When a major AI laboratory releases a new flagship model, the international reaction tends to follow a predictable script. Analysts rush to compare benchmark scores, while investors and policymakers debate what the technology signals about national capability. This script played out again when Moonshot AI, a leading Beijing-based AI startup, unveiled its latest model, Kimi 3. Almost overnight, the model became a global talking point, framed as evidence of China’s rapidly advancing frontier AI capabilities. Curious to see what the excitement was about, I decided to try Kimi myself. Instead of exploring its frontier capabilities, I found myself faced with subscription requirements, a priority queue with no guarantee of access, and eventually server overload messages even after switching to an earlier version. It was difficult not to smile at the irony. One of the most talked-about AI models in the world was largely unavailable to many of the people eager to evaluate it. This experience led me to a conclusion that the true significance of Kimi 3 will not be recorded in benchmark tables, but as the point where global AI competition expanded from a pure engineering sprint into a complex struggle over physical access, compute infrastructure, and digital sovereignty. The first phase of generative AI centred on which entity could build the most capable model. The next phase centres on who can host, finance, and control these systems. A couple of years ago, a new model release was a technical milestone discussed primarily in research papers. Today, it gets treated like a geopolitical event. Financial markets react before independent evaluations are completed. Governments assess national competitiveness implications. Competitors adjust their API pricing structures. Social media platforms flood with selective benchmark charts, while policymakers begin drafting briefs on supply chains and security risks. Kimi 3 entered an environment already deeply attentive to Chinese technological momentum. Last year, DeepSeek proved that smart engineering could bypass massive compute requirements. Kimi 3 brought a different reality into focus: the sheer friction of serving a frontier model to a global audience. Every frontier model is eventually surpassed. What remains is the fact that AI launches are no longer standard software updates. They serve as direct signals of industrial capacity, state power, and geopolitical leverage.
Technical performance remains important, but high scores on static test suites no longer guarantee real-world influence. A model can perform exceptionally in a controlled sandbox, yet without massive physical infrastructure including fibre optics, cloud capacity, inference optimisation, and power grids, it remains confined to the laboratory. This fact often gets buried during launch week, when radar charts dominate public commentary. However, capability and availability represent two entirely different metrics. When a model captures global headlines yet stalls under routine user traffic, it reveals a fundamental constraint of modern AI. Access to compute infrastructure now dictates technological leadership just as much as model architecture. This creates a paradox for open-weight models. Open-source software frameworks are frequently championed as an equaliser, offering smaller nations and independent developers a way to escape vendor lock-in and achieve technological autonomy. While openness eliminates licensing fees, it does not alter the physical realities of hardware economics. Hosting a trillion-parameter model requires immense capital, energy, and specialised engineering expertise. By releasing model weights without accompanying access to affordable compute, the primary beneficiaries of open frontier releases are rarely independent developers or developing states. Instead, the benefits flow to major cloud providers, state-funded research centres, and large enterprises, as these are the only entities with the hardware required to run them. Permission to use a model is meaningless without the physical capacity to host it. These operational realities explain why state actors and international bodies do not simply procure whichever model tops the current leaderboard. When a government or international organisation evaluates an AI deployment, raw technical performance is often secondary to broader governance questions. The primary decision criteria include: In the public sector, governance requirements carry far more weight than raw speed or parameter counts. The most advanced model on paper is unusable if it introduces unacceptable political or regulatory risks. Model distillation is a technique where a smaller or newer AI model is trained using the outputs and logical step-by-step reasoning generated by a larger, more advanced AI system. Instead of learning directly from raw data, the student model learns by imitating the responses of the teacher model, allowing developers to replicate advanced capabilities at a fraction of the cost and compute. Kimi 3 has landed squarely in a diplomatic dispute over how its training was conducted. U.S. officials recently alleged that Moonshot AI relied on large-scale model distillation using outputs from Anthropic’s flagship models. Furthermore, regulators suspect Moonshot routed workloads through overseas entities to access restricted U.S.-designed cloud hardware. Moonshot AI has not accepted these claims, and many details remain disputed. Regardless of the facts in this specific case, the controversy exposes a rift in international technology law. For decades, reverse engineering was accepted as legitimate competitive research. AI breaks that consensus. When one entity trains a system by collecting millions of synthetic outputs from a competitor’s API, determining the boundary between competitive learning and intellectual property infringement becomes highly controversial. While distillation is standard practice for building lightweight, efficient tools, using it to rapidly replicate a foreign rival’s flagship capabilities pushes the technique into a diplomatic grey zone. Depending on how international regulators respond, this practice could trigger fresh supply chain sanctions, export bans, and trade disputes that outlast any single model’s lifecycle. The situation surrounding Kimi 3 highlights the growing limits of physical export controls. For several years, Western policy has focused on blocking the physical transfer of advanced microchips to rival states. Yet reports alleging that Moonshot accessed high-end Nvidia hardware through third-party cloud data centres in Thailand illustrate how easily geographic containment can be bypassed. Compute power no longer exists solely as physical hardware inside national borders. It operates as fluid, borderless cloud capacity that can be leased internationally. This creates an acute dilemma for neutral third-party states attempting to develop their own technology sectors. When cloud infrastructure flows freely across borders, hosting a commercial data centre risks triggering secondary sanctions. As a result, smaller nations may be forced to choose between international digital integration and strict compliance with global superpower mandates. Tracking hardware shipments is no longer sufficient. Regulators will eventually need to monitor software workloads in real time, which presents a much harder enforcement challenge. Whether Kimi 3 secures long-term market dominance is largely beside the point. Its release says more about the current state of global AI governance than about the model itself. The mechanisms previously relied upon to measure and control technology, including licensing frameworks, export bans, physical borders, and benchmark rankings, are losing their efficacy. A license does not guarantee usable access. A physical border does not contain compute power. A benchmark score does not guarantee real-world adoption. What replaces these mechanisms will not be decided solely in computer science departments. It will be determined through trade agreements, procurement annexes, and national energy policies. That is the true frontier of AI competition. Author: Slobodan KovrlijaAI launches have become geopolitical events
The infrastructure constraint: why openness is not access
Procurement as an instrument of sovereignty
Distillation and the grey zone of global governance
The limits of hardware containment
Beyond Kimi 3
