Fixstars Corporation
Category: AI Infrastructure
Japanese performance-engineering company that accelerates AI/LLM training and inference, GPU utilization, and embedded AI, and delivers secure on-premises AI infrastructure. Fixstars Corporation was founded in 2002. The company is led by Satoshi Miki. Based in Tokyo, Japan. Team size: 101-500. Total funding raised: $1.02M. Latest round: IPO. Key investors include Toshiba Corporation, Mitsubishi UFJ Capital.
- Founded
- 2002
- Headquarters
- Tokyo, Japan
- Team size
- 101-500
- Total funding
- $1.02M
Value proposition
Fixstars makes AI/GPU workloads faster, cheaper, and more energy-efficient through deep performance engineering — profiling, optimizing, and porting AI models/software across GPU, FPGA, embedded and quantum-annealing hardware — including secure, on-premises deployments for confidential-data workloads.
Products and solutions
Fixstars AIBooster (GPU/AI workload profiling & auto-optimization, claims up to 80% GPU cost reduction), Fixstars Vega (secure on-prem AI development environment for embedded teams), AI Model Porting & Optimization for Embedded Systems, Fixstars Amplify (quantum/Ising-machine combinatorial optimization cloud SDK, 1,000+ organizations), DC-PKG liquid-cooled GPU container data-center package (with NTTPC Communications & Getworks, launched Sept 2026), FPGA, flash memory, and firmware development services.
Unique value
20+ years of deep performance-engineering expertise (since 2002) spanning GPU/CPU/FPGA/quantum hardware, applied to modern AI/LLM acceleration and secure on-prem GPU infrastructure — a rare combination of low-level systems optimization skill and enterprise trust (99% client repeat rate, TSE Prime-listed).
Target customer
Enterprises running large-scale AI/LLM training and inference (GPU cost/performance pain points), embedded AI/hardware developers needing on-device model porting, and organizations with confidential-data constraints needing secure on-prem AI infrastructure (finance, manufacturing, automotive, government).
Industries served
AI/LLM infrastructure, Automotive (embedded AI, R-Car SoC), High-performance computing (Fugaku supercomputer), Mining & logistics (optimization), Finance (Mizuho derivatives, quantum optimization), Robotics (Qualcomm Japan Robotics Center), Healthcare (Smart Opinion AI breast-cancer analysis)
Technology advantage
Deep multicore/GPU/FPGA/quantum optimization heritage (Cell/PS3 clusters, OpenCL, Fugaku supercomputer work) repurposed for AI/LLM acceleration; AIBooster's automated bottleneck detection and cost reduction; secure on-premises deployment model for confidential-data AI use cases.
How they differentiate
20+ year track record in raw performance/systems engineering (not just an ML tooling wrapper), spans GPU+FPGA+quantum hardware, and offers secure on-premises AI infrastructure for confidential-data enterprises — a niche most cloud-native AI infra competitors don't serve.
Main competitors
MosaicML/Databricks (GPU/AI training optimization platforms), Baseten and Together AI (AI inference/GPU optimization infrastructure), Preferred Networks (Japan AI infrastructure & compute optimization).
Key partnerships
NTTPC Communications & Getworks (liquid-cooled GPU DC-PKG data center), Qualcomm Japan Robotics Center (launch partner, Aug 2026), AIST (Japanese quantum computer 'System F' support in Fixstars Amplify SDK), D-Wave Systems (quantum computing collaboration since 2017), OpenPOWER Foundation, Toshiba Corporation (strategic investor/partner), Nallatech (OpenCL FPGA cards), Waseda University Venture Fund (Fixstars is an LP investor)
Notable customers
Mizuho Securities (derivatives calculation system), US Air Force Research Lab (PS3 cluster, 2010), RIKEN/Fugaku supercomputer project, automotive OEMs via R-Car SoC AI optimization work.
Major milestones
2002: Founded in Yokohama, 2013: Toshiba/Mitsubishi UFJ Capital investment, 2014: IPO on TSE Mothers market, 2016: Moved to TSE First Section, 2017: Fugaku predecessor 'Kei' supercomputer #1 in Graph 500 (5th consecutive year), 2021: Established Fixstars Amplify Corporation (quantum cloud), 2022: Transferred to TSE Prime Market, 2024: Launched AI Booster, 2026 (Sept): Launched DC-PKG liquid-cooled GPU container data center with NTTPC and Getworks.
Growth metrics
TTM revenue ~$68.2M (as of Mar 2026, PitchBook); net sales +22.4% YoY in H1 FY2025 with record-high quarterly revenue; market cap ~¥87.21B (~$580M) as of 2026; Fixstars Amplify surpassed 1,000 registered organizations and 100M cumulative executions.
Market positioning
Established (TSE Prime-listed, 2002-founded) enterprise performance-engineering vendor pivoting its deep hardware-optimization expertise toward the AI/LLM acceleration and secure on-prem GPU infrastructure market in Japan and via US subsidiary (Fixstars Solutions, Irvine CA).
Geographic focus
Primarily Japan, with a US subsidiary (Fixstars Solutions, Irvine, California) extending North American reach.
Patents and IP
US patents assigned to Fixstars Corporation include: US12334211 (breast CAD/diagnosis support with neural-network recognizers, granted Jun 17 2025); US12314153 (information processing / duplicate sub-block detection, granted May 27 2025); US9904483 (file management device/method, inventor Satoshi Miki, granted Feb 27 2018); plus related US applications on NV memory file systems and flash memory temperature/wear management. Also JP/WO filings on combinatorial optimization hardware mapping (e.g. WO2021172593, inventors Matsuda et al.).
About Satoshi Miki
Founded Softworld in 1998 (VP of the board); handled online securities trading systems; B.A. Waseda University School of Commerce, 1995.
Latest news about Fixstars Corporation
More AI Infrastructure companies
Official website: https://www.fixstars.com/en/
