The essence of the US–China AI competition lies not in any technological generation gap, but in the fact that the two countries’ fundamental logics for embedding AI into society are completely different: China treats it as new-quality productive forces, while the United States has turned it into new-quality finance. This divide determines two endings — a community with a shared future for mankind, or Cyberpunk 2077.
Overall Thesis and Main Arguments
All analysis of the US–China AI competition should shift from a “comparison of technological capabilities” to a “comparison of social-embedding logics” — what the two countries are contesting is not AI itself, but the direction in which AI is put to service and the power to define AI.
Main argumentative framework (nine insights):
- A flawed historical premise: there has never been a true “golden age of free trade” between China and the US; the historical premise behind US chip controls is a fabrication.
- The collapse of strategic assumptions: all three assumptions underlying the US chip war (“choke it and it dies,” “racial superiority,” “penalty in place of governance”) have proven self-refuting.
- An institutional structure of corruption: “compliance-as-corruption” makes it impossible for the US to build a genuinely effective system of strategic controls.
- The essential divide between two routes: the logic of productive forces (China) versus the logic of finance (the US) yields two different definitions of AI success.
- The “ghost story” mechanism: AGI, as an unfalsifiable vehicle of financial narrative, serves the perpetuation of the capital bubble.
- The real constraint is electricity: the endpoint of AI competition is not chip performance, but the combined constraint of electricity cost plus geopolitical security.
- Open source as a historical turning point: DeepSeek’s open-sourcing was a deliberate declaration of strategic direction, not merely a technical choice.
- Two ultimate endings: Cyberpunk 2077 (financial monopoly) versus a community with a shared future for mankind (universal and inclusive).
- The self-contradiction of Western ideology: championing “freedom” while inevitably moving toward “monopoly” is a fundamental incoherence it cannot resolve.
I. A Flawed Cognitive Premise — The “Golden Age” Never Existed
Shen Yi’s first structural argument: there has never been a so-called “golden age of free trade” between China and the United States.
The real-world significance of this judgment is that US chip-control policy rests on a false historical memory — it assumes that China once had smooth access to advanced US technology and has now been “cut off.” But the actual history is that from the Paris-based Coordinating Committee (CoCom) to the Wassenaar Arrangement, controls have never been interrupted. The Sino-US honeymoon of the 1980s was an exception, made possible only by the political assumption that China would “take the Gorbachev path,” and it always carried the attached premise: “you don’t challenge me, you pose no threat to me.”
“From 1949 onward — after the founding of the People’s Republic of China — there has never been any so-called golden age of China-US trade. As if advanced US technology could flow smoothly and freely between China and the United States — that thing does not exist; it is something people imagined.”
China’s expectations toward the United States were, from the very beginning, a fantasy built on a mistaken perception. In realist international relations, advanced technology has never been a “commodity in circulation” — it is a “vehicle of power.”
II. The Three Strategic Assumptions of the US Chip War — and Their Collapse
US chip sanctions against China rest on three mutually reinforcing assumptions, each of which has proven self-refuting in reality:
Assumption One: “Choke it and it dies” — China will voluntarily abandon development under sanctions
That is: once the US imposes a blockade, China says “fine, then we won’t build it — we give up,” and the game is over. This assumption sets the precondition for sanctions’ success as “China’s self-abandonment” rather than America’s technological lead.
Assumption Two: The “racial superiority” thesis — the Chinese cannot build advanced chips
An assumption of racial supremacy holding that the Chinese are innately incapable of innovation and can only imitate and follow. This assumption supplies the moral justification for continuously escalating controls.
“Suppose the United States firmly believes in race-based superiority — that America is smarter than China, that the Chinese cannot build this thing themselves. So I choke; once I choke, you have nothing — you cannot possibly survive, you are certain to die. If you haven’t died, then I simply haven’t choked enough, so I keep choking.”
Assumption Three: “Penalty in place of governance” — the commercialization paradox of sanctions
The Commerce Department’s Bureau of Industry and Security (BIS) has become the most revenue-generating agency, for a simple reason: if the flow is genuinely blocked, there are no fines to collect. What actually operates, therefore, is a “gray market” logic — the nominal blockade becomes a toll-collection network rather than a true embargo.
All three assumptions of US chip strategy are self-refuting. A strategy that can succeed only if the adversary voluntarily gives up is, in essence, the externalization of a failed strategy.
III. The Deeper Structure of “Penalty in Place of Governance” — Compliance as Corruption
With a “feed trough” metaphor, Shen Yi exposes the operating mechanism of institutional corruption in the United States:
“Anti-corruption in America — what is America? America is the legalization of corruption. So America has no corruption — everyone is corrupt; America has no illegal corruption, it is all legalized. In America people never say ‘legal’ — what they always say is ‘compliant.’ What is compliance? Compliance is me pointing you to a doorway, and everyone gathers around the — to put it bluntly, there is a metaphor used on the island of Taiwan: when a certain somebody used to engage in corruption, people saw the feed trough of a pig farm — feed poured into the trough, and everyone eats from it together, all getting a share. Corruption in America is exactly this spoils system.”
The essential difference between the two countries’ approaches to anti-corruption:
- China: removes the feed trough entirely, so that corruption has nowhere to occur.
- United States: ensures that only “compliant” pigs get to eat from the trough — corruption is part of the institutional design.
Mapped onto the sanctions domain: every new export control creates a new “feed trough” — a new license, a new revenue opportunity. The “stricter” the sanctions, the worse the actual blocking effect — because “blocking the flow” means “losing the people who pay the tolls.”
The “compliance-ization” of corruption in the US system makes it impossible to build a genuinely effective system of strategic controls. Any external sanctions tool will ultimately be internalized as a revenue source for domestic interest groups in the United States.
IV. Two Routes for AI Development — Productive Forces vs. Finance
This is the core framework of Shen Yi’s analysis and the main axis running through the entire piece:
“What is the essence? The essence is that China is developing artificial intelligence as new-quality productive forces, while the United States is developing artificial intelligence as new-quality finance.”
| The China Route | The US Route | |
|---|---|---|
| Essential positioning | New-quality productive forces | New-quality finance |
| Constituent elements | Infrastructure + power operations & maintenance + vertical applications | Large models + compute centers + capital narratives |
| Core goal | Lower barriers, improve accessibility, achieve endogenous growth | Sustain monopoly, harvest belief, maintain financial hegemony |
| Definition of success | Verifiable real-world output | Continuation of the narrative; the bubble never bursting |
| Core competitiveness | Real industrial system, infrastructure capacity, engineer dividend | Narrative capacity, dollar hegemony, the SWIFT system |
The essential difference: AI is a tool in Chinese society, and an end in American society.
- When a tool serves society, the standard of success is verifiable real-world output.
- When the end itself is financial return, the standard of success is the maintenance of the narrative.
To judge the nature of a country’s AI development route, do not look at what it says — look at who gets to define its success: whether it is priced by capital, or defined by social need.
V. The “Ghost Story” Mechanism — A Structural Need of Financialized AI
The core mechanism of AI financialization in the United States is the “ghost story” — a grand narrative that can never be verified, needed to keep capital flowing in.
Why must AI be packaged as “artificial general intelligence” (AGI)?
- AGI is unfalsifiable: what cannot be falsified never has to face empirical testing.
- AGI is unpriceable: what cannot be priced can be valued without limit.
- AGI makes for good storytelling: the bigger the ghost story, the more capital is willing to hold the bag.
“Why does the United States push this so-called artificial general intelligence? It’s because — I’m not saying there’s anything wrong with strong AI itself — but for a financial market like America’s, it simply needs a ghost story. Strong AI is a ghost story that’s very easy to build a pitch around, convenient for everyone to make money together — every kind of capital showing its own magic, everyone lifting the sedan chair together. But in reality it can never land, and people even hold a psychological expectation that it cannot land — becoming a situation in which everyone willingly, happily, consents to mutually swindling money.”
“Stargate” is the textbook case: a $500 billion promise → $200 billion in the first phase → shrunk to $40 billion → ultimately $10 billion for building a single data center. Its core function is to give the capital narrative a bubble vehicle that never bursts.
When the development goal of a technology shifts from “solving problems” to “sustaining a narrative,” its substance is no longer technological progress but financial engineering.
VI. “The Tyranny of the Watt” — The Real Constraint on AI Competition
The real constraint on AI is not chips, but electricity.
The key metric of a compute center is not performance but power consumption. Reaching equivalent compute, China needs lower-end chips → higher power draw → higher electricity bills. But electricity prices are not equal around the world: China’s electricity cost is roughly one quarter of America’s.
“You use one kilowatt-hour, I use two — but my electricity cost is a quarter of yours. My two kilowatt-hours are half as cheap as your one. Same performance, twice the power draw, operating cost only fifty percent of yours. Do you want to compete with me on operations? You say your compute — do you want to compete on the price of compute? You cannot beat me.”
This means:
- At equivalent performance, the operating cost of Chinese compute centers is fifty percent of that of US centers.
- The dimension of competition is not “chip performance” but “effective compute per kilowatt-hour.”
The siting of compute centers must also take geopolitical security into account — compute centers in the Middle East face missile risk, and China is the safest option.
“Today, anywhere in the world that goes to build a compute center — who dares not factor geopolitics into the calculation? Once you do — where in the whole world is safest? China!”
AI competition is, on the surface, a contest of chips; in the middle layer, a contest of electricity; at the base, a contest over the completeness of the industrial system and geopolitical stability. Any analytical framework that focuses only on “chip performance” misses the real constraints.
VII. Open Source as a Historical Turning Point — What Liang Wenfeng’s Choice Means
DeepSeek’s decision to open-source is characterized by Shen Yi as a historic decision — one that “held the key to the course of history.”
“Liang Wenfeng isn’t doing charity either — he’s here to make money. But if I ever get the chance in the future, I really want to know the answer to one question — how did he ever come up with open-sourcing DeepSeek? That was a historic turning point. History was heading this way — in that sense, whether history goes this way or that way, in that moment this man held the key to the process of history.”
Its strategic significance lies in:
- Breaking capital’s pricing power: open source makes it impossible to impose monopoly pricing on technological diffusion.
- Creating a second option: it provides Global South countries with an alternative that does not depend on the United States.
- Changing the nature of the game: it expands the US–China bilateral contest into an ideological competition of “a community with a shared future for mankind versus a world empire.”
Open source is not merely a technical choice but a declaration of position: had DeepSeek chosen closed source, it could have been co-opted, brought into “partnership” — it would not have disrupted the existing order. Open-sourcing meant voluntarily giving up the possibility of co-optation and choosing a “road of no return.”
In technological competition, the choice between open source and closed source is essentially a choice of “alliance strategy” — it decides who stands in your ecosystem, and in whose ecosystem you stand.
VIII. Two Endings — Cyberpunk vs. a Community with a Shared Future for Mankind
Shen Yi explicitly lays out two ultimate endings for AI development:
“When it comes to artificial intelligence in the international system, in the competition of international relations — humanity today, facing future development, in fact faces two roads, two models, two endings. One of the endings is already quite clear to us: Cyberpunk 2077 — Silicon Valley tech, the right-wing neo-technological reactionaries, the future Arasaka Tower. This trajectory is already very obvious.”
Ending A (the US route): Cyberpunk 2077
- Algorithms decide what you can learn and what you can know.
- Knowledge is pay-per-use; those who cannot pay remain permanently at the cognitive bottom.
- Social stratification is defined by algorithms; mobility disappears entirely.
- Humans become appendages of the system rather than ends in themselves.
“The more value you create for me, the more profit you supply, the more exploitable you are, the more I reward you — and the closer you get to my system. As for those out in the wilds, the vast masses — what are they? They provide the reproduction of the population; they are a breeding pen — I raise them like pigs.”
Ending B (the China route): a community with a shared future for mankind
- AI lowers barriers and improves accessibility.
- The gains of development diffuse to broader social groups.
- Technology serves “making life better for the majority,” not “deepening monopoly for the few.”
This is not a question of “which technology is better” but a value choice about “whom AI serves.” The weight of this choice is rising in technological competition, because the convergence point of technological routes is approaching.
IX. The Self-Contradiction of Western-Centered Ideology
Shen Yi points out a fundamental contradiction in how the West promotes AI-governance norms:
| The Western Position | |
|---|---|
| Principle | All people are created equal |
| Practice | Only AI monopolized by the West is safe; open source is unsafe |
| Logic | If everyone can see the source code, the world will “abuse” it — ordinary people have no capacity for self-governance |
The essence of this contradiction: they assume that some people are “fools who cannot be trusted to see the truth” — which collides directly with the liberal values the West itself professes.
“We must avoid one thing — building, out of Western-centric arrogance, a misguided regime of controls over the open-source ecosystem. That violates the principle that all people are created equal. You assume some people are fools — that it would be very unsafe for them to see the source code, and that only when you, the few, monopolize the source code and artificial intelligence is it safe — that violates the basic moral principles that you in the West yourselves hold.”
Technical norms have never been merely technical matters — they are vehicles of values. Every technology-governance framework embeds a fundamental judgment about whether ordinary people are capable of self-governance. The West’s present predicament is that it can no longer coherently champion both “freedom” and “monopoly.”
The essence of the US–China AI competition is a competition between two social logics: China positions AI as a tool serving society (the logic of productive forces), while the United States positions AI as a means of sustaining monopoly (the logic of finance). These two routes have already formed a clear divide in historical evolution, and in 2025–2026, through two landmark events — DeepSeek’s open-sourcing and the faltering of Stargate — the divide became public.
The methodological value of Shen Yi’s analysis:
- It is not technical analysis but social-structural analysis — the choice of technological route is a reflection of the choice of social system.
- It reveals why “chip performance” is not the decisive variable in AI competition — because the real constraints lie outside the chip (electricity, geopolitics, institutions).
- It offers a simple criterion for judging the authenticity of an AI narrative: whatever requires you to “believe it will succeed” is a ghost story; whatever can be verified through output follows the logic of productive forces.
This analysis is a typical application of the “history–structure–institutions” framework in Shen Yi’s methodology: starting from historical facts, constructing structural hypotheses, analyzing the operating logic of institutions, and finally arriving at judgments about historical tendencies. The framework itself is highly non-time-sensitive, and its core conclusions hold up for the foreseeable future.
Supplement: A Complementary Perspective from Wushi Caijing — Black Tech vs. White Tech
The “black tech vs. white tech” framework of Wushi Caijing (a popular finance-focused video essayist) mirrors Shen Yi’s “new-quality productive forces vs. new-quality finance” framework: Shen Yi approaches from the logic of social embedding (whom AI serves and who defines it), while Wushi approaches from the industrial route (how technology reaches the masses). Only placed together do the two frameworks form a complete picture of US–China AI competition.
The core dividing line in Shen Yi’s framework:
- New-quality productive forces → AI as a factor of production serving the whole of society → a community with a shared future for mankind
- New-quality finance → AI as a vehicle of financial narrative serving capital → Cyberpunk 2077
The core dividing line in Wushi’s framework:
- Black-tech route (US) → 0-to-1 breakthroughs, technological glamour, serving elites (0.3% paying users)
- White-tech route (China) → mass democratization at scale, costs driven down to the point of being disposable, serving everyone (75% of people who have never touched AI)
Where the two intersect:
| Shen Yi’s Framework | Wushi’s Framework | |
|---|---|---|
| China’s AI logic | New-quality productive forces (serving production) | White tech (democratization at scale) |
| US AI logic | New-quality finance (serving the capital bubble) | Black tech (technological originality + high unit price) |
| Shared judgment | Electricity is the real hard constraint | The electricity bottleneck + equipment dependence lock in the US ceiling |
| Shared conclusion | China’s AI will go further | China’s AI — both the “face” and the “lining” — will inevitably surpass America’s |
Two insights that Wushi uniquely contributes:
- The inversion of cost structure: in the AI era, free users cost more than free users on the traditional internet — this is not the old “round up users first, monetize later” playbook.
- The endgame of technological invisibility: AI will become vanishing infrastructure (electricity → internet → WeChat → AI). Black tech makes humanity look up in awe; white tech makes humanity live in the future.