Crusoe's Energy-First Vision for the AI Cloud | Chase Lochmiller | RAISE Summit 2026

Crusoe Business Model & Vertical Integration

• Crusoe operates as a vertically integrated AI infrastructure platform spanning from "electrons to tokens."
• Core Capabilities:
  • Energy & Land: Develops stranded energy sources and land acquisition to power compute infrastructure.
  • Data Center Construction: Designs next-generation data centers capable of housing large-scale training and inference clusters (e.g., Abilene, Texas).
  • Crusoe Cloud: A high-performance managed AI cloud platform for deploying and managing GPU clusters.
  • Developer Tools: Provides managed inference services and endpoints to help innovators scale applications cost-effectively.
• Strategic Philosophy: Prioritizes an "energy-first" approach, moving compute to energy sources rather than the traditional model of bringing energy to compute hubs.

Infrastructure Scale & Geographic Shift

• Strategic Pivot: AI workloads have lower latency sensitivity than legacy internet applications, enabling data centers to relocate from dense hubs like Northern Virginia to new energy-rich geographies.
• Capacity Comparison (End of 2024):
  • Total historical capacity in Northern Virginia: ~4.5 GW.
  • Crusoe contracted capacity: >5 GW.
  • Crusoe development pipeline: >40 GW.
• Market Impact: A single company (Crusoe) is building a power footprint comparable to the entire historical Northern Virginia corridor.
• Construction Speed (Abilene Project):
  • Construction started: June 2024.
  • Time to first megawatts: 11 months.
  • Competitive benchmark: Next fastest bid was 2.5 years.

Hardware & Cooling Trends

• Liquid Cooling Adoption: Shifted from a niche solution for experimental supercomputers to the industry standard for large-scale AI infrastructure due to high-density chip architectures (e.g., Blackwell).
• Supply Chain Resilience:
  • Cultivated a "mountaineer" mindset (planning for failure and having backup plans) to manage supply chain bottlenecks.
  • Utilized in-house manufacturing of critical electrical components to bypass external supply chain delays during the Abilene build.

Acquisition: Atero (Inference Technology)

• Acquisition Focus: Completed acquisition of Atero to strengthen the inference and orchestration layer.
• Technology Integration: Embedded "Crusoe Memory Alloy," a distributed memory caching layer.
  • Function: Stitches together HBM, DRAM, and VME object storage to manage KV caches across memory layers.
  • Objective: Optimize GPU utilization by reducing time spent waiting for data, thereby accelerating token generation and improving cost efficiency.
  • Strategic Rationale: Monetizing hardware investments by transitioning focus from training to inference, creating value across multiple market cycles.

Product Strategy: Crusoe Spark

• Concept: A strategy to deploy inference nodes as distributed "AI factories" (self-contained units).
• Differentiation from Gigawatt Campuses:
  • Gigawatt-scale campuses are optimized for training.
  • Inference requires distributed, smaller clusters (10-50 MW) closer to end-users to deliver value.
• Energy Access: Enables access to decentralized energy sources that are difficult to aggregate for single gigawatt sites.

Forward-Looking Statements & Outlook

• Market Phase: The industry is transitioning from a phase of "announcements and press releases" to one of "execution" over the next 24 months.
• Growth Driver: The commercialization of AI is defined by the shift from training models to serving inference (tokens), which represents the actual creation of consumer value.
• Risk Assessment: Not all announced complex infrastructure projects will be delivered due to execution challenges.
• Prediction for Raice 2027: Execution capability will be the primary metric of success rather than new announcements.
• Core Value: The company's culture emphasizes resilience, treating the inability to deliver on commitments as a critical failure point in the ecosystem.

Ecosystem Adoption

• Developer Engagement: High adoption of Crusoe's managed inference platform by developers (e.g., participants at the Akaton hackathon) seeking ease of integration and full-stack control.
• End-User Obsession: Product design is driven by the necessity of delivering tangible value to end-users rather than just infrastructure metrics.