Solidigm targets the intelligence layer as agentic inference pushes storage to center stage

2026-07-08 · Source: AI – SiliconANGLE · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

Solidigm, a trademark of SK Hynix NAND Product Solutions Corp., is positioning high-capacity solid-state drives (SSDs) as the "intelligence layer" for agentic AI inference, a critical shift from traditional model training. Greg Matson, SVP at Solidigm, highlighted at RAISE Summit 2026 that storage, once an afterthought, is now central to AI infrastructure, extending GPU memory. Hyperscalers began replacing legacy storage 12-18 months ago to ensure GPUs, the most expensive infrastructure component, remain 100% utilized. A 15-word prompt can generate 40,000 tokens, equating to 5-10 gigabytes of context data, pushing storage needs into petabytes for enterprise AI clusters. Solidigm offers drives up to 122 terabytes per unit, including the industry's first cold-plate-cooled enterprise SSDs for fanless Nvidia GPU servers, anticipating a future where all AI racks are liquid-cooled. This new architecture, moving from PCIe Gen 4 to Gen 5 and then Gen 6, is essential for sovereign AI deployments and efficient token generation.

Key takeaway

For AI Architects designing next-generation inference infrastructure, your storage strategy must prioritize high-capacity, high-performance SSDs as a core "intelligence layer." You should evaluate solutions like Solidigm's 122TB drives and cold-plate-cooled options to maximize GPU utilization and optimize token economics, especially for sovereign AI deployments. Failing to integrate advanced storage will lead to costly GPU idle time and inefficient AI factory operations.

Key insights

Agentic AI inference elevates high-capacity SSDs to an "intelligence layer" essential for continuous GPU data feeding and efficient token generation.

Principles

• GPU utilization demands continuous data flow.
• Storage extends HBM and DRAM footprint.
• AI infrastructure is the new asset.

Method

Hyperscalers are undertaking rip-and-replace of legacy storage with high-capacity SSDs to prevent GPU idle time and optimize token generation economics, moving towards liquid-cooled, fanless rack designs.

In practice

• Deploy 122TB SSDs for petabyte-scale context.
• Adopt liquid-cooled SSDs for fanless GPU servers.
• Upgrade to PCIe Gen 5/6 for direct-attach SSDs.

Topics

• Agentic AI Inference
• High-Capacity SSDs
• AI Infrastructure
• GPU Utilization
• Liquid Cooling
• Sovereign AI
• Token Economics

Best for: MLOps Engineer, VP of Engineering/Data, Investor, AI Architect, Director of AI/ML, CTO

Related on AIssential

• 2026.22: Luceing Their Mind — The future of AI compute will diverge, with agentic inference prioritizing memory capacity and cost over raw speed.
• SambaNova Abandons Intel Acquisition, Raises Funding Instead — SambaNova secured $350M funding and partnered with Intel, launching the SN50 chip for economic agentic AI inference.
• AMD targets system-level AI infrastructure optimization as agentic workloads reshape enterprise compute — Agentic AI demands system-level optimization and heterogeneous architectures for cost-efficient, scalable compute.
• Rebellions Bets on Memory-Centric Architecture as It Weighs IPO Options — Memory-centric AI architectures and supply chain security are critical for large-scale LLM inference and market traction.
• Nvidia BlueField-4 STX adds a context memory layer to storage to close the agentic AI throughput gap — Nvidia's BlueField-4 STX architecture optimizes AI agent performance by accelerating KV cache access.
• A PhD pivot, a stint in VC and a rumoured Anthropic deal: why Europe is watching Walter Goodwin — AI inference demand is growing exponentially, creating a strategic opportunity for specialized chip development to challenge incumbents.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI – SiliconANGLE.