Silicon Shadows: How Black Markets Help AI Chips Reach China Despite US Export Controls

The story of advanced AI chips moving across borders is no longer just about factories, pricing, and demand. Increasingly, it’s about paperwork, classification disputes, shipping routes, and the gray zones that appear when governments try to slow the flow of cutting-edge technology. A new Financial Times Film report, “Silicon shadows: inside the black market for AI chips,” spotlights a reality that export-control policy often struggles to fully anticipate: when restrictions tighten, incentives don’t disappear—they migrate. In this case, they migrate into intermediaries, rerouting strategies, and opaque trading channels that can blur the line between legal supply chains and illicit transfers.

At the center of the report is a familiar policy goal. The United States has tightened export controls on advanced AI semiconductors, aiming to limit their reach to China and thereby constrain the pace at which certain high-end AI capabilities can be developed. The logic is straightforward: if the most powerful chips are harder to obtain, then training runs become more expensive, timelines lengthen, and some applications may be delayed. But the report argues that the real-world effect is more complicated than a simple “less supply equals less capability.” Instead, the tightening of official pathways can create a parallel market—one that thrives precisely because it offers what the formal system withholds.

What makes this story particularly consequential is not only the existence of a black market, but the way it reshapes the ecosystem around the chips themselves. When enforcement becomes the bottleneck, the supply chain becomes a battlefield. And when the battlefield shifts, so do the tactics of everyone involved: legitimate distributors who try to stay compliant, brokers who claim they are “facilitating trade,” and illicit networks that treat compliance language as something to navigate rather than obey.

The report describes how export controls can generate demand for workarounds. That demand doesn’t necessarily come from a single actor or a single scheme. It can be distributed across many participants—some knowingly, others at arm’s length—each playing a role in keeping transactions moving while reducing the visibility of end users and end use. In practice, the difference between a lawful transaction and an unlawful one can hinge on details that are easy to obscure: who ultimately receives the chips, where they are installed, how they are labeled, and whether documentation matches the physical reality.

One of the most important insights in the reporting is that export controls don’t just reduce availability; they change the economics of procurement. If a chip becomes scarce through official channels, its value rises—not only in terms of performance, but in terms of leverage. Scarcity creates room for markups, favors, and risk premiums. It also creates a market for information: knowledge about which shipments are likely to be scrutinized, which intermediaries have reputations for “getting things done,” and which routes are less likely to trigger delays or investigations.

In other words, the black market isn’t merely a shadow version of the supply chain. It is a response to the friction introduced by policy. Where there is friction, there is profit for those willing to absorb risk. And where there is profit, networks form.

The report’s framing also highlights a key enforcement challenge: export controls are designed to regulate specific categories of technology, but trade is a living system. Chips can be repackaged, reclassified, or routed through multiple jurisdictions. Even when the underlying intent is to comply, the complexity of global logistics can make it difficult to verify every step. Illicit actors exploit that complexity by focusing on ambiguity—using layers of intermediaries, shifting paperwork, and sometimes relying on the fact that different countries interpret and implement rules differently.

This is where the “thriving” aspect of the black market becomes more than a sensational claim. A market can only thrive if it can reliably deliver goods. That reliability can come from specialization: networks that understand the technical requirements of the chips, the practicalities of shipping, and the documentation patterns that tend to pass through. It can also come from relationships—between traders, freight forwarders, and local buyers—where trust is built over repeated transactions. In such systems, the black market becomes less like a one-off smuggling operation and more like a parallel commercial infrastructure.

The report points to intermediaries and opaque trading channels as central mechanisms. These intermediaries may present themselves as brokers, resellers, or logistics facilitators. Their role can range from sourcing chips from surplus inventories to arranging cross-border transfers that are difficult to trace. Some may operate in ways that are technically defensible while still enabling outcomes that the original export-control policy was meant to prevent. Others may be openly illicit. The difficulty for enforcement is that the boundary between those categories can be hard to draw from the outside.

There is also a broader strategic dimension. Export controls are often justified as a tool to slow down technological progress. But technological progress is not only about access to hardware; it’s also about the ability to iterate, test, and scale. If chips are diverted through unofficial channels, then the policy’s intended delay may shrink. Even partial access can matter, especially in a field where training cycles, experimentation, and deployment all depend on compute availability. A small number of high-end chips can still accelerate learning curves, support benchmarking, and enable experiments that would otherwise be postponed.

That said, the report does not imply that export controls are meaningless. Rather, it suggests that they may produce uneven effects. Official supply may become more constrained, but unofficial supply can fill some of the gap. The result can be a patchwork: some organizations may face shortages, while others find ways to obtain enough compute to keep momentum. This unevenness can distort competition and shift power toward those with better procurement networks—networks that may be embedded in the very regions and industries that export controls aim to influence.

Another unique angle in the reporting is the policy feedback loop. Governments tighten controls to reduce access. Markets respond by finding routes around them. Enforcement then faces new challenges: more complex transactions, more hidden end users, and more sophisticated attempts to conceal intent. Over time, the policy can become a game of adaptation, where each tightening triggers new workarounds. That dynamic can lead to a cycle that is costly for regulators and disruptive for legitimate businesses.

For companies trying to comply, the environment becomes harder to navigate. Compliance teams must assess not only the product and its specifications, but also the customer, the end use, and the chain of custody. When black markets expand, the risk of inadvertent involvement increases. A legitimate distributor might sell to a reseller that later diverts the chips. Or a shipment might be routed through a partner whose practices are unclear. Even when a company acts in good faith, the presence of illicit networks can raise the probability of being caught in the middle.

This is one reason the report’s implications extend beyond geopolitics and into the day-to-day functioning of the semiconductor ecosystem. Semiconductors are already complex products to move: they require careful handling, specialized packaging, and precise documentation. When export controls add another layer of scrutiny, the cost of compliance rises. That cost can discourage smaller players from participating in legitimate trade, concentrating activity among those with the resources to manage risk. Meanwhile, illicit networks can attract participants who are willing to operate without those overheads.

The global semiconductor ecosystem also depends on predictability. Manufacturers and distributors plan production and inventory based on expected demand and stable supply routes. If the market becomes more volatile—because official channels are constrained and unofficial channels are unpredictable—then planning becomes harder. That can affect lead times, pricing, and even the willingness of firms to invest in certain markets. In the long run, the policy may not only shift chips across borders; it may shift investment decisions and supply-chain strategies.

There is also the question of what “black market” means in practice. It can include outright smuggling, but it can also include transactions that are technically legal in one jurisdiction while undermining the spirit of the control regime. It can involve misrepresentation of end users, diversion after purchase, or the use of intermediaries that obscure the final destination. The report’s emphasis on opaque channels underscores that the problem is not always a dramatic border crossing. Sometimes it is a series of ordinary commercial steps that, when combined, produce an outcome that violates the policy’s intent.

This is why the enforcement challenge is so persistent. Export controls are designed to be targeted, but trade networks are adaptive. They can route around constraints by exploiting differences in oversight capacity, regulatory interpretation, and the speed at which information travels. Even when authorities share intelligence, the sheer volume of global commerce makes it difficult to inspect every shipment. As a result, enforcement often relies on risk-based targeting—focusing on transactions that appear suspicious. Black market networks, in turn, learn what triggers suspicion and adjust accordingly.

The report’s narrative also invites a deeper look at incentives. Why would anyone take the risk? Because the chips are valuable, and because the demand for advanced AI compute is not theoretical. Organizations in China and elsewhere are building models, deploying AI systems, and competing in markets where compute translates into capability. If official access is restricted, the demand doesn’t vanish. It becomes more expensive and more politically sensitive. That combination—high demand plus restricted supply—creates a strong incentive for intermediaries to offer “solutions.”

But the “solutions” come with costs. Illicit procurement can be unreliable. Chips may arrive late, in incomplete batches, or with uncertain provenance. There can be risks of counterfeit components, mismatched specifications, or products that are not supported by the intended software ecosystem. Even if the chips function, the lack of transparency can complicate maintenance, warranty claims, and long-term planning. For buyers, the black market can be a bridge—but bridges are not infrastructure.

Still, in fast-moving AI development cycles, a bridge can be enough. If a buyer needs compute for a critical training run or a time-sensitive deployment, the ability to obtain chips—even through risky channels—can outweigh concerns about provenance. That is part of what makes the black market resilient: it meets urgent needs.

The report also implicitly raises a question about the limits of export controls as a tool for shaping technological