Pentagon Sets Up AutoWarCom as Robotaxi Scaling Hits an Infrastructure Wall

US Defense Secretary Pete Hegseth used his annual State of the Force address at Marine Corps Base Quantico on Wednesday to announce Autonomous Warfare Command, or AutoWarCom, a new four-star-led body that will develop and field drones, robots and autonomous systems across all branches. The Next Web reported the announcement on Thursday, noting the Pentagon has not stood up a new command since US Space Command in 2019.

Hegseth said AutoWarCom would carry powers similar to the armed services, letting it drive design, development and fielding of its own systems. He said it would stand up over several months and launch within a year, according to Air & Space Forces Magazine. Until then an interim measure called Project Agincourt will operate, and the services will create new job specialties for personnel who run autonomous systems.

Money is moving with the mandate. The Pentagon's fiscal 2027 budget proposal requests about $54.6 billion for the Defense Autonomous Warfare Group, up from roughly $226 million now, though Congress has not approved the request.

"The pace of war is changing faster than the process to support it," Hegseth said.

The command builds on a June decision to consolidate almost all unmanned programmes under a single drone office, including the Defense Autonomous Warfare Group and Joint Interagency Task Force 401, which focuses on countering enemy drones. The Next Web also notes that security experts in both the US and China have called for common rules on AI in weapons systems, including nuclear weapons.

Robotaxis: the demo was never the hard part

On the commercial side, the same week brought a blunt reminder that autonomy is a logistics business. Writing for The Robot Report on 26 September, Aseon Labs co-founder Dan Keene argued that the bottleneck for robotaxis is not driving software but what happens between rides. A robotaxi that finishes a trip with a spilled latte on the back seat may drive up to 15 miles empty to a central depot to be cleaned, charged and inspected, a reset that costs roughly two hours. California data shows the cost: of 86 million Waymo miles reported through 2025, only 54% carried a passenger.

Keene lays out three constraints. Industrial-zoned land with power is being outbid by data centre developers, so an operator entering a new city competes with the likes of Microsoft and Amazon. California's regulator targets six months just to energize a standard charging site; a new circuit runs 1.9 years, a substation upgrade 2.8 years and a new substation 8.9 years. And every city differs, from Phoenix sprawl to Zurich's medieval street plan to Riyadh heat and dust. Waymo, he notes, is in 14 cities while Uber is in more than 15,000 and did 3.9 billion trips last quarter. Keene discloses that his own company is building robotic servicing stations, and says the argument rests on arithmetic rather than on his product.

Two days later, on 28 September, The Robot Report published a second piece on the economics of machine autonomy. It reports that analysts at the AI Robotics Alliance of America Summit in June hypothesised capital itself as a form of robot governance, and that lawyers at Baker McKenzie say no European jurisdiction will accept robot economic autonomy in contract law. The article also cites the €1,000 to €1,500 price band most humanoid makers see as the mass market, and futurist John Anderson's push for crypto wallets for machines that cannot hold bank accounts.

Capital flows into construction autonomy

On 30 September The Robot Report reported that Autonomous Solutions Inc. and SoftBank Group announced a joint venture to commercialise brand-agnostic autonomous heavy equipment, with SoftBank providing $225 million in new funding to ASI. CEO Mel Torrie said skilled equipment operators are hard to find and coordinating mixed fleets across a large jobsite is a real bottleneck. The Mendon, Utah company, founded in 2000, sells its Mobius fleet orchestration system and says it works across mixed vehicle environments.

None of this settles the governance question. The Robot Report's 28 September analysis notes that in multi-segment neural models that reconfigure on the fly, distinguishing a malfunction from fatigue or low energy becomes difficult, raising the prospect of anthropomorphic arguments in court. The Pentagon's own announcement arrived against the same backdrop, with experts warning about military AI developing without limits.

The through-line is less about vehicles than about everything around them: depots, charging, land, standards and liability. Robotaxis can already drive. The industry's next few years will be decided by whether anyone builds the pit.