Graid Technology Inc.

Home
United States
Sunnyvale, CA
Graid Technology Inc.

Graid Technology builds GPU Accelerated Storage and CPU-native performance for AI, HPC, and enterprise infrastructure. Visit graidtech.com

Creator of SupremeRAID™, the GPU Accelerated Storage engine built for the AI era, and VROC™ by Graid Technology. 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 h

as 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

09/01/2026

Graid Technology is heading to Split, Croatia as a Gold Sponsor of Supermicro Innovate! EMEA, September 14-15, 2026.

This invitation-only event brings together Supermicro's top EMEA customers, channel partners, and technology leaders for two days of executive and technical content focused on enterprise AI and data center infrastructure.

Find us at Booth G1. Our theme for the show: Data Protected & Performance Optimized with Graid Technology.

At the booth, our team will be showcasing:
-- SupremeRAID™, Graid Technology's GPU Accelerated Storage engine built for the AI era
-- VROC™ by Graid Technology, our CPU-native RAID platform

As AI and HPC workloads push infrastructure harder than ever, data protection cannot come at the cost of performance. That's the trade-off our portfolio is built to solve.

If you'll be in Split, stop by Booth G1, or reach out to book time with our team at the event.

👉 Full details: https://zurl.co/o2pOC

08/28/2026

Last week, Graid Technology participated in Dell Technologies Forum Shanghai, showcasing how SupremeRAID™ is transforming NVMe storage for the AI era. During the theater session, “SupremeRAID™ Evolution: GPU-Powered Innovation in NVMe Storage Architecture,” we introduced comprehensive high-performance storage solutions for AI workloads—from cloud and enterprise data centers to private AI and edge environments.

One highlight was how Graid Technology-powered Smart JBOF enhances Dell Pro Max with GB10 by enabling external KV Cache offloading over a 100GbE network with NVMe-oF over RDMA. With RAID 5 data protection, the solution delivers up to 24× faster TTFT compared with MD RAID, significantly improving the responsiveness of local AI inference while maintaining data protection.

Thank you to everyone who joined our session and visited us at the event! We look forward to continuing the conversation and accelerating AI infrastructure together.

08/25/2026

KV cache offload is supposed to help your inference stack. But what if it’s actually making things worse? 🤔

Most teams dealing with GPU memory pressure know they need to offload KV cache to local NVMe. The instinct is right. But the ex*****on matters — a lot. We ran a controlled benchmark to determine whether RAID-protected local NVMe actually improves inference or just adds complexity. The results were clear:

❌ No KV cache offload: 29.4s mean TTFT, 95.1s query round time

❌ Linux MD RAID5 (default settings): 36.6s TTFT, 117.7s query round time — worse than no offload at all

✅ SupremeRAID™ AE RAID5: 9.0s TTFT, 29.7s query round time — 3.26x faster than no offload, 4.06x faster than Linux MD RAID5

The lesson: RAID protection only delivers value if the storage path is fast enough to preserve it. Linux MD RAID5 at default settings added overhead without preserving NVMe performance. SupremeRAID™ AE — GPU-accelerated RAID software — kept over 95% of raw NVMe performance intact while adding RAID5 protection against drive failure.

For production inference under memory pressure, that difference isn’t incremental. It’s the difference between a KV cache tier that helps and one that hurts.

📄 Read the full whitepaper → https://zurl.co/eXsvn

🌐 Learn about SupremeRAID™ AE → https://zurl.co/6RIcT

🤖 Explore our KV Cache portfolio → https://zurl.co/9sFv1

08/20/2026

Join Graid Technology at Dell Technologies Forum Shanghai on August 21 to discover how the Dell × SupremeRAID™ joint solution breaks through AI storage bottlenecks while delivering exceptional performance and enterprise-grade data protection.

Four featured application scenarios:

🔹 On-Premises Private AI
Achieve 2–10× higher storage performance with GPU-accelerated RAID, providing a fast and secure data infrastructure for enterprise private AI.

🔹 Small and Medium-Sized HPC Data Center Clusters
Our integrated appliance, powered by SupremeRAID™ HE and BeeGFS, delivers up to 80 GB/s per node, supports up to 32 HGX systems, and scales to 1,000+ nodes.

🔹 Edge AI and KV Cache Offloading
SupremeRAID™-powered Smart JBOF supercharges Dell Pro Max with GB10, delivering up to 24× faster TTFT than MD RAID with RAID 5 data protection for faster, more reliable local AI inference.

--DTF Shanghai 2026--
📅 21, Aug. 2026
📍 Pudong Shangri-La, Shanghai
🔗 Learn More : https://zurl.co/PkZHw

08/18/2026

GPU memory pressure is a real infrastructure bottleneck — and it's getting worse as context windows grow. 📈

As long-context LLM inference scales, KV cache demand scales with it. When the cache working set outgrows GPU memory, serving stacks either degrade or teams are forced to scale compute nodes just to get more local SSD headroom. That's an expensive and inflexible answer.

We built the SupremeRAID™ KV Cache for Rack to offer a better one.

The architecture is straightforward: a dedicated SupremeRAID™ NVMe storage appliance connects to GPU servers over high-speed Ethernet and exports a shared NFS cache path via LMCache. GPU servers stay focused on inference. Cache capacity scales independently — without touching the compute fleet.

We validated it with a production-grade benchmark:

🖥️ Model: Qwen3-235B on 4x NVIDIA H200 GPUs

⚙️ Stack: vLLM + LMCache + EvalScope multi-turn workload

📦 Storage: 10 × KIOXIA CM7-V NVMe SSDs in RAID 5 on Supermicro SSG-221E-DN2R24R

🔗 Network: Dual 200 Gb/s Ethernet

Total Throughput improvement vs. no KV cache offload:

→ +32.7% at 512-token prefix length

→ +47.0% at 2,048-token prefix length

→ +51.6% at 4,096-token prefix length

→ +53.4% at 8,192-token prefix length

Zero fails across all 512 requests, at every prefix length tested.

The trend is the story: the longer the reusable context prefix, the more the external cache tier contributes. And inference workloads are moving toward exactly this shape — longer contexts, higher concurrency, more multi-turn depth.

Graid Technology has qualified SupremeRAID™ KV Cache for Rack on 20 storage server platforms across AIC, Dell, Giga Computing, Lenovo, and Supermicro — giving architects validated options across a range of form factors, processor platforms, and NVMe densities.

📄 Read the full white paper — benchmark methodology, test configuration, and complete results — here: https://zurl.co/gaomR

Join Graid Technology at Dell Technologies Forum Shanghai on August 21 to discover how the Dell × SupremeRAID™ joint so...

08/14/2026

08/13/2026

The Summer 6 Pack Promotion is here! ☀️ Buy 5 SupremeRAID™ licenses, get the 6th free.

Summer is too short to wait 30 weeks for a Broadcom RAID card. And your NVMe array shouldn't be running at 12–18% of line rate — the ceiling every hardware RAID controller quietly imposes on the flash you paid full price for. The good news? SupremeRAID™ ships now, no delays. GPU-accelerated RAID that unlocks full NVMe bandwidth, returns host CPU cores to your applications, and delivers enterprise-grade RAID 5/6/10 protection on a single card.

The Summer 6 Pack Promotion is live through September 30: get 6 SupremeRAID™ licenses for the price of 5, discount applied at quote. Bring your own NVIDIA GPU and use SupremeRAID™ AE, or bundle the license with a GPU sized to your workload.

Crack open the promo before summer's over! →
https://zurl.co/8CmpF

08/11/2026

AI demands a fundamentally different storage architecture—one that is parallel, resilient, and efficient.

In our latest joint whitepaper with Innogrit Corporation, we show how modern AI workloads break traditional storage—and how GPU-accelerated RAID changes the equation.

With SupremeRAID™ 2.0 + InnoGrit N3X SLC NVMe:
⚡ Up to 6.4M IOPS random write (RAID5)
⚡ Sustained 200+ GB/s throughput even in degraded mode
⚡ Up to 107× CPU efficiency improvement

No trade-offs between performance, protection, and efficiency—even under failure conditions.

If you’re building AI infrastructure, this is what next-gen storage looks like.

📄 Read the full white paper: https://zurl.co/bOA4R

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.

Contact The Business

Send a message to Graid Technology Inc.:

Want your business to be the top-listed Computer & Electronics Service in Sunnyvale?