EdgeCortix, Unigen Expand RAIDEN AI Partnership

The companies signed a Memorandum of Understanding to broaden collaboration across product development, manufacturing and go-to-market activity.

Updated on Nov. 19, 2026 in Semiconductors

Bold flat-color vector editorial illustration of stacked server modules in front of a silicon wafer disc, in teal, mustard, and coral.
EdgeCortix and Unigen signed a memorandum of understanding to expand their RAIDEN AI platform partnership across product development, manufacturing, and go-to-market activity. AI Illustration. Upload story photo >

EdgeCortix Inc. and Unigen Corp. signed a Memorandum of Understanding to expand their collaboration to the RAIDEN platform. Unigen plans to design, manufacture, market, and sell next-generation E3.S modules and server compute built on RAIDEN.

Why it matters

The agreement pairs a fabless AI chip designer with a manufacturer that handles production, supply-chain management and aftermarket services, linking chip design to volume hardware. For regions targeted for expansion, including Japan, North America, Europe and Asia, it signals a path for energy-efficient AI infrastructure to reach enterprise and industrial buyers.

The four-die RAIDEN configuration delivers more than 3,300 TFLOPS of AI compute, about 55 times the compute of SAKURA-II, with up to 256 GB resident memory and 548 GB/s memory bandwidth. Performance figures are company-stated, and independent validation has not been reported.

The players

EdgeCortix Inc.

EdgeCortix is a fabless semiconductor company focused on energy-efficient AI infrastructure, and it designs the SAKURA-II accelerator and RAIDEN platform.

Unigen Corp.

Unigen is a global technology and manufacturing company specializing in enterprise and industrial electronics, with capabilities spanning design, volume production, assembly, test and aftermarket services.

The details

EdgeCortix's RAIDEN pairs its DNA-X processor architecture with the MERA software stack that runs SAKURA-II, scaling by die count across single-die, two-die and four-die configurations on one architecture and software stack. Unigen's Amaretti E1.S / E3.S AI Modules, announced in April 2026, use the SAKURA-II accelerator, and its Poundcake servers support up to 24 AI modules with 32 GB of LPDDR4.

Timeline

  1. April 2026: Unigen announced the Amaretti E1.S / E3.S AI Modules.

  2. Early 2027: RAIDEN customer sampling is scheduled to start.

  3. Second half of 2027: Volume production of RAIDEN is planned.

The Tech Race

RAIDEN extends EdgeCortix's SAKURA-II accelerator line, reusing its MERA software stack and application code so existing SAKURA-II deployments can move to the new platform. The partnership adds a volume manufacturing partner to a field where AI accelerator makers compete on power efficiency and deployable compute.

For enterprise and industrial buyers, RAIDEN-based modules and servers are planned to be sold by Unigen once volume production begins, so purchasing decisions are likely to follow the 2027 sampling and production timeline. Buyers evaluating AI infrastructure can look to the Amaretti modules' 10-watt, 60 TOPS profile as a reference for power-constrained deployments.

The takeaway

The partnership shows AI accelerator designers increasingly pairing with manufacturers to move from chip design to deployable hardware. Buyers planning AI infrastructure should track the early 2027 sampling window as the first concrete point to evaluate RAIDEN in real deployments.

What happens next

RAIDEN customer sampling is scheduled to begin in early 2027, with volume production planned for the second half of 2027.

Further reading

Explore more coverage in our Semiconductors section for context on AI chip development and production.

Source note: This article includes information reported by DQ.