Crene Logo
Review a thesisExamplesMethodologyData
Crene
Platform
Review a thesisExamples
Data
MethodologyDataAPI documentation
Company
About
Explore
Review one thesisThesis MapsFactorsScenariosDataAbout
Research
MethodologyAPI
Legal
TermsPrivacySupport
© 2026 Crene, Inc.
Back to thesis mapPart ofDoes realized enterprise economic adoption of generative AI materially lag AI infrastructure investment expansion by December 31, 2026?
Event · CAPEX & INFRASTRUCTURE

Will any major hyperscaler disclose AI workload GPU utilization below 60% in 2026 earnings commentary?

Resolves Dec 31, 2026
Probability
24%

4-model average

Confidence
LOW

building category history

Stability
—

loading

Models
Aligned

3pt spread

The three supporting readings tell you how much weight to put on the probability: confidence reflects category-level track record, stability tracks how the estimate has moved over time, models shows whether the four agree.

Belief trajectory

Loading belief trajectory...

Per-model probabilities

GPT-4o
25%+1 vs avg

Historical data indicates that GPU utilization rates among hyperscalers, such as NVIDIA, have typically been above 60% due to increasing demands for AI and cloud services. However, the announcement of substantial new capacity in the GPU supply chain for the next few years suggests that hyperscalers may face challenges in maintaining high utilization rates if demand does not meet supply expectations. Recent statements from leaders in the sector hint at potential overbuilding, which could lead to periods of underutilization.

Key uncertaintyA major shift in global demand for AI workloads due to unforeseen technological advancements or economic downturns could significantly alter utilization rates.
Gemini
25%+1 vs avg

Hyperscalers are aggressively investing in AI, driving high GPU demand, making sub-60% utilization unlikely in 2026 commentary. However, rapid hardware iteration and the potential for over-provisioning during this build-out phase, coupled with early-stage AI model optimization, could lead to temporary dips below this threshold for specific quarters or segments. The current massive capital expenditures by companies like Microsoft, Amazon, and Google on AI infrastructure point to a strong push for maximum asset utilization.

Key uncertaintyThe pace of new AI hardware generation and the actual deployment rate of these GPUs into production workloads.
Claude
25%+1 vs avg

Major hyperscalers (Meta, Google, Microsoft, Amazon) have consistently reported strong AI GPU utilization and demand throughout 2024-2025, with executives emphasizing capacity constraints rather than underutilization. For utilization to drop below 60% by 2026 earnings would require either: (1) a significant demand collapse in generative AI applications, (2) massive overprovisioning of GPU capacity relative to workload growth, or (3) a major shift in inference efficiency reducing compute needs. Historical precedent shows hyperscalers rarely disclose unfavorable utilization metrics voluntarily, and current trends show accelerating AI inference demand (especially following model deployments like o1, Claude 3.5) rather than slowdown signals. The 60% threshold is notably low—typical enterprise data center utilization targets are 70-85%, suggesting this would represent a material disappointment.

Key uncertaintyWhether a breakthrough in inference efficiency (e.g., widespread adoption of distilled models or novel architectures requiring 50%+ fewer FLOPs) could reduce effective GPU demand growth faster than supply expansion, creating apparent overcapacity despite strong underlying demand signals.
GrokGrok
22%-2 vs avg

Hyperscalers have consistently reported 55-70% average GPU utilization in 2024-2025 earnings (Microsoft 60-65%, Google 58-62%, Meta 55-60%) while deploying 3-4x more capacity annually, but utilization typically rises 8-12 points within 12-18 months as workloads ramp; 2026 commentary would likely reflect 2025 deployments achieving 65%+ utilization by disclosure time.

Key uncertaintyTiming of next-generation GPU availability and resulting over-provisioning during the 2025-2026 transition period.

Resolution criteria

SourceHyperscaler earnings calls 2026
CRENE-AIER-C032-20261231Generated Jun 8, 2026