GigaIO

GigaIO, founded in 2012 by Joey Maitra, develops composable infrastructure solutions for data centers, focusing on Artificial Intelligence (AI) and High-Performance Computing (HPC) markets. The company is headquartered in Carlsbad, California, with Alan Benjamin serving as CEO. Maitra, the Founder and Chief Architect, invented the company's core intellectual property related to using PCI Express as a unified system area network.

The company's core technology is FabreX™, an AI memory fabric that allows for the dynamic composition and disaggregation of data center resources like GPUs, FPGAs, CPUs, storage, and memory. This is designed to overcome the limitations of static server architectures by creating a flexible, software-defined infrastructure. By using the native PCIe protocol, FabreX aims to reduce latency and eliminate the performance bottlenecks associated with traditional networking fabrics like Ethernet or InfiniBand. The software stack includes open-source, Redfish-compliant APIs to integrate with orchestration tools like Ansible, Chef, and Puppet.

GigaIO's product line targets clients in sectors requiring intensive computation, such as AI/ML, defense, and research, including customers like the San Diego Supercomputer Center. Key products include the SuperNODE™, a 32-GPU AI supercomputer, and Gryf™, a portable, suitcase-sized supercomputer designed for edge computing environments. These products are marketed as vendor-agnostic, allowing customers to integrate accelerators from various manufacturers like NVIDIA, AMD, d-Matrix, and Tenstorrent. The business model focuses on reducing the total cost of ownership for data centers by improving resource utilization and energy efficiency. The company has raised a total of $45.9M in funding over several rounds, with a recent Series B round in July 2025 raising $21M to scale production and expand sales.

Keywords: composable disaggregated infrastructure, PCIe fabric, data center architecture, AI infrastructure, high-performance computing, GPU pooling, software-defined hardware, memory fabric, rack-scale solutions, NVMe-oF, edge computing, AI inferencing, FabreX, SuperNODE, Gryf, workload-defined infrastructure, data center automation, low-latency interconnect, resource utilization, accelerator pooling, dynamic infrastructure