State of AI Infrastructure: What's Powering AI Today | IREN, Mirantis & NVIDIA | RAISE Summit 2026

Transcript Analysis and Summary

Note: The initial segment of the provided transcript contains approximately 90 seconds of nonsensical, repetitive boilerplate regarding cake-making (e.g., "You can use any kind of flour... You can make this cake without any flour"). This content is excluded from the substantive summary below as it appears to be a glitch or placeholder error preceding the actual panel discussion.

Core Theme: The AI Infrastructure Race

• Fundamental Bottleneck: Energy and physical infrastructure availability are currently the primary constraints, outpacing supply relative to exponential demand for compute.
• Competitive Landscape: The "intelligence race" is defined by the ability to convert electrons into tokens (compute) faster and more efficiently than competitors.
• Three-Layer Architecture:
  • Physical Layer: Data centers, grid connections, and substations (identified as the critical foundation).
  • Compute Layer: GPUs, backend networking, and storage.
  • Software Layer: Orchestration and services that unlock value from the first two layers.
• Time-to-Power Metric: Reducing the latency between power availability and operational compute is a key differentiator for new infrastructure providers.

Strategic Moves and Partnerships

• Iron (Kent):
  • Vertical Integration Strategy: Owning both data centers and compute layers to eliminate third-party dependencies and accelerate capacity deployment.
  • European Expansion: Recently acquired a Spanish data center platform, marking its first official entry into Europe to leverage high demand from hyperscalers and North American enterprises.
  • Sovereign AI Focus: Targeting European markets driven by strict data sovereignty, security, and the need for localized sovereign compute.
• NVIDIA (Diane):
  • Platform Pivot: Repositioning the company as an "AI infrastructure platform" rather than just a chip manufacturer, emphasizing API layers for partners.
  • API Standardization: Launching the "AI-ready cloud ISV program" (with Mirantis as an inaugural member) to establish open standards for AI consumption, comparable to AWS's impact 20 years ago.
  • DSXOS Ecosystem: Developing a data center operating system (DSXOS) to maximize performance-per-watt and expose capabilities to developers like Mirantis.
• Mirantis (Alex):
  • Open Source Foundation: Leveraging 20 years of open-source dominance (Linux, Kubernetes, OpenStack) to ensure ecosystem interoperability and rapid innovation adoption.
  • Partnership Synergy: Collaborating with Iron and NVIDIA to create a reference implementation of a fully orchestrated AI platform using NVIDIA hardware.
  • Market Expansion: Moving beyond bare-metal customers (e.g., Microsoft) to target enterprise and AI-native customers seeking fully managed, orchestrated services.

Operational Challenges and Market Trends

• Infrastructure Lag: Building physical infrastructure (power, grid, cooling) takes significantly longer than the rapid cycle of digital innovation and model advancements.
• Demand vs. Supply Gap: Technological efficiency gains in GPUs often trigger surges in demand that further exacerbate the supply shortage of physical compute and energy.
• Software Orchestration: Essential for "democratizing" AI access, ensuring that diverse workloads can efficiently utilize available infrastructure without manual intervention.
• Speed of Innovation: Infrastructure must evolve to keep pace with weekly leaps in model capabilities (e.g., DeepSeek, OpenAI, Claude) across both closed and open-source sectors.

Forward-Looking Statements

• Hyperscaler Ambitions: Iron aims to become an independent major hyperscaler, offering flexible consumption models ranging from bare metal to systematic platforms.
• Holistic Ecosystem: Success requires a coordinated effort across energy providers, chip manufacturers, data center operators, and software builders; no single entity can succeed in isolation.
• Scalability Roadmap: The current six-gigawatt pipeline is viewed as a starting point for widespread infrastructure availability.
• Open Standard Adoption: The industry is moving toward a standardized "off-take API" that will dictate how AI is consumed and integrated across the ecosystem.