Cloud Providers Face Compute Capacity Shortages
Infrastructure supply is limited through 2028, forcing companies to commit over $1 trillion in long-term leases.
Updated on Oct. 21, 2026 in Data Centers
Live Poll
Is now a good time for big tech companies to commit billions to AI infrastructure?
Major cloud providers including Alphabet and Amazon Web Services face significant compute capacity shortages, with some providers fully booked through 2028. To secure space, companies signed over $1 trillion in non-cancellable data center leases in 2026.
Why it matters
The supply crunch highlights the challenge of aligning long-term data center construction with the sudden surge in demand for agentic AI. As infrastructure requires years of planning, current compute limits are forcing tech giants to prioritize internal research and development over immediate market availability.
Alphabet announced 2026 capital expenditure guidance between $195 billion and $205 billion to address infrastructure needs. The industry has committed over $1 trillion in non-cancellable lease contracts for facilities scheduled to come online in 2027.
The players
Alphabet
The parent company of Google and DeepMind, currently scaling massive GPU clusters and custom silicon for large-scale AI model training.
Amazon Web Services
The dominant global cloud provider, currently managing significant infrastructure constraints as demand for its proprietary and third-party AI services exceeds capacity.
Sundar Pichai
The CEO of Alphabet who has confirmed that securing sufficient compute capacity remains a persistent challenge for the company.
The details
Data centers require multi-year lead times between initial planning and final delivery, making it difficult for providers to scale supply to meet current AI-driven spikes. To guarantee availability, firms are increasingly turning to non-cancellable lease contracts that lock in capacity years in advance. Google is currently managing compute shortages by balancing its 2024 infrastructure planning against the unforeseen emergence of widespread agentic AI—autonomous software that performs complex, multi-step tasks—which has rapidly consumed available resources.
Timeline
2024: Internal planning set baseline capacity levels that are now insufficient.
2025: Google began utilizing short-term leases to mitigate capacity issues.
2026: Alphabet issued $195-205 billion in capital expenditure guidance.
2027: Target year for tens of gigawatts of new data center capacity.
2028: Expected period through which Amazon Web Services remains sold out.
The Tech Race
The current infrastructure shortfall follows a period of rapid, unplanned growth in demand for agentic AI applications. This race marks a departure from historic data center scaling, as providers commit over $1 trillion in long-term contracts to prevent losing their competitive edge in compute.
Cloud computing customers will likely face restricted capacity or longer wait times for high-end AI compute resources through 2028. Businesses relying on scalable cloud infrastructure should expect continued supply tightens as providers prioritize internal research and development over external demand.
The takeaway
The race for compute remains constrained by the physical reality of building power and cooling infrastructure, forcing companies to look ahead to 2028 and beyond. Readers should monitor upcoming quarterly earnings reports from major cloud providers to track the realized capacity versus initial expansion projections.
Further reading
Explore more analysis of infrastructure growth in the Data Centers section.
Live Poll
Is now a good time for big tech companies to commit billions to AI infrastructure?
