Solidigm

At Technology Live! in Munich, Solidigm covered 2 Qs enterprise teams are wrestling with: 1. How do you build AI infrastructure that's truly efficient? 2. How do you get your organization to genuinely adopt AI rather than just license it? The answers? news.solidigm.com/en-WW/268302-t…

Our @RaiseSummit recap has the full story, including recorded interviews & what came of the main stage panel. 📹 A big thank you to @bjbaumann2014 and @furrier for their collaboration and for having @Solidigm on @theCUBE with @NYSEWired. What a show!(2/2) news.solidigm.com/en-WW/268161-s…

"Senior leaders from organizations of all sizes - from hyperscale to the edge - discussed efficiency, data sovereignty, and build out speed...But storage was the hottest topic!" 🔥 🎤 That's Greg Matson, straight from the show floor at @RaiseSummit 2026 in Paris. (1/2)

What if memory capacity is the third axis of AI intelligence, sitting right alongside model size and raw compute? During the session, “Scaling Intelligence Through the Memory Hierarchy,” at #AIFD8, Kapil Karkra tackled that question head on. See more: solidigm.com/products/techn…

@Solidigm's proud to join the world's top founders, builders & executives shaping the future of AI at @RaiseSummit with @TheCUBE and @NYSE Wired in Paris! Can't wait to share more about why storage is such a critical component underlying AI infrastructure raisesummit.com

In a recent episode of Future Ready Leadership 🎧 🎙️ Greg Matson joined Jacob Morgan to explore what enterprise AI requires at the infrastructure level and why storage is an essential part of the AI boom. Listen to the full ep. for more valuable insights! open.spotify.com/episode/2WZPRv…

To modernize its infrastructure, @RoseHulman turned to @DataON & @Solidigm for an all-flash Azure Local solution built on high-cap SSDs. The upgrade helped reduce maintenance windows, cut server deployment time, & free IT resources. Read the white paper: solidigm.com/products/techn…

The AI bottleneck has moved from compute to context. #TeamSolidigm's Jeff Harthorn and Ace Stryker break down why AI infrastructure needs a dedicated context tier, and what flash must deliver to support it. @VentureBeat: venturebeat.com/orchestration/…

A single AI query may trigger a ~42,000-token context bundle before a response is generated. As context windows hit 1M tokens, prompts become state objects. Storage architecture has to be designed for that. Our new white paper breaks down why. solidigm.com/products/techn…

Underutilized GPUs are expensive GPUs. Context windows are growing. HBM fills up. GPUs recompute. Utilization drops. The fix starts with storage architecture. @solidigm's Avi Shetty broke it down with @ryanshrout and @Thesixfivemedia at @computextaipei. 🎙️ bit.ly/4a4kkKA

@Solidigm recently joined @Akamai & @DDNintelligence at the @BBGIntelligence Generative AI Forum to tackle one of AI infra's most pressing questions. What happens to the data layer when AI gets exponentially more demanding? Our upcoming Bloomberg Tech Disruptors Podcast dives in.

@Solidigm, @MinIO, & @Intel partnered to design and benchmark S3-compatible object storage, with a focus on performance and concurrency scaling. This white paper goes deeper into what this benchmark means for AI builders. Check out the full results! solidigm.com/products/techn…

And we're out! Thanks to everyone for tuning in.

RAG pipelines don't have to be DRAM-constrained. SSD-offloaded vector retrieval delivered 65% lower memory footprint and 14% higher query throughput vs DRAM-bound alternatives while maintaining 99%+ recall. Storage is earning its place as an active AI memory tier. #AIIFD5

Well, what if your system could remember instead of recompute? With KV cache offload to NVMe, @Solidigm and @MetrumAI cut context switching from 39 seconds to 1.43 seconds! #AIIFD5 #AIInfrastructure #SSD

As context windows grow, GPU memory can't keep up. Long-context AI inference forces systems to recompute the same work over and over, burning cycles and slowing response times. At scale, that's not sustainable. #AIIFD5 #Inference #GenerativeAI

Forget bolting KV cache onto enterprise storage. MemKV strips it down — KV-addressed, RDMA-native, NVMe-backed. Purpose-built for the way AI inference actually works. #AIIFD5 #GenerativeAI #SSD

The solution? @MinIO MemKV is built like memory, not enterprise storage. GPU registered memory + raw NVMe blocks + RDMA. No metadata database. No overhead. Just fast, direct KV access via io_uring. #AIIFD5 #ObjectStorage #MinIO

The problem? At just 8 concurrent 32K sessions, KV cache requirements blow past GPU HBM capacity. At 32 concurrent sessions? You're at 4.3x the HBM ceiling. The memory wall is real, and it's happening now. #AIIFD5" target="_blank" rel="nofollow noopener">techfieldday.com/appearance/sol… #AIInfrastructure #Inference #AIIFD5

Modern object storage and high-density SSDs are helping organizations build scalable, efficient AI factories. Looking forward to sharing insights alongside the @MinIO team! techfieldday.com/appearance/sol…