How We Design AI-Ready Data Centers | Hayden

Overview

The video explains a 3-phase design process for AI-ready data centers, highlighting how detailed requirements and iterative validation help avoid inefficient “false economies” and improve CAPEX/OPEX optimization.

Phase 1: Initial inquiry / discovery

• Clients typically provide a basic dataset and request a quick initial response.
• The team conducts at least one interview to gather missing details before deeper design work begins.

Phase 2: Installation requirements & engineering detail (longest phase)

• Covers installation phasing, modularization potential, and system airflow strategy.
• Focuses on:
  • Hot air recirculation allowances
  • Integration with chiller packs
• Typically takes ~6 to 10 weeks, often driven by client requirements.

Phase 3: Optimization & validation

• Performs CAPEX optimization and OPEX optimization.
• Includes ambient profiling (site environmental conditions used for design decisions).
• May use CFD services (computational fluid dynamics) to validate and confirm the selected design.

Key design drivers for AI-scale data centers

As AI demand increases and data centers grow:

• Hot air recirculation allowances increase
• Heat islands become a bigger concern

Design therefore expands to include:

• Equipment layout decisions
• Required raise/height (e.g., elevating equipment to manage airflow)
• Segregation of airflow passes (“one pass from another pass”)

Underlying principle: understand the application requirements well enough to optimize the full solution.

Review / process guidance emphasized

• Strong, transparent, collaborative design is important.
• An early kickoff meeting is preferred over relying only on email/data sheets with thermal loads or temperatures.
• Without deeper questions—especially if designers lack historic project experience—solutions may be built only to basic thermal requirements, creating false economies of scale.
• The team keeps the same execution project team involved “end-to-end,” so the original equipment specification and design intent carry through.
• The approach emphasizes a single-point-of-contact / flawless supply in execution model to reduce complexity across a complex build (whether delivering a “call”/partial solution or a full end-to-end data center).

Main speakers / sources

• The excerpt does not name a specific individual.
• It appears to be spoken by a data center design/execution team member; “Hayden” is referenced only in the video title.