The global interconnected economy
getty
Silicon Valley, long considered the capital of the tech world, still has the funding, talent and frontier labs driving much of the AI industry. What it no longer has is the old assumption that global tech markets will depend on the Valley’s tech. As governments realize the critical nature and increasing dependency on AI, they are no longer letting the Valley shape the market. They are blocking deals, finding ways to keep talent and deciding which companies can own the most valuable AI assets.
China’s sudden move to block Meta’s reported $2 billion acquisition of Manus, an AI agent startup with Chinese roots and Singapore operations, is not just a narrow regulatory dispute. Reuters reported that China ordered Meta to unwind the Manus purchase on national security grounds, with authorities seeking to stop foreign access to Chinese AI talent and intellectual property. Cross-border AI is economically and politically charged, and the market is splitting into national and regional camps.
For decades, the global tech market ran on a Silicon Valley-centric premise that the best startups wanted U.S. capital, U.S. customers and, in many cases, a U.S. exit. AI is breaking that pattern. The Valley remains the richest and most powerful single spot for AI investment and development, but it no longer owns the default path for global commercialization. The ties are starting to loosen that have traditionally made Silicon Valley hard to unseat in its central leadership position.
China is building a parallel stack around domestic chips and models. Singapore has become a staging ground for AI talent and intellectual property. Europe, Canada and the Gulf are pushing sovereign AI as a matter of industrial survival. The map of innovation now looks less like a hub-and-spoke system and more like increasingly separate power systems that connect only where governments allow it.
AI companies usually think that their biggest risk is a better model, a cheaper chip or a rival with deeper pockets, but the Manus case is a sign that the largest risk to AI companies is a government’s ability to stop acquisitions, hiring, capital flows and expansion. Deals can be blocked, founders can be held in place and infrastructure choices can be forced by export controls. AI is now treated less like software and more like nuclear technology: strategic, scarce, dual-use and too powerful to leave fully to markets.
Reuters reported that China’s National Development and Reform Commission cited national security concerns and ordered Meta to unwind the Manus transaction, even after Manus had moved operations to Singapore. The scrutiny focused on Chinese-origin technology, Chinese AI talent and intellectual property that Beijing views as strategic. Reuters also reported that Manus co-founders had been summoned by regulators and barred from leaving China during the review.
Usually regulators are most concerned with ownership and integration related issues. But now governments are focusing on who can leave, who can build and who gets to carry tacit knowledge across borders. AI models sit on chips and data centers, but the real bottleneck still sits in researchers, founders, infrastructure engineers and product teams who know how to turn research into usable systems.
In the case of AI agents, that sort of knowledge has unusual leverage. Agents promise software that can plan, browse, write code, handle workflows and act on behalf of users. The companies that master that layer may own the interface through which millions of workers interact with AI. Meta’s interest in Manus fits that pattern, and Beijing has realized its strength in resisting.
Silicon Valley’s classic flywheel is simple and hard to unseat. Universities attract skilled talent, who create companies that attract yet more talent. The talent attracts venture capital and investments, which enable these companies to grow and attract more talent and resources. The tech companies become bigger, spinning off new companies and attracting even more labor, new money comes in to fund these companies, and the flywheel continues. The power of it all makes it seem like there’s no way to stop or even slow this flywheel.
Stanford’s 2025 AI Index found that U.S. private AI investment hit $109.1 billion in 2024, nearly twelve times China’s $9.3 billion and twenty-four times the U.K.’s $4.5 billion. U.S. institutions produced forty notable AI models in 2024, far ahead of China’s fifteen and Europe’s three.
However, AI is showing that things might not be that inevitable. Silicon Valley remains the strongest AI cluster, but it is losing the old privilege of being the automatic buyer, funder and organizer of global technical ambition. New models coming out of China, Europe and other places are keeping pace with the Valley in terms of performance and power. Governments are increasingly shifting their own investments to native AI companies, offsetting Silicon Valley’s funding dominance. And governments are increasingly flexing their muscles when it comes to keeping talent and technology inside their own borders.
Stanford’s AI Index captures the changes. The U.S. still leads in top models and investment, but China is closing the performance gap on major benchmarks. The report also notes that model development is becoming more global, with notable launches from regions such as the Middle East, Latin America and Southeast Asia.
This broadening of global competition applies to the hardware layer just as much as software. DeepSeek’s reported shift toward Huawei chips says that Chinese AI firms are being pushed toward a domestic infrastructure base, even when Nvidia remains the global benchmark. Manus shifted its headquarters to Singapore, a hub often used by startups seeking access to global capital, Asian talent and a more neutral regulatory posture. Singapore is positioning itself as a neutral AI hub that gives founders room to maneuver outside the tech superpowers.
Reuters described Singapore as an emerging neutral ground for AI firms navigating Sino-U.S. rivalry. Chinese startups have used the city-state to ease concerns from international customers and investors, and U.S. firms have been drawn by its business environment, visa system and IP rules. The same report warned that Singapore’s intermediary role could come under pressure as both superpowers tighten controls over technology transfer and talent flows. Beijing still treated the Manus deal as falling under its national security interests.
For years, founders in sensitive sectors used offshore structures to court foreign investors, list abroad or serve multinational customers. In AI, that structure may fail. Governments can assert control based on founder nationality, original research location, data provenance or perceived strategic value. This has direct implications for venture capital. A U.S. fund that backs a Chinese-origin AI startup in Singapore may now have to ask if the company could ever be sold to a U.S. buyer, given Beijing’s heavier hand in regulation.
AI is not the first technology to be dragged into geopolitics. Chips, telecom networks, social platforms, EV batteries, rare earth minerals, clean energy supply chains, app stores, cloud infrastructure and data flows are already being treated as strategic assets. AI sits on top of all of them, which makes it the sharpest expression of the shift in technology sovereignty. So the latest moves aren’t AI-specific, but AI is accelerating the shift.
In fact, the U.S. has been equally forceful in demands for technology and AI sovereignty. The U.S. has pressed controls on advanced chips and computing technology. Other countries and regions including India, Europe, the Gulf and Southeast Asia are pushing cloud regions, model developers and data center investments with their own political conditions.
AI makes questions of geographic positioning harder because it bundles every strategic anxiety into one market. It needs advanced chips, giant cloud systems, scarce talent, sensitive data, consumer interfaces and military-adjacent capabilities. Silicon Valley-centric technologies such as search engines, social networks, data center, enterprise tech and operating systems may also have been as critical to economies, but they don’t bundle all these concerns together at once. Governments have not been shy to exert control over these technologies either, especially social networks that challenge free speech controls, so it comes as little surprise they see AI as even more critical and strategic.
The AI industry has spent the last few years optimizing for more compute, larger contexts, better agents, faster inference and cheaper training. The next era requires a different discipline. Companies need to think geopolitically, with a plan for ownership, data, chips, talent and government review.
That means knowing which government stack they belong to, in addition to their technology stack. It means mapping ownership risk, founder mobility risk, data residency risk and chip dependency risk before a financing round or acquisition process starts. It means building product road maps that assume certain markets may close with little warning.
Some think that this is just regulatory friction, but that thinking would be a mistake. Friction is just about slowing adoption and finding ways to move faster in more bureaucratic regions. The issues here are more strategic than simply pushing paper. Governments are acting like gates. A state can open it, close it or force a company to walk backward through it.
For Meta, Manus is a painful case study in cross-border AI deal risk. For founders, it is a warning that relocation does not erase origin or national security concerns. For Silicon Valley, it is a message that capital and talent alone no longer set the rules of the game.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。