Via 247wallst.com
The AI infrastructure giant locks in priority access to ultra-high-capacity solid-state drives as demand for GPU cloud services accelerates
CoreWeave signed a multi-year agreement with Solidigm, the Intel-spun-off flash storage maker, guaranteeing priority access to solid-state drives as CoreWeave races to build out data centers fast enough to keep up with demand.
What the partnership actually looks like #
This isn’t a fresh relationship. CoreWeave and Solidigm have been working together for years, deploying QLC (quad-level cell) SSDs across CoreWeave’s GPU infrastructure. Jacob Yundt, a CoreWeave executive, has previously described the partnership as “incredible” for data center effectiveness.
The two companies are now co-engineering drives specifically optimized for AI workloads. That includes native direct-to-chip liquid cooling for SSDs, a technology that addresses the heat density that increases when cramming 122-terabyte drives into GPU clusters.
For context, 122TB in a single drive is enormous. Most consumer SSDs top out around 4TB. Enterprise drives typically max out at 30TB or so.
Why storage is the quiet bottleneck in AI infrastructure #
QLC SSDs offer a compelling balance for AI workloads. They’re denser and cheaper per terabyte than other NAND flash architectures, which makes them well-suited for the read-heavy data pipelines that characterize AI training. The trade-off is typically lower write endurance, but for workloads that are mostly about reading massive datasets, that’s an acceptable compromise.
CoreWeave’s bigger picture #
Earlier in 2025, CoreWeave secured an $11.9 billion commitment from OpenAI. In March 2026, the company added a multi-year partnership with Perplexity. CoreWeave’s acquisition of Core Scientific for approximately $9 billion gave it access to substantial power capacity. Core Scientific was previously a Bitcoin mining operation, meaning CoreWeave essentially converted crypto mining infrastructure into AI compute capacity.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our