Where Are All The Data Centers?

2026-05-12 · Source: Ed Zitron's Where's Your Ed At · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

The article argues that the reported rapid buildout of AI data centers, particularly gigawatt-scale facilities, is largely unsubstantiated. It highlights a significant discrepancy between public claims by hyperscalers like Microsoft, which reported adding 1GW of capacity per quarter in FY26, and the actual observable progress of construction projects. The author's investigation reveals that most announced data centers from 2023-2024 are either stuck in construction limbo, facing delays, or are only partially operational, often with a single phase completed. The article suggests that building a data center typically takes 18-36 months, with larger projects facing more delays. This slow pace implies that NVIDIA's reported shipments of millions of Blackwell GPUs (e.g., 3 million B200 GPUs, totaling 3.6GW IT load) far exceed the world's current operational data center capacity, suggesting many GPUs are likely in warehouses. This disconnect between reported capacity and physical reality poses significant financial risks for hyperscalers, AI companies like OpenAI and Anthropic, and their investors, as massive capital expenditures are not translating into revenue-generating assets.

Key takeaway

For investors and AI/ML directors evaluating hyperscaler infrastructure claims, recognize that reported data center capacity often reflects planned or partially built projects, not fully operational assets. Your investment decisions or compute procurement strategies should account for typical 18-36 month construction timelines and the high likelihood of delays. Scrutinize public statements and financial depreciation figures, as a disconnect between reported capacity and physical reality poses significant financial risks and could lead to compute scarcity.

Key insights

The rapid buildout of AI data centers is largely an illusion, with most projects facing significant delays and underreported operational capacity.

Principles

• Public claims of data center capacity often misrepresent operational status.
• Large-scale data center construction is inherently complex and time-consuming.
• Financial reporting can obscure true infrastructure deployment pace.

In practice

• Investigate actual operational status of announced data center projects.
• Cross-reference reported capacity with physical construction progress.
• Scrutinize financial depreciation figures for signs of delayed asset deployment.

Topics

• AI Infrastructure
• Data Center Construction
• Hyperscaler Capacity
• NVIDIA GPUs
• Financial Reporting
• Compute Scarcity

Best for: CTO, Entrepreneur, VP of Engineering/Data, Investor, Director of AI/ML, Tech Journalist

Related on AIssential

• The billion-dollar infrastructure deals powering the AI boom — The AI boom drives unprecedented infrastructure investment, creating complex financial and environmental challenges.
• Speed and Scale: How Today's AI Datacenters Are Operating Through Hypergrowth — AI infrastructure hypergrowth creates complex logistical and operational challenges requiring robust data management and automation.
• Capital, Not Compute, is the Real AI Bottleneck — Capital, not compute, is the primary bottleneck in AI infrastructure expansion, driven by financing and build-out timelines.
• SpaceX IPO Filing Ties AI Growth to Water Access — AI infrastructure growth faces critical physical limits, with water access now a key constraint for data center expansion.
• World’s Largest AI Datacenter — $100B Disaster — Building leading AI models now demands extreme infrastructure decisions, prioritizing compute, power, and rapid deployment.
• 📈 Data to start your week: The AI buildout — AI's exponential growth faces physical resource limits, sparking local resistance despite corporate environmental pledges.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Ed Zitron's Where's Your Ed At.