Meta and Nvidia highlight the increasing complexity of global AI politics

The New Currency in Tech: Permission

Artificial intelligence is now shaped by shifting regulations. While innovation remains crucial, the real challenge is adapting to ever-evolving definitions of what’s “permitted”—whether in hardware, business deals, data handling, or deployment. By 2026, the most prized asset in technology won’t be a cutting-edge model or a powerful GPU, but rather the authorization to operate—a privilege that can vanish between placing an order and receiving a shipment.

Regulatory Risk Moves to the Forefront

What was once considered routine business is now fraught with regulatory risk. For example, Nvidia now requires Chinese buyers of its H200 AI chips to pay the full amount upfront, with no option for refunds, cancellations, or last-minute changes. In some cases, collateral or insurance may substitute for cash. These payment terms reflect a new reality: regulatory uncertainty is now a cost of doing business, and payment is demanded before products even leave the warehouse.

Regulatory scrutiny isn’t limited to hardware. China’s Ministry of Commerce announced it would review Meta’s acquisition of AI agent developer Manus, citing compliance with foreign investment, technology export, and cross-border data transfer rules. Manus, based in Singapore but with Chinese origins, reportedly generates over $100 million in annual recurring revenue—making it both significant and a useful test case for regulators.

The underlying business landscape is growing more complex, with unpredictable regulatory shifts disrupting executive plans. Questions like “Can you sell U.S. chips and actually deliver them?” or “Can you acquire a company with Chinese ties and retain it?” are in constant flux. AI firms must now compete not only on speed but also on resilience—how well their strategies withstand regulatory upheaval.

Chips: From Commodity to Compliance Challenge

If semiconductors were still treated as ordinary products, Nvidia wouldn’t need to operate as if it were selling restricted goods. The H200 chip trade is now entangled in political calculations and supply constraints.

Although the dispute between the U.S. and China over Nvidia chips is ostensibly about technology, it has evolved into a permissions-based system with inventory implications. Nvidia CEO Jensen Huang recently noted that demand from Chinese customers for H200 chips is “very high,” prompting the company to ramp up its supply chain. However, Chinese authorities have reportedly instructed domestic tech firms to pause new orders while they determine the appropriate balance between domestic and imported chips. Chinese companies have placed orders for over 1.2 million H200 chips, each costing around $27,000, but Nvidia’s available stock is closer to 700,000 units. This is a clear example of government policy directly shaping market demand and the technological landscape within China’s data centers.

The Cost of Regulatory Shifts

Nvidia’s insistence on upfront payment is more understandable in light of the risks: the company recently wrote off $5.5 billion due to a U.S. export ban. With multiple export restrictions to navigate, misjudging regulatory winds can leave companies with unsellable inventory. In traditional markets, a misstep with customers can be corrected; in the AI sector, misreading regulators can be far more costly.

China is also tightening its grip on the AI supply chain. Recent policies have shifted the focus from “which chips can be purchased” to “which chips are allowed in national infrastructure.” In late 2024, Beijing banned foreign AI chips from state-funded data center projects, forcing early-stage initiatives to abandon or revise their plans. Nvidia’s market share in China’s AI chip sector plummeted from 95% in 2022 to zero by 2025, underscoring the real-world impact of these regulatory changes.

As Huang commented at CES, “If the purchase orders come, it’s because they’re able to place purchase orders.” Nvidia reportedly told Chinese clients it aimed to begin shipping H200 chips before the Lunar New Year, pending Beijing’s approval.

Regulatory dynamics in the U.S. are also in flux. In December, former President Donald Trump announced that Nvidia could export H200 chips to China with a 25% tariff and Commerce Department oversight. Yet, this was quickly followed by a review of advanced chip sales, illustrating how quickly permissions can be redefined. The regulatory environment is never static; it’s always subject to change.

Deals as Vectors for Technology Transfer

As chips become strategic assets, so does corporate ownership. China’s Commerce Ministry is scrutinizing Meta’s acquisition of Manus through the lens of foreign investment, technology export, and data sovereignty. Manus’s Singaporean registration and Chinese roots highlight the complexities of modern corporate structures, but Beijing is signaling that legal maneuvering is no longer sufficient protection.

Meta has stated that there will be no ongoing Chinese ownership after the acquisition and that Manus will cease operations in China. Even if this satisfies both U.S. and Chinese authorities, it raises deeper questions about what constitutes an “export” when the asset is not just software, but a team, a system, and the expertise to build AI agents capable of real-world impact.

This trend suggests that AI acquisitions are increasingly viewed as transfers of capability rather than standard mergers and acquisitions—transactions that governments may want to scrutinize, delay, or block. Europe is following suit, with the EU recently strengthening its foreign direct investment screening. By February 10, regulators will decide whether to approve Alphabet’s $32 billion purchase of cybersecurity firm Wiz or launch a more thorough investigation, highlighting that the “AI stack” includes critical security layers that underpin trust in cloud infrastructure.

Data: The New Borderline

The most significant regulatory changes rarely arrive as outright bans. Instead, they come as new definitions, frameworks, and timelines that force companies to adapt today in anticipation of tomorrow’s constraints. The European Commission’s AI Act is a case in point: bans on certain practices and requirements for AI literacy take effect in February 2025, obligations for general-purpose models begin in August 2025, and the Act is fully enforceable by August 2026, with some provisions extending to 2027. When companies requested a delay, the Commission made it clear there would be no pause—demonstrating the EU’s relentless approach to regulation.

Brussels is also examining the infrastructure supporting AI. In November, the European Commission launched investigations under the Digital Markets Act to determine if AWS and Microsoft Azure should be classified as gatekeepers in cloud computing. If cloud services become regulatory chokepoints, AI oversight will extend beyond model behavior to encompass the infrastructure, data, and distribution networks that support deployment and training.

In the U.S., the regulatory uncertainty centers on jurisdiction. A December executive order from Trump directed federal agencies to review state AI laws that might conflict with national policy, including those that could mandate disclosures or alter system outputs in ways the administration argues are unconstitutional. This effort to standardize compliance across states introduces its own volatility for companies navigating a shifting federal landscape.

Financial regulations are also tightening. The Treasury Department’s outbound investment program now restricts and requires notification for certain U.S. investments in sensitive technologies, including AI, in countries of concern. The question “Can you fund it?” is now intertwined with “Can you ship it?” and “Can you own it?”—all governed by the same permissions logic, just through different mechanisms.

Adapting to a Rule-Change Economy

AI is increasingly defined by regulatory shifts, as policymakers recognize they are shaping the next era of global power. Compliance has become a strategic imperative, contracts serve as risk hedges, and corporate structures are now entangled with geopolitics. Success will not only belong to those with superior technology, but to those who can continually adapt and thrive amid ongoing regulatory transformation—offering answers to “Can you?” that go beyond uncertainty.