Artificial intelligence (AI) pushes military to next-gen computer performance with energy efficiency
Artificial intelligence (AI) pushes military to next-gen computer performance with energy efficiency
Publish Date: 2026-02-03 05:45:00
Source Domain: www.militaryaerospace.com
ARLINGTON, Va. – U.S. military researchers are asking industry to develop next-generation computing technology using logic devices with the same performance as complementary metal-oxide-semiconductor (CMOS) transistors, but with 100 to 1,000 times greater energy efficiency.
Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., issued a solicitation (DARPA-PA-25-07-01) on Friday for the Fast and Curious program.
Such a technology breakthrough is necessary to support the rapidly growing computational demands of artificial intelligence (AI), data-intensive workloads for simulation and modeling, and edge computing, DARPA officials say.
CMOS transistors form the foundational technology for essentially all modern microelectronics computing, from mobile devices to exascale supercomputers, researchers say.
Few alternatives
To date, no alternative technology has been able simultaneously to match CMOS in speed, switching energy efficiency, scalability, and integration density. CMOS gain enables it to achieve large voltage swings at gate output sufficient to drive three to seven subsequent gates. As a result, CMOS has remained dominant for digital computing for more than five decades.
At the same time, however, continued improvements in CMOS energy efficiency are approaching a fundamental limit. Energy consumption from microelectronics and data centers is expected to reach as much as 50 percent of total energy consumption at the current levels of energy efficiency.
Thermal management takes up a large percentage of the work, energy, and space in military supercomputing facilities. Furthermore, military edge computing that operates under severe constraints in power and thermal management are being asked to perform AI inference, sensor fusion, autonomy, and real-time decision-making.
Unlike data centers, edge systems cannot rely on abundant grid power, active cooling, or large physical footprints; power consumption in these applications…
