Graid Technology Inc.

Home
United States
Sunnyvale, CA
Graid Technology Inc.

Graid Technology keeps AI infrastructure moving. We remove the bottlenecks that slow training and inference, so AI runs at full speed.

Headquartered in Silicon Valley. Learn more at graidtech.com. Chosen by CRN as one of the Ten Hottest Data Storage Startups of 2021 and a 2022 Emerging Vendor in the Storage & Disaster Recovery category, GRAID Technology Inc. has developed the world's first NVMe and NVMeoF RAID card to unlock the full potential of enterprise SSD performance, and has redefined performance standards for enterprise data protection. We're headquartered in Silicon Valley, with an R&D center in Taiwan, and are led by a dedicated team of experts with decades of experience in the SDS, ASIC and storage industries. Our extraordinary software plus hardware solution has redefined the value of SSD RAID cards and makes SupremeRAID™ the most powerful and flexible NVMe SSD RAID in the world. SupremeRAID™ tri-mode NVMe/NVMeoF RAID cards offer enterprise and OEM customers a unique blend of flexibility, superior performance, and scalability: single SupremeRAID™ card is capable of delivering 19 million IOPS and 110GB/s of throughput. Learn more: visit www.graidtech.com

11/03/2026

We’re proud to share the first release of VROC™ by Graid Technology! If you’ve relied on to protect your OS volume, you’re in good company. It's been deployed on more than one million CPUs since 2017. Version 26.H2 builds on that with a few things we think you’ll appreciate:

👉 UEFI-based licensing, so you rely less on physical hardware keys
👉 GPUDirect Storage support for GPU-accelerated, data-intensive workloads
👉 Coexistence with SupremeRAID™, so you can run both in the same system
👉 A 24-month, OEM-driven roadmap focused on expanding platform support

Read the full release blog: https://zurl.co/Jqe6b

One more thing: we’re getting ready to challenge some long-held assumptions about RAID. We’ll share more at SC26 in mid-November, and we’d love to see you there. Contact us today for a complimentary pass, and stop by Booth 1326 to meet the team.

10/29/2026

Parity RAID has long meant giving up NVMe performance in exchange for data protection. Hardwareluxx shows what changes when GPU acceleration removes that bottleneck.

After outgrowing its legacy ESX and SATA environment, Hardwareluxx rebuilt on Proxmox with Gigabyte servers, Intel® Xeon® 6 processors, Kingston DC3000ME Gen5 NVMe SSDs, and SupremeRAID™ by Graid Technology. The platform now delivers more than 100 GB/s of RAID 5 read performance and keeps latency consistently low in both RAID 5 and RAID 6 configurations, with CPU cores left free for application workloads.

Read the full case study: https://zurl.co/eJsEK

Media Gmbh Technology

10/28/2026

Thank You for Visiting Us at Japan IT Week Autumn 2026, and a special thank you to our partner Electronics Corp. for hosting us at their booth.

Here’s a quick look at the solutions we showcased:

SupremeRAID™ AE uses existing GPUs to accelerate NVMe storage with enterprise-grade data protection, with no extra RAID hardware, cables, or PCIe slots. Paired with OEC’s server solutions, it powers demanding AI workloads.

SupremeStore combines SupremeRAID™ HE and BeeGFS to deliver high-performance cluster storage for AI and HPC, offering up to 80 GB/s throughput, linear scalability, and high availability.

Thank you, OEC, for a strong partnership. Contact our team to learn more.

10/27/2026

When AI context outgrows GPU memory, most systems throw it away and rebuild it from scratch. That means wasted GPU cycles and a higher cost per token.

At SC26, we’re showing how Graid Technology’s GPU-accelerated I/O turns the NVMe already inside your GPU servers into a high-speed KV cache tier, so AI can retrieve context instead of recomputing it. For CPU-based systems, VROC™ by Graid Technology delivers data protection engineered for every architecture and performance tier.

It’s one portfolio for any workload, whether it runs on GPUs or CPUs.

Find us at Booth 1326, November 17–19, and enter our raffle for a chance to win a pair of Meta glasses.

Want to meet with our team at the show? Contact us to set up a meeting: https://zurl.co/OlqsA

10/22/2026

Deskside AI systems have a storage problem that goes beyond capacity. Every model and dataset copied onto a local drive sits outside the backup, access controls, and audit trail that live in the data center.

Storage Review tested a different approach: eight NVIDIA DGX Spark systems mounting one governed, RAID 5 protected Solidigm QLC array over 100GbE, with SupremeRAID™ handling parity on a GPU.

The results:
👉 24.9GiB/s aggregate sequential read across all eight systems
👉 10.8GiB/s from a single system, matching the fastest internal M.2 drive Storage Review has measured in a GB10 system
👉 A 78GB model loaded from shared storage within 13% of its local NVMe load time

The result is one copy of every model and dataset, kept under data center control and available to every system on the fabric.

Read our recap and the full Storage Review report: https://zurl.co/zKneM

10/20/2026

Over the next two years, buyers will judge storage for AI on time to first token, cache-hit rate, and cost per million tokens served. Raw capacity will still matter, but it will stop being the headline number.

That is Graid Technology Field CTO Randy Kreiser’s view in TechArena’s new report, The Integrated Machine. The report traces how rack-scale integration is changing AI infrastructure economics, from compute and memory to power, cooling, and firmware.

Randy explains why inference turns storage into a memory problem. Every active request carries a KV cache that grows with context length and concurrent users, and agentic workloads keep adding to it. Once the rack is the unit of deployment, storage has to be designed in from the start.

Read our recap of the report: https://zurl.co/5r3xV

.ai

10/15/2026

🇯🇵 Bringing GPU-accelerated storage to Japan IT Week Autumn 2026!

Join Graid Technology at the OEC booth to explore storage solutions for demanding AI and HPC workloads.

🔹 SupremeRAID™ AE
Using existing GPUs, SupremeRAID™ AE accelerates NVMe storage with enterprise-grade data protection—without extra RAID hardware, cables, or PCIe slots. Explore AE with OEC’s server solutions for demanding AI workloads.

🔹 SupremeStore
Combining SupremeRAID™ HE and BeeGFS, SupremeStore delivers high-performance cluster storage with scalable capacity and cross-node high availability—supporting the growing storage demands of AI and HPC environments.

📅 October 21–23, 2026
📍 Makuhari Messe, Japan
🤝 OEC Booth A7-5
🌟 Learn More:https://zurl.co/XTp3d

Meet us to discuss storage solutions for your next AI or HPC deployment.

10/13/2026

"You've got a four-lane highway, and to get to your storage, you have to go down to a one-lane bridge."

That's how Garrett McKibben, Senior Director of Technical Marketing at Graid Technology, described the storage bottleneck choking AI inference today, on TechArena's Data Insights podcast with Allyson Klein and Jeniece Wnorowski (Solidigm).

The topic: KV cache, the working memory an inference model holds for every active request. As context length and concurrent users grow, teams run out of room fast. McKibben's math: a Blackwell GPU tops out around 180GB of HBM at $50 to $80 per gig, and concurrency collapses after three or four users. Buying another GPU for every few users isn't a real answer.

Graid Technology's approach moves that overflow to NVMe and accelerates the IO path on the GPU itself, using a small slice of GPU resources before releasing them back. The result, per published testing: faster time to first token than recomputing the context from scratch.

Watch the full episode: https://zurl.co/jP3LJ

10/08/2026

Ships, submarines, aircraft, and remote edge sites need servers that protect and process mission data locally while withstanding shock, vibration, and temperature extremes, all within tight space and power budgets. Systems built the TRS14-18.5 for that environment: an ultra-rugged, short-depth 1U server on Intel® Xeon® 6, rated from -40°C to 50°C, and fully designed and manufactured in the USA.

Trenton Systems uses both Graid Technology storage platforms in the same server:

➡️ VROC™ by Graid Technology provides CPU-native, bootable NVMe RAID for boot and OS volumes, such as a mirrored RAID 1 boot volume, without a dedicated RAID card.
➡️ SupremeRAID™ provides GPU-accelerated RAID for a separate, performance-critical NVMe data array where sustained throughput and low latency are mission requirements.

Each storage tier gets the RAID architecture best suited to its role. Trenton Systems has also demonstrated a configuration pairing SupremeRAID™ with PCIe Gen5 NVMe storage and dual 400GbE networking, designed to scale to nearly 1 PB of raw storage in a short-depth 1U footprint.

Read the blog: https://zurl.co/NnLBY

Attending AUSA in Washington, DC, Oct 12-14? Connect with Mike Flender of Trenton Systems to set up a meeting and explore a configuration built for your mission.

10/06/2026

AI infrastructure doesn't fail because the GPUs are slow. It fails when the storage behind them can't keep up.

Graid Technology CEO Leander Yu and VP of Product Development Alven Yen sat down to explain why the company rebuilt RAID from the ground up, moving core computation onto the GPU to eliminate the bottleneck traditional controllers create, without giving up the redundancy and data protection enterprises depend on.

The result is two complementary platforms built for the AI era: SupremeRAID™, our GPU Accelerated Storage flagship, and VROC™ by Graid Technology, extending that same protection to CPU-native environments from the data center out to the edge.

Full interview and breakdown here: https://zurl.co/xzYQf

Address

150 Mathilda Place, Suite 100
Sunnyvale, CA
94086

Website

https://graidtech.com/

Alerts

Be the first to know and let us send you an email when Graid Technology Inc. posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.