Global AI power consumption reaches 38GW

What AI 2027 Predicted

The scenario’s “KEY METRICS 2026” sidebar lists global AI power consumption reaching 38 gigawatts of peak power by end of 2026. This reflects the massive datacenter buildout depicted in the scenario, where infrastructure investment pushes AI compute capacity to unprecedented levels.

How We Track This

We monitor:

• IEA datacenter electricity demand reports and projections
• Goldman Sachs, SemiAnalysis, and other analyst estimates of datacenter power draw
• Individual datacenter project tracking (capacity announcements, construction timelines)
• Distinction between total datacenter power and AI-specific power consumption
• DOE and EIA reports on US electricity demand from datacenters

Current Evidence

Estimating global AI-specific power consumption is challenging because most reporting covers total datacenter power (which includes non-AI workloads like cloud computing, streaming, and traditional enterprise IT).

Total datacenter power (global):

• The IEA’s 2026 report estimates total global data-centre electricity consumption at 485 TWh in 2025 and projects approximately 950 TWh in 2030. These are annual energy quantities for all workloads, not AI-specific peak power. IEA

AI-specific peak power: The reviewed evidence does not provide a comparable global total. The earlier assumed AI workload share and 20-25 GW extrapolation are insufficiently supported to judge the 38 GW target. Annual energy divided by hours gives average load: 240 TWh/year is about 27.4 GW average, not peak power.

Stargate Project Pipeline: OpenAI’s Stargate project has expanded to approximately 10GW of planned capacity across multiple sites: Abilene TX (1.2GW, partially operational Sep 2025), five additional sites (~7GW announced Sep 2025), Michigan (1.4GW approved Dec 2025), plus a 4.5GW Oracle partnership (Jul 2025). Overlapping commitments must not be added twice, and planned capacity is not operational peak demand.

US datacenter power:

• Pew Research reports US datacenters accounted for 4% of total US electricity in 2024 (Pew Research)
• The DOE projects US datacenter energy demand to double or triple by 2028 (DOE)

Recent qualitative pressure: Reuters/Bloomberg reported in May 2026 that Microsoft may revise its 2030 clean-energy target because AI data center growth is raising electricity demand and changing the feasibility of pre-AI climate commitments. This supports the broader energy-constraint narrative, but it does not provide a global AI-specific GW estimate and should not be treated as direct evidence that global AI power is approaching 38 GW.

TrendForce’s May 2026 forecast materially strengthens the plausibility of high AI-related power demand by putting total installed global data-center power capacity around 155 GW in 2026 and projecting AI servers to surpass general-purpose servers in electricity consumption this year. Wood Mackenzie’s warning, reported by ITPro, adds qualitative support that grid and transmission constraints are becoming a binding bottleneck. This still does not directly measure global AI-specific peak power, so the 38 GW target remains “emerging” rather than “on-track.”

Bloom Energy’s 2026 Data Center Power Report adds mechanism evidence that power availability is shaping deployment strategy. The report describes power availability as a defining boundary on data center growth and says its research covered hyperscalers, colocation providers, utilities, and equipment suppliers. It supports the infrastructure bottleneck mechanism but does not directly estimate global AI-specific GW load.

Sources:

• Reuters/Bloomberg: Microsoft may shelve 2030 clean energy target as AI lifts power use
• IEA Electricity Mid-Year Update 2025
• Goldman Sachs: AI to drive 165% datacenter power demand increase
• Pew Research: US datacenter energy
• DOE datacenter energy report
• TrendForce: 2026 global data center installed power capacity forecast
• ITPro: Wood Mackenzie warning on AI data center power demand
• Bloom Energy: 2026 Data Center Power Report

Counterevidence & Limitations

• A comparable estimate must measure AI-specific peak power; annual averages and all-workload totals do not establish the distance to the target.

What Would Change Our Assessment

• Strengthen: Credible, consistently scoped estimates of operational global AI-specific peak power approach the target.
• Confirm: Comparable evidence reaches 38 GW by the deadline.
• Weaken: Comparable deadline-period estimates fall materially below the target; missing measurement does not establish a shortfall.

Update History

Date	Update
2026-09-06	Current evidence and assessment explanation reviewed. Clarified measurement scope, source interpretation and limitations.
2026-06-22	Bloom Energy’s 2026 Data Center Power Report described power availability as a defining boundary on data center growth, based on research across hyperscalers, colocation providers, utilities, and equipment suppliers. This supports the mechanism behind rising AI power demand without directly verifying the 38 GW target.
2026-05-25	TrendForce forecast total installed global data-center power capacity at about 155 GW in 2026 and projected AI servers to surpass general-purpose servers in electricity consumption this year. ITPro separately reported Wood Mackenzie’s warning that AI data-center power demand is outpacing grid buildout. This strengthens the plausibility of the 38 GW AI-power target, but neither source directly measures global AI-specific peak power. Confidence adjusted 0.50 → 0.55.
2026-05-11	Reuters/Bloomberg reported that Microsoft may revise its 2030 clean-energy target because of AI data center power growth. This supports the broader energy-constraint narrative but does not directly estimate global AI-specific GW; status and confidence unchanged.
2026-04-06	Tech Insider reports US data centers now consume approximately 176 TWh annually (4.4% of US total power), with 550 planned projects totaling 125 GW of capacity (Tech Insider). BCG projects a 50-80 GW capacity shortfall in the US by 2030 (BCG). JLL estimates 100 GW new capacity coming online between 2026-2030, requiring ~$1.2T investment. Google is building a dedicated 933 MW gas plant for its AI datacenter campus, signaling a shift away from grid dependency and climate commitments (Guardian). IEA projects datacenter demand >250 TWh in 2026, >300 TWh in 2027. If 176 TWh is current US consumption and AI is ~40% of datacenter power, US AI power alone could be ~8 GW average (~20 GW peak). Global AI power could plausibly approach 25-30 GW peak by end 2026. The 38 GW target remains ambitious but increasingly within reach. No status change.
2026-03-23	The Atlantic (Apr 2026 issue) reports OpenAI alone has announced plans for facilities requiring 30+ GW total, more than New England’s largest recorded demand (The Atlantic). White House event features xAI committing to 1.2 GW of dedicated power per supercomputer site (White House). UPI reports 10 largest US AI datacenter projects alone require 5.26 GW. Confidence adjusted 0.50 → 0.55 — planned capacity increasingly supports the 38 GW target, but gap between announced and operational capacity remains large.
2026-03-16	Programs.com (Mar 2026): AI workloads projected at 44 GW for 2026, non-AI at 38 GW — total 82 GW. If accurate, the 38 GW AI-specific target may actually be exceeded. Reuters reports largest US sites now consume >1 GW continuous. IEA projects >1,000 TWh datacenter consumption by 2026 under high-growth scenarios. Confidence adjusted 0.45 → 0.50 — newer estimates suggest the target is more achievable than previously assessed, though estimates vary widely.
2026-03	Industry estimates place AI-specific power consumption at 20-25 GW globally (end 2025). The 38 GW target for 2026 appears aggressive but datacenter buildout is accelerating rapidly.
2025-10	Anthropic’s 1M TPU deal (October 23) targets “more than 1 gigawatt” of capacity online in 2026. Q3 Big Tech capex at $114B/quarter with power availability cited as primary constraint — demand outpacing electrical infrastructure build-out.