to strict export controls to China, initiated under the Biden administration and maintained under Trump.

The Chinese $295 billion plan calls for using 80% domestic chips — primarily those developed by Huawei (the Ascend 910 series) — to work around this embargo. These chips are significantly less powerful than their NVIDIA equivalents. According to available analyses, Huawei's Ascend 910B delivers approximately 60 to 70% of the NVIDIA H100's performance on certain AI workloads, with higher energy costs. The performance gap per dollar invested is therefore even larger than the investment ratio alone suggests.

China is investing massively in domestic chips that are less powerful. That is not a reason to dismiss the effort — it is precisely what you must do when you are cut off from the world's best chips by an embargo. And Huawei is moving fast. But there is a difference between "advancing quickly in a narrow corridor" and "competing in open space." American export controls on semiconductors may be the most important technology-strategy decision of the last decade. They deserve to be named as such.

Alibaba, Tencent, ByteDance, Baidu: an additional $200 billion

The comparison between the Chinese state plan and American Big Tech investments often omits an important reality: China's private technology companies have also announced massive AI investments for 2026. Alibaba announced investments of approximately $53 billion in cloud and AI. Tencent, ByteDance, and Baidu each announced significant investment programs. The total for all Chinese private hyperscalers is estimated at more than $200 billion additional in 2026.

If you add the state plan ($59 billion per year in annualized projection over 5 years) and the private investments ($200 billion in 2026), the total Chinese AI and infrastructure investment in 2026 would be on the order of $260 billion. That is still significantly below the $725 billion from American Big Tech, without counting the hundreds of other American actors. But it is a gap of 2.8 to 1, not 12 to 1. The nuance matters.

The structural constraints on China's private tech companies

China's private hyperscalers face constraints their American counterparts do not encounter. First, access to cutting-edge chips is limited by export controls — like the state operators, they must turn to less capable domestic alternatives or to complex and costly supply routes through third-party countries. Second, the Chinese regulatory environment is less predictable — the regulatory crackdown launched by the government against Alibaba, Tencent, and Didi between 2021 and 2023 demonstrated that China's private tech companies operate under a structural regulatory uncertainty absent from the American environment.

Third, the target markets are different. American tech companies serve a global market — their cloud and AI services are deployed on every continent, generating revenues that fund the next round of investments. China's private technology companies primarily serve the domestic market and certain Global South markets, with limited access to Western economies for geopolitical and regulatory reasons. This difference in accessible market size is a structural factor that affects long-term investment capacity.

There is something troubling about the trajectory of China's private tech companies. These firms — Alibaba, Tencent, ByteDance — are extraordinary private creations, genuine engines of innovation. But they operate under the permanent threat of unpredictable government regulation, possible partial nationalization, and a geopolitical glass ceiling limiting their international expansion. Investing massively in this context requires trust in the stability of the political framework that these companies' leaders must cultivate with care. That is not the soil of unconstrained innovation.

What official budgets do not reveal about the military AI race

The AI investment figures discussed in this fact-check are civilian figures — company and state plan investments for commercial infrastructure. They do not reflect military and defense spending on AI, which involves separate, often classified budgets in both countries. In the United States, the Department of Defense and intelligence agencies have their own AI programs distinct from the Big Tech budget. In China, the People's Liberation Army has an explicit military AI strategy directly tied to the doctrine of "military-civil fusion."

China's military-civil fusion doctrine is particularly important for the comparison. It stipulates that civilian technological advances must be directly transferable to military applications, and vice versa. In practice, this means that Chinese civilian AI investments — including the $295 billion state plan and private hyperscaler investments — carry an implicit military dimension that has no direct equivalent in the American tech ecosystem, where the separation between civilian and military is, at least in theory, more clearly drawn (even though defense contracts with Big Tech do exist).

NVIDIA locked out of China: the hidden weapon of export controls

The fact that NVIDIA is "locked out of China" — as the source article's headline states — is one of the most significant developments in Sino-American technological competition. American export controls prohibit NVIDIA from selling its most advanced chips (H100, H200, Blackwell) to Chinese entities. This decision, perceived by some as a commercial sanction, is in reality a strategic decision of the first order: it deprives China of access to the tools that make cutting-edge AI possible.

China's response — developing its own domestic chips, building data centers with 80% Huawei chips — is the rational alternative in the face of this embargo. Huawei has made notable progress, and the $295 billion plan is partly designed to accelerate that progress by creating massive domestic demand for Chinese chips. It is an ambitious technological bet. But the current gap between the best NVIDIA chips and the best Huawei chips remains significant, and closing it will take several development cycles — probably 3 to 5 years at minimum according to industry experts.

American export controls on Chinese semiconductors are the technological equivalent of a strategic military decision. By denying China the best chips, the United States created an artificial but real technological gap in the AI race. It is a double-edged weapon — it slows China, but it also pushes Huawei to innovate faster on domestic chips. In ten years, this decision may be judged as the most consequential of the decade in global technological competition.

The research ecosystem: where the real AI race is happening

Infrastructure investments — data centers, chips, cables, energy — are the "muscles" of the AI race. But the "brains" are the models themselves: the large AI architectures whose performance depends on both available computing power and the quality of the underlying research. On this terrain, the comparison between China and the United States is more nuanced and more competitive.

China has produced high-quality models. DeepSeek, developed by a Chinese startup funded by the High-Flyer investment fund, was presented in early 2025 as capable of rivaling the best American models on certain benchmarks — at a fraction of the computational cost. Alibaba has published models in the Qwen series comparable to many Western open models. Baidu maintains its ERNIE program. Chinese AI research, measured by the number of academic publications and patent filings, is among the most active in the world.

Where the American lead remains structural

Despite these real Chinese advances, America's lead in AI research remains structural across several dimensions. Access to talent: the United States attracts the best AI researchers from around the world — many engineers at Google DeepMind, OpenAI, Anthropic, and Meta AI are foreign-born, including many from China. The open ecosystem: the culture of open source and academic publication in the American tech ecosystem accelerates collective innovation in a way that the Chinese system, more closed on certain strategic files, cannot fully replicate.

Access to training data: large language models require massive quantities of high-quality English-language data — the dominant language on the global internet. The English-language data advantage structurally favors American models, even as Chinese models advance rapidly in Mandarin and other languages. And finally, the commercialization pipeline: American companies have access to a global market to monetize their AI models — a crucial advantage for generating the revenues that fund the next generation of research.

What impresses me most in this race is not the investment figures. It is that DeepSeek produced competitive models with a fraction of American computational resources. That means algorithmic efficiency can compensate for part of the chip gap. That is not trivial. China may not be catching up with America in terms of raw computational brute force. But in terms of algorithmic efficiency, the gap is narrowing faster than investment figures suggest.

AI data center energy consumption: a real constraint

The AI race is also an energy race. AI data centers are extraordinarily energy-intensive — training next-generation large models can consume as much electricity as a medium-sized city for several weeks. China's $295 billion plan explicitly includes investments in energy distribution networks around data centers. The chosen ratio of domestic chips (80% Huawei) also implies higher energy consumption per unit of computation — current Huawei chips are less energy-efficient than their NVIDIA equivalents.

For the United States, the explosion of Big Tech capex spending in 2026 is also running into a real energy constraint. States like Virginia (the hub of American data centers), Texas, and Georgia face electricity demand that exceeds the current capacity of their power grids. Data center projects are being delayed by insufficient available electrical capacity. This structural constraint is one of the main risks to the pace of American AI infrastructure expansion in the years ahead.

The role of nuclear power in both countries

Facing the energy constraint, both countries are turning to nuclear power as a long-term solution. The United States is experiencing renewed interest in nuclear energy — notably in small modular reactors (SMRs) that could directly power data centers. Microsoft signed an agreement with Constellation Energy to reopen a unit of the Three Mile Island plant. Google and Amazon have similar programs in development.

In China, the existing nuclear fleet is one of the world's largest, and many new reactors are under construction. The AI infrastructure plan likely includes a nuclear energy component in the power mix for future data centers. This energy dimension of the AI race is often absent from financial analyses — but it may be one of the most important limiting factors for both powers over the next five years.

The AI race is an energy race as much as it is a race for computing power. And energy is geopolitics in its most fundamental form. Whoever masters clean, abundant, stable energy to power its data centers will have a durable competitive advantage in AI. This is another dimension of the race that tech investment figures do not capture. It deserves separate, serious attention.

The five truths this fact-check establishes

At the end of this analysis, here are the five factual assertions this fact-check establishes with confidence. First truth: China's $295 billion is a state plan over five years, not yet officially adopted at the time of the announcement, financed by public debt, involving primarily state operators. Second truth: America's $725 billion represents the infrastructure capex spending of the four American Big Tech companies for the single year 2026, representing a 77% increase over 2025.

Third truth: comparing over the same five-year period and including Chinese private hyperscaler investments (approximately $200 billion per year), total Chinese investment runs to approximately $260 billion per year versus $725+ billion per year for American Big Tech — a gap of at least 2.8 to 1, without counting other American actors. Fourth truth: the chips used in the Chinese plan (80% Huawei Ascend) are less powerful than the NVIDIA chips used in the United States, widening the effective gap beyond the financial ratio. Fifth truth: China is making real algorithmic advances (see DeepSeek) that can partially offset the computing power gap.

The two interpretation risks to avoid

This fact-check also identifies two interpretation risks to be avoided. The first: underestimating the Chinese technological threat by relying solely on the investment ratio unfavorable to China. Investment figures do not capture algorithmic advances, the national mobilization behind this effort, Huawei's effectiveness in developing domestic chips, and the geopolitical ambition driving this investment. China is a serious technological competitor that deserves to be taken seriously.

The second risk: overestimating America's lead by presenting the competition as more balanced than it is. China's $295 billion over five years is not "comparable" to America's $725 billion in one year. This erroneous framing can lead to disproportionate response policies, hasty and poorly targeted American public spending, and an alarmist narrative that harms democracy rather than protecting it. Factual truth is the best foundation for informed policy decisions.

The danger is not that China is more advanced than America in AI — it is not, not yet, not structurally. The danger is that China's pace of progress — more efficient algorithms, rapidly improving chips, national mobilization — could close the gap faster than American and Western decision-makers anticipate. That is a problem of dynamics, not statics. And investment figures are a useful but insufficient indicator for measuring those dynamics.

AI as a national power asset

The AI race between China and the United States is not an ordinary commercial competition. It is the arena of a national power rivalry that will determine who controls the cutting-edge technologies of the next decade — and with them, the economic, military, and geopolitical advantage. The United States has always understood this dimension: export controls on semiconductors, the "Chip Act" policy, restrictions on Chinese technology investments in the United States are all strategic responses to this reality.

China articulated the same understanding in its doctrine of the "AI society" — the official goal of becoming the global leader in AI by 2030. The $295 billion plan is an instrument of this national strategy. It is not a commercial investment guided by market signals. It is a strategic investment guided by a long-term geopolitical vision. This difference in nature — American private market vs Chinese strategic state — is one of the most important in the comparison.

What Europe must understand about this race

Europe, watching this race between the United States and China, must draw a lesson that directly concerns it: the absence of a European tech ecosystem of comparable size to either of these two blocs is a real strategic risk. America's $725 billion and China's $260 billion are to be compared with a few tens of billions in annual European tech investment. That is not a flattering comparison. Europe risks finding itself a user of technologies developed by others, with implications for its digital sovereignty, its security, and its economic competitiveness.

The European response — the AI Act, French-German sovereign cloud projects, investments in deep tech startups — is necessary but insufficient in the face of the scale of investment by the two technological superpowers. Europe must decide whether it genuinely wants to be an independent technological power or whether it is content with the role of sophisticated regulator in a world where technologies are developed elsewhere. This strategic choice deserves a far more serious public debate than currently exists.

What I take away from this fact-checking exercise is that numbers, poorly understood, can lead to bad decisions. America is not losing the AI race. But China is advancing — differently, with different constraints, at a pace it would be imprudent to underestimate. And Europe, meanwhile, is regulating with excellence what it does not build. That is a problem no AI regulation can solve on its own.

Talent, education, and human capital

Neither the $295 billion nor the $725 billion captures one of the most decisive resources in the AI race: human capital. The number of AI researchers trained each year, the quality of technical universities, the capacity to attract the world's best talent, and the conditions that allow innovation to develop freely: these are factors that cannot be reduced to dollars invested in data centers. On these dimensions, the comparison is more nuanced.

China trains more AI engineers per year than the United States. But the United States attracts a significant proportion of the world's best AI talent — including Chinese nationals. Academic freedom, a culture of experimentation, and the absence of censorship on training data and research topics give American researchers a working environment that the Chinese system cannot offer for certain areas of AI. This qualitative factor, difficult to measure, may be as important as the billions invested in infrastructure.

Governance and trust: AI for whom, controlled by whom

The final dimension that investment figures do not capture is the question of governance and trust. AI systems are not neutral — they reflect the values and power structures of the societies that develop them. An AI developed in an authoritarian system, trained on censored data, deployed to monitor a population: it is not the same tool as an AI developed in a democratic framework with protections for individual liberties. This governance difference is the most important difference of all — and no investment figure captures it.

For democracies that will need to choose where their AI comes from — American platforms, Chinese systems, or locally produced alternatives — this dimension is decisive. An AI optimized for surveillance is not an AI optimized for freedom. That may be the deepest reason why the United States and its allies must maintain their technological lead in AI: not only for economic or military reasons, but because the value of AI systems depends on the values that built them.

Ultimately, the real question in the AI race is not "who spends more." It is "what kinds of AI are we building and for what purpose." An AI that helps doctors, frees workers from repetitive tasks, improves education, and accelerates scientific research — that is an AI in service of humanity. An AI that surveils, controls, censors, and optimizes repression — that is a weapon. America's $725 billion and China's $295 billion are not in the same race. They are in two different visions of the world.

DeepSeek and the algorithmic efficiency effect

DeepSeek, the Chinese startup that produced language models in early 2025 comparable to the best American models at a fraction of the computational cost, illustrates a dynamic that Western analysts often underestimate: constraint can be a source of innovation. Cut off from the best NVIDIA chips by export controls, Chinese research teams were forced to optimize their algorithms more aggressively — finding more efficient architectures, training techniques that achieve comparable results with less computation. DeepSeek R1 and its successors are the product of that forced constraint.

This phenomenon is known in innovation economics as "creative constraint" — the absence of infinite resources that forces engineers to be more ingenious. In the specific case of AI, it suggests that the chip gap between China and the United States does not mechanically translate into a proportional gap in model performance. A Chinese research team that must achieve as much with half the chips will develop optimizations that could ultimately benefit the entire AI field — including American actors who adopt these techniques.

China's AI startup ecosystem: beyond the hyperscalers

Beyond DeepSeek, a dense ecosystem of Chinese AI startups has developed in recent years — in computer vision, natural language processing, robotics, and medical and industrial applications. Companies like Zhipu AI, Moonshot AI, MiniMax, and dozens of others have raised significant funds and developed commercial products deployed at scale. This ecosystem is less visible internationally than the American Big Tech players — partly because of a language barrier, partly because of geopolitical restrictions on their international expansion — but it is real and dynamic.

This level of Chinese AI startup activity is not captured in the $295 billion of the state plan nor in the estimated $200 billion from private hyperscalers. It is a third layer of investment and innovation that standard comparisons ignore. If one were to include investments in Chinese AI startups — venture capital funding, regional government incubation programs, corporate investment by large companies in startups — the total annual Chinese AI ecosystem figure would draw even closer to the full American ecosystem. The gap remains, but it is more complex to measure than a simple two-figure comparison implies.

DeepSeek sent shockwaves through American financial markets in early 2025 — NVIDIA's stock plunged the day its performance was made public. That reaction reveals both the weight attached to Chinese competition and the nervousness of markets toward any threat to American Big Tech's dominance in AI. This nervousness is healthy to a degree — it maintains competitive pressure. But it should not lead to policy decisions driven by fear rather than by precise factual analysis. This fact-check tries to be that precise factual analysis.

When four companies control the infrastructure of global intelligence

The concentration of $725 billion in investment among four private companies (Amazon, Microsoft, Alphabet, Meta) raises important questions about the power structure of the global AI ecosystem. These four companies together control the majority of global cloud infrastructure, the most widely used AI models, and unprecedented quantities of personal data in human history. This concentration creates structural dependence — governments, businesses, and public institutions that cannot function without these platforms.

This is not without risk. Even though these companies operate within democracies with regulatory frameworks, their scale gives them a political and economic power that exceeds that of most nation-states. The question of how to govern these giants — ensuring they serve the public interest as much as their shareholders — is one of the major policy challenges of our era. Massive AI investment accelerates this concentration rather than reducing it.

The symmetric Chinese risk: a state that controls everything

Concentration also exists on the Chinese side, but with a different structure: it is the State that controls the infrastructure, not private companies. The $295 billion plan deployed through China Mobile and China Telecom — two state entities — creates a model where the Chinese government owns the infrastructure on which the country's AI applications run. This state concentration has advantages in terms of coordination and rapid deployment. It also carries major risks: AI as a tool of social control, mass surveillance facilitated by centralized infrastructure, the absence of private or civil counterweights against state excess.

These two models of concentration — American (private Big Tech) and Chinese (sovereign State) — represent two different risks for global AI governance. Neither is satisfactory from the standpoint of humanity's collective interest. The challenge is to build alternative models — with public participation, democratic oversight mechanisms, and a broader distribution of benefits — that reduce neither to Big Tech dominance nor to authoritarian state control.

The real challenge of the AI race is not figuring out whether America or China will "win." It is ensuring that, whatever the outcome of this competition, the technologies developed serve all of humanity and not the interests of a handful of corporations or an authoritarian regime. That is a global governance challenge we have barely begun to take seriously. And time is pressing, because the systems are already being deployed.

The AI Safety Summit: a multilateral governance under construction

The race for AI investments between China and the United States is unfolding in a context of global AI governance that is still largely unfinished. The AI Safety Summit at Bletchley Park in 2023, followed by meetings in Seoul and Paris, laid the first stones of an international dialogue on AI risks and norms. These initiatives are valuable. They are also fragmentary, non-binding, and insufficient given the speed at which systems are being deployed.

China participated in some of these discussions, signaling a minimal willingness to engage on AI safety questions. But its governance positions remain fundamentally different from those of Western democracies on questions of free expression, surveillance, and military use of AI systems. This divergence of values makes it difficult to build a truly global governance regime — we risk ending up with two parallel standards: a "Western AI" and a "Sino-Russian AI," with incompatible values and rules.

The EU as a governance actor: the European regulatory model

In this still-murky global governance space, the European Union plays a particular role with its AI Act — the world's first comprehensive artificial intelligence regulation, adopted in 2024. The European model — based on risk categorization, fundamental rights, and transparency — differs from both the American model (less regulated, guided by market innovation) and the Chinese model (state-governed, oriented toward national security and social control).

The European AI Act will have a global ripple effect — as the GDPR did for data protection. Global companies that want to access the European market will need to comply with its requirements, which will influence their global practices. In the AI race between China and the United States, Europe does not win on the dimension of investment or computing power. But it may win on the terrain of norm-setting — establishing the rules that others will need to follow if they want to access its 450 million consumers. That is a less visible but real form of power.

Europe regulating what it does not build — it is an easy and partially fair criticism. But there is another reading: Europe establishing the norms that others will be required to follow. The GDPR forced American Big Tech to change their global practices to access the European market. The AI Act could do the same. That is not sufficient in the face of America's $725 billion and China's $295 billion. But it is more than purely local regulation. It is an ambition for global governance that Europe should claim more clearly.

Disinformation is not an accident: it is a strategy. And as long as we treat it as a technical problem to be solved by algorithms, we will lose the information war. It is a political war that demands political answers.

What this fact-check has established

The comparison of China's $295 billion vs America's $725 billion is accurate in its raw numbers and misleading in its typical framing. It compares a five-year state plan with the annual spending of four private companies. It omits the investments of China's private hyperscalers. It ignores the chip performance gap. It overlooks the algorithmic dynamics that partially offset the investment differential. And it says nothing about the qualitative dimensions — talent, governance, trust, values — that may ultimately be the most determinative over the long term.

The reality is this: the United States maintains a significant structural lead in the AI race, in terms of investment, computing capacity, and research ecosystem. China is a serious competitor that is advancing rapidly, particularly on algorithmic efficiency and domestic chip development. And this race is not only a competition between two powers — it is a question of which vision of artificial intelligence — free or controlled, open or closed, in service of humanity or the state — will come to dominate the world of tomorrow.

The fact-check's final verdict

VERDICT: The comparison as typically presented is PARTIALLY TRUE but MISLEADING IN ITS FRAMING. The figures are real. The side-by-side presentation without specifying duration, the nature of the actors, and the competitive context misleads on the true scale of the gap between the two powers. America's lead is larger than the formula suggests. China's progress is more real and more qualitative than the financial comparison alone implies. The truth lies in both nuances simultaneously.

By Maxime Marquette, columnist

Editorial stance

I am not a journalist but a columnist and analyst. My expertise lies in observing and analysing the geopolitical, economic and strategic dynamics that shape our world. My work is to take political strategies apart, to understand global economic movements, to put the decisions of international actors in context and to offer analytical perspectives on the changes that are redefining our societies.

I do not claim the cold objectivity of traditional journalism, which stops at reporting facts. I claim analytical clarity, rigorous interpretation and a deep understanding of the complex issues that concern us all. My role is to give meaning to the facts, to place them in their historical and strategic context, and to offer a critical reading of events.

Method and sources

This text keeps the basic distinction between verified facts and interpretive analysis. The rule never changes: a factual claim is published only when a verifiable source supports it, and the sources this article actually used are listed under Sources, never here.

Primary source families used by the publication, when the story has them: official statements from governments and international institutions, public statements by political leaders, reports from intergovernmental organisations, dispatches from recognised international news agencies.

Secondary source families: specialised publications, internationally recognised news media, analyses from established research institutions, reports from industry organisations.

When an article cites statistical, economic or geopolitical data, it comes from the institutions that produce it (intergovernmental organisations, central banks, national statistics offices), and the exact institution is named under Sources.

Nature of the analysis

The analyses, interpretations and perspectives in the analytical sections of this article are a critical, contextual synthesis based on the information available, the trends observed and the expert comments cited in the sources consulted.

My role is to interpret these facts, to set them in the context of today's geopolitical and economic dynamics, and to give them a coherent meaning within the larger story of the changes shaping our time. These analyses reflect expertise built through continuous observation of international affairs and an understanding of the strategic mechanisms that drive global actors.

This article describes a documented state of affairs on its publication date, not a prediction: later developments may shift its outlook. No update is promised in advance; when an article is corrected or completed, the change is dated in the text.