The Trump administration is pouring $46 billion into what it calls a "smart wall" along the U.S.-Mexico border, and what is going into the ground right now is less a wall than a 2,000-mile autonomous surveillance array. Construction crews are adding steel at roughly six miles a week, with 74 miles of new 30-foot fencing installed since mid-June, while the crossings the build was sold to stop have fallen to their lowest level in decades.
"It's a smart wall. It's not just a barrier," U.S. Customs and Border Protection Commissioner Rodney Scott told The Associated Press. The administration's framing is that steel, sensors, cameras, and towers together let a smaller Border Patrol workforce cover more ground than a chain-link fence and a Humvee ever could. Scott, in the same AP interview, said the system "maximizes the use of our most valuable resource, which is our agents."
That framing is now the central political question of the project. Civil-liberties groups argue that what looks like efficiency is in fact a permanent militarization of the borderland. "What we are seeing is a massive expansion of surveillance and surveillance technology across the borderlands," said Ricky Garza of the Southern Border Communities Coalition, in comments reported by AP. A wall, he added, "in all its forms is harmful to communities."
The hardware side of the build has moved quickly into a recognizable industrial program. The Department of Homeland Security announced five new Smart Wall contracts in Texas and Arizona, and CBP's Autonomous Surveillance Towers were formally declared a Program of Record along the Southwest Border, a procurement status that signals the agency intends to keep buying and fielding the systems for years rather than treat them as a one-off pilot. General Dynamics IT is now operating AI-enabled autonomous surveillance towers on the southern border, with machines rather than agents making first-pass detection calls, and DHS is recompeting a roughly $100 million contract for integrated surveillance towers that will define the next layer of the grid.
The largest single award so far has gone to Anduril, the defense-tech firm founded by Palmer Luckey. The company won a $363 million CBP contract for more than 200 Extended Range Sentry towers, mast-mounted sensor pods that fuse radar, electro-optical cameras, and onboard AI to detect and classify people and vehicles at long range, then cue agents to the spot. It is a different business model from the old fence-first border. Instead of a continuous physical obstacle, the country is laying down a continuous sensing fabric and treating steel as the visible backstop, not the spine.
That is where the timing paradox becomes politically uncomfortable. Border encounters are at their lowest in decades by every public measure CBP has released. The administration's argument is essentially that this is the moment to harden the line while operational pressure is light, and that the same sensors and towers that look wasteful in a quiet month are what makes a future surge manageable without rebuilding headcount. Critics hear the same data and ask the inverse: if the crossings have already collapsed, what is the $46 billion actually buying, and for how long?
About 535 miles of the roughly 2,000-mile border will remain unwalled no matter how fast crews work, because of mountains, rivers, and protected terrain. The build, in other words, can never be a continuous wall. It will be a continuous grid, with steel in the places steel can stand and sensors in the rest.
The open oversight question is who watches the watchers. CBP's surveillance towers, by design, sweep broad bands of borderland that include ranches, crossings, and small communities miles from any fence line. The contracts now being signed, from the Anduril sentries to the recompeted integrated tower award, will shape the southern border for years to come, and the rules for what their data can be used for, how long it is kept, and which agencies can query it are not part of the public procurement record. That is the trade-off the country is making in the dirt this summer: more steel, more sensors, fewer agents on the line, and a sensor grid whose own perimeter of authority has not been drawn.