GPU Meets RAID. The Industry Took Notice.
Leander Yu walked out of Silicon Motion in late 2019 with a clear-eyed bet: the storage industry's biggest performance ceiling wasn't the drives. It was the controller. NVMe SSDs were delivering speeds that traditional RAID hardware couldn't process fast enough, and nobody had fixed it. Yu decided to fix it by recruiting an unlikely candidate - the GPU.
Founded in December 2019 and headquartered in Sunnyvale, California, Graid Technology's flagship product SupremeRAID offloads all RAID computation from the CPU to a dedicated GPU. The result is a single PCIe card that protects up to 32 NVMe drives simultaneously while delivering up to 28 million IOPS and 260 GB/s of throughput. Those numbers sit well above what any conventional RAID controller can touch. By 2023, Yu was telling interviewers with characteristic confidence: "We don't believe we have any competition. No one can do what Graid does."
That's a brash claim. It also happens to be defensible. GPU-accelerated RAID requires expertise in CUDA development, NVMe protocol, storage architecture, and RAID algorithms - a fairly exotic combination of skills that took Graid years to assemble. Yu's three-year head start estimate was not marketing; it was an engineering reality check.
The market responded. Series A in January 2022 brought $15 million led by Capital TEN. Series B in March 2025 closed at $30 million, with Foxconn and CTBC's joint venture HH-CTBC Partnership leading alongside Yuanta Ventures, Delta Electronics Capital, and Harbinger Venture Capital. The $47.95 million total raised over six years reflects steady institutional conviction rather than speculative fever - a distinction that matters in enterprise hardware, where long sales cycles punish hype-funded companies hard.
Yu's backstory tracks in a straight line toward this moment. He graduated from National Taiwan University in 1998 with a degree in Computer Science and Information Engineering, then spent the early part of his career building expertise in cloud and storage architecture at Trend Micro's cloud division, TCloud Computing. By 2012, he was ready to found his first company - EZ Cloud Tech in Taipei. By 2014, that first chapter closed and Bigtera opened: a software-defined storage company using Ceph as its foundation, with Yu as both CEO and CTO. The company sold to NASDAQ-listed Silicon Motion Inc. in July 2017. Yu stayed on as Vice President through 2019 - long enough to understand what didn't work about existing storage approaches, then left to build Graid.
The name "Graid" isn't arbitrary. It compresses "GPU RAID" into something that sounds like a product brand - which is exactly what it became. The naming decision captures how Yu built the company: precise, deliberate, no wasted syllables.
Graid Technology operates on two continents, with commercial operations in the US, sales across Europe, and an R&D center in Taipei. That dual-geography isn't coincidence; it maps directly to where the relevant talent lives. Taiwan's deep semiconductor and hardware engineering ecosystem is precisely where you want your RAID algorithms stress-tested. Yu has kept that center active and staffed even as the company scaled in California.
The SupremeRAID product line has expanded steadily. The original SR-1000 and SR-1010 serve enterprise and cloud deployments. The SR-1001 (January 2024) targets edge computing and professional workstations. The HE edition (August 2025) addresses high-performance computing clusters where the standard industry answer to data protection - full replication across nodes - wastes storage capacity at exactly the scale where waste is most expensive. The SE (Simple Edition) takes a bring-your-own-GPU approach, letting users slot in a consumer GPU card to protect up to 8 NVMe drives. The AI Edition integrates with NVIDIA GPUDirect Storage to enable direct NVMe-to-GPU memory transfers without CPU involvement - a specification that reads like it was written for AI training workloads, because it was.
The Intel VROC deal in late 2025 marked a strategic inflection point. Intel's Virtual RAID on CPU technology had a sizable installed base. When Graid finalized the license agreement, it inherited continuity responsibility for that customer base and, in the same move, gained a credibility signal that enterprise sales teams spend years chasing. Yu framed it plainly: "Finalizing the deal allows us to ensure long-term continuity for existing customers while also accelerating innovation and value creation across the enterprise storage market."
Graid has won recognition from across the industry: CRN named it one of the Ten Hottest Data Storage Startups of 2021. The 2024 Tech Trailblazers Award followed. The company took Best of Show at Flash Memory Summit 2025, and also earned Top 10 Enterprises of the Year at Taiwan's 19th Golden Torch Awards - a nod to both the product's technical merit and Yu's roots in Taiwanese engineering culture.
The aspirations are clear. Yu wants SupremeRAID to become the default storage performance layer for AI training, HPC, and enterprise NVMe deployments worldwide. The path runs through OEM partnerships, Intel VROC ecosystem integration, and continued product expansion - the HPC and AI editions signal that Graid is following the infrastructure dollars wherever AI workloads go next. In enterprise storage, that's not a speculative bet. It's a calendar appointment.