On March 5, 1946, a politician out of power stood at a small college in Fulton, Missouri, and named a division most of the world was not yet ready to admit existed. Winston Churchill, defeated at the ballot box the year before and speaking on foreign soil with President Truman seated behind him, declared that an iron curtain had descended across Europe. He did not create the division — the Soviet consolidation of Eastern Europe was already underway — but he gave it a name, a shape, and a moral valence, and in doing so he helped turn a diffuse anxiety into an organizing framework for the next forty years. Fulton mattered less as an event than as a marker: the moment the Cold War became something you could plan around.
Xi Jinping’s first in-person keynote at the World Artificial Intelligence Conference in Shanghai on July 17 has the same structural quality. It, too, named a division. It, too, was delivered as one half of a matched pair — Xi’s coalition-building address landed within hours of a combative broadcast from President Trump, and the split screen made the point better than either man could alone. And it, too, may be remembered not for what it launched but for what it made undeniable: that the world is bifurcating into two artificial-intelligence systems, and that the terms of that split are now being set in public. For markets, that is the signal worth extracting from the rhetoric.
The Speech Beneath the Symphony
Stripped of its cadences — Xi reached for a Chinese aphorism about a single string being unable to make music, and called for a symphony of international cooperation — the address made three concrete commitments. It doubled down on open-source AI as China’s chosen strategy. It formally launched the World Artificial Intelligence Cooperation Organization, an intergovernmental body headquartered in Shanghai, with 29 founding members including Brazil, Russia, Pakistan, and Kazakhstan. And it pledged five years of Chinese support for AI capacity-building across the developing world, framed as a remedy for what Xi called the historical injustice of unequal access.
Like Churchill, Xi paired the naming of a division with the rallying of a bloc. Where Fulton summoned an Anglo-American front against an advancing East, Shanghai summoned a Global South coalition against a restrictive West. The rhetorical machinery is nearly identical. What has flipped is who claims to be standing on the open side of the curtain.
The Inversion That Matters
This is where the analogy stops being decorative and starts being analytically useful. In 1946, the open society was warning about the closed one; the West cast itself as the party of free exchange and the Soviet bloc as the party of walls. In 2026, Beijing has appropriated the language of openness and assigned the role of the walled-off, restrictive power to Washington. American export controls on advanced chips are the wall in Xi’s telling, and China’s open-weight models are the free exchange the controls are meant to suppress.
The inversion is not merely rhetorical sleight of hand, and that is precisely why it is dangerous to dismiss. It reflects a genuine strategic reality: the United States is, in fact, pursuing dominance through restriction — through control of the chokepoints in the semiconductor supply chain — while China, locked out of the highest tier of hardware, has rational reasons to compete through distribution instead. A country that cannot win on proprietary frontier compute can still win by making its models the default substrate for everyone outside the American orbit. Openness, for Beijing, is not a value. It is the optimal strategy given the constraints, and it happens to come wrapped in language the developing world finds far more appealing than the language of controls.
Where the Curtain Actually Falls
Churchill’s curtain ran along a geography — from the Baltic to the Adriatic. The AI curtain runs along a technology stack, and investors need to know which layer is dividing.
It is not, primarily, a division of applications. It is a division of foundations: compute, models, and standards. On one side sits the American stack — the most advanced accelerators, the frontier proprietary labs, and an export-control regime designed to keep the top of that stack scarce and domestic. On the other, an emerging Chinese stack built around low-cost open models in the DeepSeek mold, domestic accelerators of the kind Huawei showcased in Shanghai, and now an institutional home in the WAICO to propagate Chinese standards outward. The two stacks are becoming mutually less interoperable by design, and the countries of the Global South are the contested territory each is trying to enroll — the equivalent of the non-aligned nations both blocs courted after Fulton.
The Market Map of a Bifurcated Stack
For an investor, a bifurcating AI world is not an abstraction; it is a set of concrete second-order effects.
The first is that the American chokepoint strategy is bullish for the chokepoint holders precisely to the extent that it works. The value of controlling scarce frontier compute rises as the world divides, because scarcity is the entire point of the position. The names that sit on the toll roads of the Western stack — the dominant accelerator supplier, the custom-silicon and networking franchises, the memory and optical layers that the buildout cannot proceed without — derive their moats from exactly the concentration Xi is objecting to. A world that hardens into blocs does not erode those moats on the American side; it entrenches them, so long as the restrictions hold.
The second effect runs the other way. Every country China successfully enrolls in an open-Chinese-model ecosystem is a market where Western frontier vendors are structurally disadvantaged, not by price but by policy and alignment. The addressable market for the American stack shrinks at the edges even as it deepens at the core. The Global South, in this framing, is not a rounding error — it is the fastest-growing pool of future AI demand, and Beijing is moving to capture it at the foundation layer, where switching costs compound over time.
The third effect is the one the market has not yet priced: a bifurcated stack means duplicated capital expenditure. Two sets of foundation models, two hardware supply chains, two standards regimes, each requiring its own buildout. Duplication is inefficient for the global economy and, in the near term, additive for the suppliers of the picks and shovels on both sides. The same forces that make the world more fragmented make the aggregate infrastructure bill larger, not smaller.
What Makes a Fulton a Fulton
The caution embedded in the analogy is also the investment caveat. Fulton is remembered as a turning point only in retrospect, because the division it named went on to harden. Had East-West relations thawed in 1947, the speech would be a footnote. The same contingency applies here. WAIC 2026 becomes the AI Cold War’s Fulton only if the coalition it announced acquires substance — real governance rules, real standards adoption, real capacity transfer — rather than remaining a membership list and a headquarters.
There is also a contradiction at the center of Xi’s framing that could undercut the whole edifice, and markets should track it closely. Even as Beijing promotes openness abroad, it has reportedly been weighing restrictions on overseas access to its own most capable models, in discussions with firms including Alibaba and ByteDance. If China ends up guarding its frontier the way Washington guards its chips, the open-society mantle slips, the Global South pitch curdles into something more transactional, and the neat inversion collapses back toward a plain contest of two restrictive powers. That would be a materially different world to invest in than the one Xi described in Shanghai.
The Investable Conclusion
Two visions are now on the table, stated plainly by their principals: one that routes global AI through open Chinese models and a Beijing-anchored institution, and one that keeps the frontier concentrated, scarce, and governed on American terms. The base case an investor should plan around is no longer convergence toward a single global AI market. It is fragmentation — a silicon curtain running through the foundation layer of the most important technology of the decade.
That base case has clear directional implications. It supports the chokepoint holders of the Western stack for as long as the controls bind. It creates a long-duration demand frontier in the Global South that accrues disproportionately to the Chinese ecosystem. And it argues for a larger, not smaller, aggregate infrastructure spend as the world builds two of everything. Xi did not start this division any more than Churchill started the last one. But like Churchill, he has now named it out loud — and naming it is how it becomes something you can no longer afford to ignore.
