The Army is taking AI-enabled command and control to the Pacific. Whether it works will depend on hardware that survives the heat and keeps computing when the links go dark.
This summer, soldiers at Project Convergence Capstone 6 found a way to keep commercial gear working in the Mojave heat. They soaked T-shirts in ice water and draped them over antennas (Business Insider). That fix worked in an exercise. It will not work on a jungle ridgeline in the first island chain, and it will not work where there is no ice.
Our COO, Ryan Rebholz, wrote about that exercise in The Last Mile of AI Warfare Runs Hot. His point was simple: the AI software showed up, but the compute, power and connectivity to run it where soldiers stand did not.
This month at AUSA, Lt. Gen. Matt McFarlane, commander of I Corps, described the next test (DefenseScoop). I Corps, the Army’s Pacific corps, has begun a deliberate rollout of Next Generation Command and Control (NGC2) to its headquarters and enabling brigades. His top concern is the “tyranny of distance and disruption.” I agree, and I would add one point. Command-post AI fails at the first heat failure or the first dropped link. The compute must be rugged and local before the software matters.
// A corps fights across an ocean
NGC2 spent the past year as a prototype in two divisions. A corps is a different problem. “What’s really different is the scope and scale,” McFarlane said, describing command posts “in Honshu in Japan and in Brisbane in Australia.” Those two cities sit more than 4,000 miles apart. At that scale, no command post can count on reaching central compute. Processing has to move out to the force.
The mission changes with the echelon. “Corps echelons do not fight localized gun battles or tactical fights,” he said. The corps runs joint fires, operational endurance and multi-domain campaigns. Predictive analysis and modeling at that level demand far more processing than a tactical app. McFarlane said some applications validated at the tactical level will need adjustment, and that I Corps will work with industry on more apps and more hardware.
// Heat is a hardware problem
PCC6 put about 9,700 soldiers in the Mojave in July, with highs reported at 106°F (Breaking Defense). Leaders called the exercise a success and were candid about what broke. “We see that we have some shortcomings in some of the technology’s ability just to handle the heat,” said Maj. Gen. Patrick Ellis, commander of the 4th Infantry Division. Gen. Christopher LaNeve, then the acting Army chief of staff, was blunter: “We’re gonna have some failures out here in the desert, and that’s good” (DefenseScoop).
Failure in an exercise is cheap. McFarlane said I Corps is fielding only systems validated at PCC6. That is the right standard, and the Pacific raises the bar. Arctic terrain in Alaska and humid, triple-canopy jungle in the first island chain will stress gear in ways the Mojave did not.
The Army knows where the hard part sits. Joe Welch, the C2 portfolio acquisition executive, said data and application layers are more straightforward to deliver than tactical edge items such as radios and edge compute devices. The Army’s FY2026 budget request included $718.6 million for tactical edge compute hardware for vehicles and mobile command posts (MeriTalk). The priority is clear. The open question is whether the hardware survives.
// Links will drop
McFarlane’s other priority is automatic switching across transport, from “fiber to radio, the different connectivity there, LEO and milsat,” in contested spectrum, “without humans having to be in that loop.”
Resilient transport is essential. It is not sufficient. An adversary that degrades every path at once, even for an hour, isolates any command post that depends on distant compute. A node that processes its own sensor data, runs its own models and keeps planning while disconnected loses bandwidth, not its critical capabilities.
Transport moves the data. Local compute keeps the decision alive when transport fails.
// What a distributed node needs
McFarlane described small, distributed, mobile command post nodes. Each one needs four things:
- Survive the environment. Heat, humidity, cold, dust and shock, without improvised cooling.
- Compute locally. Enough GPU power for operational-level analysis and modeling, not only tactical apps.
- Move with the force. Small enough to relocate quickly once a site is found.
- Run on available power. A draw the node can supply in the field.
Any one of these is achievable. All four in one system is the hard part.
// Where VectorEdge fits
UET builds VectorEdge for this problem.
VectorEdge Carbon is two-person portable, with four NVIDIA H200 or four RTX PRO 6000 Blackwell Server Edition GPUs and 1TB of memory. It is designed to meet MIL-STD-810H and draws about 3.5 kW. I know of no other two-person-portable system offering even two H200s or two RTX PRO 6000 Server Edition GPUs. Carbon fits the small mobile node McFarlane described: data-center-class compute that moves when the command post moves.

VectorEdge Large covers the larger operational node: 24+ NVIDIA H200 NVL GPUs with single-phase immersion cooling, running without a data hall or external HVAC.
I have spoken with former military personnel who were assigned to keep servers running in the desert. They told me they spent more time keeping the cooling systems alive than doing IT work. Every hour spent nursing an air conditioner is an hour not spent on the mission. VectorEdge is built so cooling is not a second job. Carbon and Large both use single-phase immersion cooling built into the system, rugged enough for the heat and paired with the GPU power that operational-level AI demands.
UET does not sell command and control software. We build the hardware underneath it, so the software the Army chooses has somewhere rugged and local to run.
// Decision dominance starts below the software
McFarlane wants “decision dominance at the operational level in the Pacific.” That goal depends on hardware that survives the heat and keeps computing when the links go dark.
Software wins the decision. Hardware decides whether there is one to make.
To see VectorEdge Carbon in person, connect with our team.

