A Hoboken company that used to make its name in biotechnology has signed a one-year research agreement with Virginia Commonwealth University to develop a new kind of computer memory built from microscopic magnetic swirls. Rocket One Inc. says the deal, which kicks off November 1, will push forward a technology called SkyMRAM, which the company hopes could eventually power AI chips, satellites and defense hardware.
The sponsored research agreement was first reported by ROI-NJ, which detailed the scope of the one-year program and its milestones. VCU will lead the project, with the university's Jayasimha Atulasimha serving as principal investigator, according to the outlet. The agreement also includes a Northwestern University subaward supporting thin-film growth, materials characterization and optimization work tied to the project.
Rocket One is not exactly a household name in semiconductor circles. The company was formerly known as Hoth Therapeutics, a biotechnology firm trading under the ticker HOTH, before its corporate name change took effect on May 26, 2026, and its ticker changed to RKTO at market open on May 28, according to a release distributed via PR Newswire. The company retained its legacy biotech pipeline under a wholly owned subsidiary while pivoting its core public identity toward space economy hardware and AI chip architectures, the release states.
Building on an Existing VCU Patent Relationship
The VCU deal did not come out of nowhere. Rocket One entered two exclusive license agreements with VCU in May 2026.
Rocket One also brought VCU expertise directly into its leadership structure. The company appointed VCU Commonwealth Professor Supriyo Bandyopadhyay, director of VCU's Quantum Device Laboratory and a frequent research collaborator with Atulasimha, as its Lead Technical Advisor for AI Nanomagnetic Technology in June 2026. Per the company's advisor page, Bandyopadhyay has authored more than 400 research publications focused on spintronics and nanomagnetic devices.
What SkyMRAM Actually Is
SkyMRAM uses nanoscale magnetic structures called skyrmions as the basis for a proposed magnetic memory architecture, per ROI-NJ's reporting. The underlying physics traces back further than this deal: foundational skyrmion research conducted at VCU and published in Nature Electronics in June 2020 demonstrated that electric fields could create and manipulate non-volatile magnetic skyrmions using voltage control of magnetic anisotropy, according to VCU News. That matters because conventional magnetic memory relies on current-driven switching that produces excess heat, while voltage control significantly reduces energy consumption, the university reported.
The efficiency gap is substantial. Academic research evaluating spintronic memory architectures, compiled in a study hosted on PMC, shows that voltage-controlled magnetic anisotropy switching requires only a few femtojoules of energy per operation, compared to roughly 100 femtojoules per bit for standard spin-transfer torque MRAM devices. High energy consumption in traditional silicon memory limits performance on power-constrained platforms like satellites and remote defense hardware, which is part of why the approach has drawn interest beyond the lab.
VCU-led work has already pushed the concept past simple on-off switching. A July 2024 study led by Atulasimha proposed a skyrmion-mediated ternary memory architecture capable of storing three magnetic states within a single cell, with modeled switching showing a 99% probability under room-temperature thermal noise, according to findings published on PubMed. Multi-state ternary memory could allow AI models and deep neural networks to operate in smaller physical footprints with higher storage density, the research notes. Stabilizing those nanoscale skyrmions in thin-film stacks requires a specific interfacial effect between ferromagnetic and heavy-metal layers, a materials physics detail outlined in patent filings on Google Patents, and optimizing that thin-film growth and layer interfaces is a primary focus of Northwestern's subaward work.
What the One-Year Program Will Actually Do
Per ROI-NJ's account of the agreement, the planned work includes developing and refining thin-film structures for device testing, investigating control of the soft magnetic layer, and patterning, etching and characterizing magnetic material stacks designed to host skyrmions. The program aims to refine structures capable of hosting skyrmions roughly 100 nanometers in size. Materials development and device measurements will operate on a feedback loop, with the program setting quarterly objectives that progress from initial material-stack fabrication and characterization toward switching studies and optimized nanoscale structures, according to the outlet.
The project's proof-of-concept milestone is demonstrating switching of SkyMRAM's soft magnetic layer, ROI-NJ reported. Rocket One has an option to negotiate licenses to any intellectual property arising from the project, subject to the agreement's terms, the outlet noted.
Knie framed the deal as a practical step rather than a finished product. “Building the next generation of AI computing requires advances in the underlying hardware, including how information is stored and switched,” he said, per ROI-NJ. He added that the agreement establishes a concrete development path for SkyMRAM, with fabrication, characterization and switching milestones for device-level testing, and said Rocket One's goal is to build the technical foundation for future chip architectures serving AI infrastructure, space and defense.
Rocket One itself describes the program as foundational memory-device research rather than completed AI processors or validated system-level performance, according to ROI-NJ. Applications in AI infrastructure, space and defense will require further development and application-specific validation, and may require additional testing to establish radiation tolerance, the company said.
Why Space and Defense Markets Matter Here
Rocket One's pivot toward orbital and military applications is not incidental. The global space economy is forecast to reach $1.8 trillion by 2035. High-radiation environments in low-Earth orbit cause conventional silicon chips to experience data loss and hardware degradation, which is part of the appeal of a lower-power, voltage-driven memory architecture.
The company has also been expanding its defense footprint outside the memory-chip effort by launching the Swarm Stage AI platform.
Rocket One trades on the Nasdaq Capital Market under the ticker RKTO, with a market capitalization of approximately $13.6 million as of October 2026, according to data from Robinhood. The agreement gives Rocket One a way to advance its early-stage hardware IP through university research. For now, the SkyMRAM program remains squarely in the prototyping stage — soft-layer switching at 100-nanometer scales — well short of a commercialized chip ready for mass production.
