AI, radar and drones: the US wall costs $46.5 billion, but fentanyl comes through the legal door

By Zendoric · July 27, 2026.

The U.S. Department of Homeland Security (DHS) has earmarked $46.5 billion, approved in the July 2025 budget legislation, to complete and modernize the border wall with Mexico, as the agency itself confirmed. The money doesn't just fund steel: it also pays for solar-powered autonomous towers, radar, thermal cameras, buried vibration sensors, patrol drones and artificial intelligence systems that classify in real time whether what is moving is a person, a vehicle or an animal, before alerting a human agent. By February 2026, Washington had already signed $4.5 billion in contracts for some 370 kilometers of that infrastructure, according to what the U.S. government itself has reported.

The stated goal is not to seal the border but to concentrate agents there: "it's more than a barrier, it maximizes the use of our most valuable resource, which is our agents," summed up CBP Commissioner Rodney Scott, in remarks reported by El Universal. On paper it's a sensible idea: computer-vision AI does not replace the border agent, it multiplies his reach by automatically filtering thousands of hours of sensor feeds that no human team could review in real time.

The problem is what the official data themselves show. CBP seized close to 8,890 kilograms of fentanyl in fiscal year 2024 —a figure for national seizures, not just border crossings— and a good share of those seizures take place at official ports of entry, not in the desert watched by sensors: there the drug travels mixed in among the millions of legal crossings by workers, tourists and freight trucks that no system can inspect one by one. The U.S. Sentencing Commission adds another uncomfortable data point: 86.4% of those convicted of federal fentanyl trafficking in fiscal year 2023 were U.S. citizens, people whom the organizations can hire or deceive into crossing through legal checkpoints.

Security analyst Jaime Ortiz, consulted by El Universal, describes a criminal organization that adapts with the same logic as any adversarial system: it splits large shipments into small ones, switches checkpoints, hides drugs in private vehicles or legal freight trucks, opens sea or air routes, digs tunnels, changes chemical suppliers or bribes a port worker. "To have any chance of shutting down an entire criminal organization, you would have to simultaneously control land crossings, checkpoints, the subsoil, airspace, coastlines, transport companies, financial networks, the supply of chemical substances and, of course, detect and attack corruption," Ortiz sums up. No radar tower covers that entire list.

Even the surveillance technology itself creates new risks that have to be managed. In February 2026 the Federal Aviation Administration (FAA) temporarily closed El Paso's airspace after an anti-drone laser system was activated near Fort Bliss without proper interagency coordination, according to sources cited by Associated Press. An episode that recalls something we already see in cybersecurity: deploying detection capability without clear governance over when and how to use it can create as much risk as not having it.

Here it is worth applying the same standard we use to judge any deployment of AI in security: separating demonstrated capability from the narrative. The smart wall demonstrates something real and valuable —automatic classification of imagery and radar at a scale no human team could match— but that technical achievement solves only the stretch of the problem that happens in the desert between checkpoints. The U.S. Treasury itself sanctioned Mexican companies and executives in October 2025 accused —without that amounting to a court conviction— of allegedly supplying chemical precursors to Los Chapitos, a faction of the Sinaloa Cartel; and even so, blocking a company does not dismantle the network, which can keep operating under other corporate names, front men or new suppliers.

Our reading: the pattern this case leaves is the same one we have been pointing to when discussing cybersecurity or models that "match X" — a system's power is not measured by its flashiest component, but by whether it closes the link that really matters. Here that link is not the desert, it is the official port of entry, the shipping line, the bank account and the bribable official, and there artificial intelligence is far less deployed than in the surveillance towers. The investment with the most upside in the medium term is not more steel or more cameras in the hills, but AI applied to detecting anomalous patterns in cargo manifests, financial flows and precursor supply chains —the same kind of analytics banks already use against money laundering and which, if scaled here, would shift human work toward risk analysis and investigation, not toward watching a monitor.

In the long run, that is also the trajectory that interests us most as an underlying thesis: the more capacity society has for AI to monitor, audit and modulate flows —of goods, of money, of chemicals— with transparency and democratic oversight, the less room there will be for financial opacity and corruption to remain the crack that no wall, however high, can seal. But that requires governance, judicial cooperation and sustained political will, not just a budget for sensors; and that part, unlike the technology, is not bought in $4.5 billion contracts.