Rocket One Partners with Virginia Commonwealth University to Develop Energy-Efficient Magnetic Computing Devices

Rocket One Inc. ("Rocket One" or the "Company") announced a sponsored research agreement with Virginia Commonwealth University ("VCU") to develop and demonstrate magnetic computing devices targeting a critical challenge in artificial intelligence: delivering greater computational capability while reducing energy consumption.

The agreement establishes a one-year fabrication and testing program focused on straintronic magnetic tunnel junctions—nanoscale devices that use voltage-induced mechanical strain to control magnetic behavior. The program targets foundational components for next-generation AI chips and accelerators, combining compact design, energy-efficient computation, and the ability to retain stored states without continuous power.

The project is led by VCU Professor Supriyo Bandyopadhyay, Ph.D., with planned device-fabrication collaboration involving Professor Jian-Ping Wang of the University of Minnesota, identified in the project plan as a subcontractor.

"AI's continued advancement depends on hardware that can deliver more computation with less energy," said Robb Knie, Chief Executive Officer of Rocket One. "This agreement establishes a concrete fabrication and testing program targeting magnetic devices for next-generation AI chips. We see potential across AI infrastructure and autonomous systems, including defense and space applications where power consumption, size and local processing capabilities are critical."

Targeting the Core Computations Behind AI

The program will pursue three complementary hardware demonstrations:

• Ultra-low-power analog multiplication: Fabricate and test a magnetic device that multiplies two voltage inputs, addressing a fundamental mathematical operation used in AI and machine learning.
• Devices for more accurate AI matrix computation: Demonstrate a magnetic crosspoint device with continuously tunable, linear electrical conductance and investigate ways to expand its operating range. This behavior is intended to support more accurate computation in crossbar-based AI hardware.
• Compact magnetic logic: Fabricate and demonstrate an exclusive-OR, or XOR, logic gate based on a single straintronic magnetic tunnel junction. XOR functions are used in arithmetic, error detection, and encryption.

Matrix multiplication is a core operation in neural networks. By targeting devices that perform multiplication and support crossbar computing architectures, the program addresses building blocks that could contribute to future AI accelerators.

Exploring Extremely Low Energy Consumption

The project builds on prior straintronic magnetic tunnel junction research, including an earlier experimental device demonstration involving VCU and University of Minnesota researchers.

The research plan cites theoretical estimates of energy consumption as low as 100 attojoules per multiplication operation for the proposed magnetic multiplier. The program will focus on fabricating devices and measuring their operation experimentally. This figure is a theoretical device-level estimate, rather than measured performance from the announced program or a complete AI processor.

The proposed devices are also nonvolatile, meaning their stored magnetic states can persist without continuous power. These characteristics could support future in-memory computing architectures that bring computation closer to stored information.

"Our objective is to translate promising magnetic computing designs into measurable device results," Knie added. "Those results would help us assess performance, define the next engineering milestones and evaluate potential commercialization pathways."

Potential Applications Across AI Infrastructure, Defense and Space

Rocket One views compact, energy-efficient processing as strategically relevant to both AI infrastructure and systems operating under constrained power budgets or limited connectivity.

Potential applications include autonomous platforms, remote sensing and defense and space systems that must process information locally. The announced program focuses on foundational device fabrication and electrical testing. Processor integration and qualification for specific applications, including space operating environments, would require additional development.

Defined Development Objectives and Intellectual Property Options

The agreement provides for a research period from October 19, 2026, through October 19, 2027.

It grants Rocket One an option to negotiate licenses to intellectual property arising from the project, including the ability to seek an exclusive license within its field of commercial interest, subject to the agreement's terms, VCU's ability to grant those rights and a separately negotiated license.