Anti Chokepoint

Globalization was built upon the ideas of economic efficiencies, comparative advantage and political convergence until the current age of fragmentation. It was fundamentally a win-win exchange: developed nations kept domestic inflation low and freed capital to push the technological frontier, while developing economies pulled millions out of poverty by supplying the goods those wealthier nations required. This order was underwritten by the post WWII unipolar system led by the US (unironically). Despite its own history of protectionism, the United States championed free markets globally, in exchange for security alignment, the US provided the world with access to its massive consumer market and guaranteed rule of law and order (in a lot of ways) for trade and investment especially. Nations discovered or created their comparative advantage through resource endowments, technological capability, or state led industrial policy, as seen in the rise of South Korea and China. This interconnectedness transformed isolated civilizations into a complex global machine where every country operated as a cog in the machine, gaining economic gains by performing its specialized role. Thanks for reading Manocha Opinion! Subscribe for free to receive new posts and support my work. Even though it is quite wishful to think that this has always been the case, this architecture is barely a century old. It is now being faced with its most adverse test. The decay started much earlier, crystallizing during the 2008 Financial Crisis and the weaponization of the global financial architecture, such as the expulsion of major powers from SWIFT but it has accelerated quite fast with the AI race and the Trump era. The world is again losing control of the 'rules of the game' and going back towards a mercantilist world, devolving towards a zero-sum game. The US is increasingly weaponizing its dominance in high tech supply chains, using export controls and tariffs as national security tools. While China is trying to benefit off its overwhelming control over manufacturing. Even though, I would agree that concept of free market is great, but only in an ideal world. Real world global trade is dictated by elite political incentives and national security strategies. Therefore, pitching for free market when your competitors are heavily subsidising their production would be quite unrealistic. But the current political response by Trump is quite shortsighted. He has used broad based tariff responses. There is a lot of literature on how besides being stupendously useless, it also ends up harming the domestic consumer and producer. An undergrad economic student can tell you that it will raise prices for domestic consumers, and reduce import demand, causing the domestic currency to appreciate. This currency strengthening makes a nation's own exports more expensive globally, undermining the very industrial competitiveness the tariffs were meant to protect. They have also imposed quite strong export controls to starve China off AI chips. But again, an undergrad economics student would tell you that by cutting off access to Western frontier hardware, the US inadvertently provides China's domestic market, such as Huawei's Ascend AI platforms and SMIC's foundry operations, with a guaranteed local customer base, removing the market pressures that once kept them dependent on American IP. I am sure that the US government is also aware about the shallowness of this economic statecraft. Therefore, I would rather argue that this is mostly theatrical. This posturing is less about technological containment and more about domestic political messaging and signaling to allies. They are just trying to delay the inevitable using whatever that tools they might have inhand. They are now trying a new trick, something that escalates this into an explicit, multi bloc technology conflict, Cold War style. A draft US State Department cable reveals Washington's plan to issue an ultimatum to 35 nations:choose between alignment with the USledPax Silicaalliance orparticipation in China'sWorld Artificial Intelligence Cooperation Organization(WAICO). Dual alignment is framed as incompatible with trusted partner status. It all started with Kazakhstan, quite an irrelevant country by optics, but a mineral rich state and a member ofPax Silica, which decided to sign onto WAICO as well earlier this June. This triggered alarm in Washington over supply chain leakage and the viability of non binding alignment. For context,Pax Silicaaggregates 25 core declaration signatories and a dozen more partners. The US strategy is to use access to advanced compute, semiconductor packaging, and AI infrastructure as leverage to force allies into harmonized export controls and mineral supply chain agreements designed to restrict China's technological development. This is of course going to fail and the US is probably aware of it. It's about imposing as much cost as you can on your enemy, while taking a cost yourself. The question comes till how long can the US economy bear the cost itself? Additionally, there are a lot of loopholes around this - shell firms, third-country trade rerouting, massively globalised procurement networks, etc. Coercing neutral third parties with binary 'all or nothing' mandates would merely drives these nations toward circumvention rather than true strategic alignment. So to enforce compliance, Washington would increasingly rely on extraterritorial mandates and secondary sanctions, threaten to cut off foreign banks, foundries, and enterprises from Western capital if they interface with non compliant Chinese stacks. For non aligned states, the primary vulnerability would be losing access to silicon and domestic software applications hardcoded to proprietary foreign hardware interfaces (such as Nvidia's CUDA). Yet, relying purely on legal coercion exposes Western policy to rapid decoupling. Faced with Washington's attempt to erect an absolute wall around frontier hardware, China is already executing a counter strategy built onAsymmetric Offsetting and Weaponized Interdependence, turning Washington's economic pressure into the catalyst for a de-Americanized global tech stack. As previously mentioned, through WAICO, China is flooding the Global South with open weight models. By offering highly capable models like Alibaba's Qwen family, DeepSeek, and Moonshot's Kimi series under permissive licenses, Beijing is drastically undercutting the switching costs for developing nations. While the United States demands that foreign governments sign audited agreements for expensive cloud subscriptions, China is handing them the code. Developing states can download, self host, and fine tune these models on local hardware without sending telemetry to Silicon Valley or risking a remote kill switch. Why would you give it for free? Beijing captures the application and intelligence layer across emerging markets by making Chinese software the global default infrastructure. It's like Microsoft giving away Windows for free. China offers open weight models today to undercut Western software, but once foreign ecosystems are dependent on Chinese frameworks, Beijing can mandate hardware telemetry, export restrictions, or licensing controls. In addition to this, Beijing is also weaponizing its position at the bottom of the value chain. China maintains an overwhelming operational monopoly over the refining of critical upstream minerals like gallium, germanium, and rare earth elements. Every time Washington expands its hardware bans, China leverages these material chokeholds to impose supply volatility and soaring manufacturing costs on Western hardware pipelines. And China is drastically closing the gap in lithography as well. China is producing lithography machines which ASML did in2004. This may seem like they are lagging a lot, but remember that innovation happens in spurts and they already have ASML to see and understand what could be done. They are a master at reverse engineering things as well. To complete the trap, China should and probably will monopolize the legacy node market. Most of the world's electronics still rely on mature semiconductor nodes. By flooding global markets with cheap, domestic mature node silicon, China locks intermediate manufacturing hubs across Southeast Asia, Central Asia, and Latin America into deeper physical dependence on Beijing. (I think this is one important gap that emerging semiconductor manufacturers could fill in - India and South East Asia hopefully). This seems like one big lecture on game theory, we could probably forecast a lot things that are going to happen and how badly it will leave all of us but we will still go down this path, until US and China reach a consensus or one of them loses the race. I am slightly optimistic on this front. I think they will come to their senses that there are no winners or losers in the highly globalised world. The thing I am highly pessimistic about is the timeline on this. I think it will take a lot of time for things to cool down especially with Trump and Xi in power. The world will hit a rock bottom before it goes back up again. I keep telling myself that we are still doing far better as a world than any world our ancestors lived in (except if climate change and tech billionaires make the world go bananas). Anyways, we still have a lot of years for this fight to be over. So an important question arises on how a country, which is not the US or China, wiggle their way out of here? Subscribe now This might be a good opportunity for a lot of nations as well since they might extract better deals by playing the both sides but it requires power to negotiation and the skills to get a good deal as well. To navigate this bipolar fragmentation, mainstream analysts usually argue that all countries face an inescapabletrilemma. In this framework, non aligned states must juggle three inherently conflicting national imperatives: Access to Frontier Hardware (Goal A):Securing cutting edge Western GPUs and capital equipment, which requires accepting strict US export control norms and embedding withinPax Silicastructures. Sovereign Control (Goal B):Running local, non audited models on domestic infrastructure without foreign telemetry, telemetry leaks, or remote kill switches. Strategic Autonomy (Goal C):Maintaining an independent foreign policy by refusing extraterritorial sanctions or binary 'all or nothing' alliances. Under conventional logic, Goals A and C are mutually exclusive. Choosing Western frontier compute forces a nation into Washington's orbit, while asserting strategic autonomy risks hardware starvation. The prevailing consensus in policy circles relies on a fundamentally flawed premise:Hardware Determinism. This view assumes that because the United States controls frontier GPU design and lithography IP, and because China controls critical mineral refining and mature nodes, the rest of the world must inevitably fall into one of two digital spheres of influence. This binary logic traps rest of the countries in a false dilemma designed by superpowers to force early compliance. This 'Frontier Anxiety' is built on a misunderstanding of how technology diffuses, and it traps developing nations in a game designed by superpowers. Leave a comment Superpowers intentionally push specific geopolitical narratives to preserve their hegemony. SinceThe United Statesholds a (temporary) lead in frontier AI models, specialized hardware IP (like NVIDIA GPUs), lithography tools, it pushes for'Arms Race'or'Nuclear Non Proliferation'narrative. Washington implies that AI progress is a dangerous, winner-take-all sprint. This framing suggests that falling behind on frontier hardware compute leads to national subordination. It justifies tight export controls, restricted hardware sales, and alliances (likePax Silica) that keep foreign states dependent on US audited cloud providers and licensed silicon. China's competitive advantage lies at the physical base of the supply chain, controlling the refining of critical minerals, dominating mature node semiconductor manufacturing, and producing low cost hardware inputs. Therefore,by emphasizing'Weaponized Interdependence'and'Material Chokepoints,'Beijing signals to the world that high tech Western stacks remain vulnerable to raw material cutoffs. Simultaneously, China offers an alternative through open weight models (like Qwen or DeepSeek) and initiative coalitions (like WAICO). This frames Western technology as locked down and expensive, while positioning Chinese offerings as accessible, sovereign alternatives for the Global South. But, like previously noted, once foreign ecosystems are dependent on Chinese frameworks, Beijing can mandate hardware telemetry, export restrictions, or licensing controls. I am sure almost all nations are cognisant of this fact. Whichever nation defines the rules, metrics, and language of a technological debate controls the policy outcomes. By convincing the rest of the world to adopt their preferred framing, superpowers make their national strengths look like global necessities and obscure their vulnerabilities. Middle powers must explicitly reject these frameworks. Accepting the arms race frame guarantees subordination, as it suggests that falling behind on frontier training compute leads to permanent geopolitical irrelevance. Instead, developing economies and middle powers must adopt theGeneral Purpose Technology (GPT) Diffusionframe, suggested byPranay Kotasthane. Treating frontier foundation models like nuclear warheads misjudges the economic nature of AI. AI is a general purpose technology, like steam engines or electricity. Historical precedents show that technological originators rarely monopolize value. Europe invented the automobile, but the US deployed the assembly line. In the end, no single nation 'won' automotive technology. But the country which is able to adapt the technology to real world enterprises, supply chains, and human capital wins massive economic gains from it. Just like how India did with the IT industry. It didn't invent SAP or cloud computing but still made billions off it. To translate this into sovereign reality, non aligned states must systematically solve three foundational bottlenecks. Firstly, a nation that hardcodes its domestic economy to proprietary foreign hardware and software remains perpetually vulnerable to extraterritorial export controls and sudden foreign bans. Nations must urgently decouple software execution from physical hardware. The ultimate anti chokepoint for software at least are open source compiler and software tools. Hardware is tricky because it would require developing actual knowledge and production bases themselves or in friendly states. The way to go about it is of course creating a union that acts together and pushes against any particular nations' hegemony. This is the most difficult task in my opinion. While there will be push for global open source architecture like RISC-V, almost countries will try to gain some sort of a manufacturing or knowledge base in their own country so no one bullies them. This is rational and an expected outcome. Secondly, running localized AI at scale is fundamentally an energy and grid capacity problem. All the countries face the bottleneck of securing clean, and continuous power. States that build resilient, green energy grids and secure physical data centre infrastructure hold geopolitical leverage of turning physical energy abundance into a sovereign magnet for global compute. Data centres are especially tricky due to their image on social media and among people in general. Each government will have to work hard on changing the narrative on them. Lastly,attempting to outspend Western or Chinese capital pools on frontier GPU training clusters is fool's mistake that would drain domestic resources without conferring strategic autonomy. The core challenge for all the nations is not generating foundational intelligence from scratch, but deploying intelligence locally at a fraction of the cost. Rather than racing for top-tier training benchmarks, countries need to solve the deployment equation, which could be done by codeveloping low cost, application specific integrated circuits (ASICs) optimized strictly for local inference and edge tasks. By combining these specialized edge chips with powerful open weight base models fine tuned on domestic datasets, non aligned nations could build tailored ecosystems that foreign tech giants cannot compete with, securing actual operational sovereignty without burning billions on frontier compute. Thanks for reading Manocha Opinion! This post is public so feel free to share it. Share