Crusoe and Supermicro partner to build purpose-built neocloud AI infrastructure

This collaboration signals a shift away from legacy hyperscaler architectures that struggle with the high-concurrency demands of agentic AI. By utilizing disaggregated storage and liquid cooling, these neocloud providers are creating a specialized infrastructure stack that prioritizes low-latency inference over general-purpose compute. For the streaming and media ecosystem, this specialized hardware layer is essential for real-time AI video processing and personalized content agents that require massive throughput without the technical debt of traditional clouds. Watch for whether these purpose-built configurations can achieve the cost-per-inference metrics necessary to lure enterprise workloads away from established cloud giants.

Crusoe has emerged as a leading neocloud operator focused on AI workloads, and its infrastructure partnerships are expanding rapidly. In April 2025, Crusoe secured a $1.4 billion Series D funding round led by Founders Fund to build AI data centers powered by stranded natural gas, valuing the company at approximately $10 billion and positioning it as a direct competitor to hyperscalers for inference-heavy workloads. The company's strategy of pairing energy sourcing with purpose-built compute aligns with the Supermicro collaboration announced at the Open Storage Summit, where the focus on liquid cooling and disaggregated architectures addresses thermal and throughput constraints that traditional air-cooled facilities cannot manage at scale. Supermicro's financial trajectory highlights both the demand and the volatility in the neocloud hardware layer. In May 2025, Supermicro reported fiscal Q3 2025 net revenue of $4.6 billion, falling short of its initial forecast as customers paused purchases to evaluate the upcoming Blackwell GPUs, a delay that CEO Charles Liang attributed to platform transition uncertainty. The company pulled its $40 billion fiscal 2026 revenue guidance and lowered fiscal 2025 expectations to between $21.8 billion and $22.6 billion, citing tariff-related uncertainty and the Hopper-to-Blackwell GPU transition. More than 70 percent of Supermicro's quarterly revenue came from GPU-powered AI systems, underscoring how dependent the company's growth is on the same inference workloads that Crusoe and Vast Data are targeting with their disaggregated architecture. Vast Data's role in this partnership reflects broader investor confidence in disaggregated storage for AI inference. Vast Data raised $300 million at a $9.1 billion valuation in late 2024 to expand its all-flash storage platform into AI inference pipelines, signaling that the shared-everything architecture at the core of this collaboration is attracting serious capital. Meanwhile, Supermicro's Q3 results showed gross margin compression to 9.6 percent from 15.5 percent a year earlier, a margin squeeze that makes the case for AI inference infrastructure growth where hardware partners can optimize for specific workload profiles rather than competing on general-purpose server volume.