Data Centers to Face Excess Capacity by 2028
As power procurement ramps up, infrastructure investment may soon outpace the cooling and compute demand of the AI era.
Updated on Oct. 23, 2026 in Data Centers
Live Poll
Will the rapid expansion of data centers lead to an excess supply of capacity by 2028?
The data center sector may transition from its current shortage to excess capacity by 2028. While GPU supply is currently easing, industry research indicates that power availability remains the primary bottleneck for infrastructure development.
Why it matters
The shift highlights a critical disconnect in the AI infrastructure pipeline, where facility expansion and power procurement could soon outpace demand for compute capacity. This trajectory could fundamentally reshape the economics of data center build-outs over the next few years.
GPUs are now easier to source than electricity, which currently serves as the limiting factor for data center scaling. While chip oversupply exists today, capital expenditure is flowing heavily into semiconductor fabrication and power-intensive facility expansion.
The players
Jeremie Ontiveros
Data center and energy research lead at SemiAnalysis, a firm specializing in the semiconductor and AI supply chain.
The details
Data centers require a massive, steady flow of electricity to operate the hardware that trains and runs AI models. As operators expand their energy footprints through 2027 to address this, the build-out rate may eventually overshoot the actual demand for processing power. This shift suggests that the current supply-chain constraints, which favor hardware manufacturers, will likely reverse as physical data center space becomes commoditized.
Timeline
2026-2027: A period of strong energy-related infrastructure expansion.
2028: The projected timeline for a shift toward excess industry capacity.
The Tech Race
The current rush to secure power for data centers mirrors previous infrastructure boom-and-bust cycles in telecommunications and enterprise storage. By scaling capacity aggressively, developers are betting on sustained demand that may ultimately face a correction by 2028.
Investors and infrastructure managers should watch for a shift in hardware availability as energy-constrained sites become the new standard for data center viability. Those waiting on compute capacity may find lower costs or higher availability as the market moves toward an oversupply by 2028.
The takeaway
The industry's current focus on power procurement is the leading indicator of a potential capacity surplus in the late 2020s. Monitor capital expenditure reports and energy grid interconnection queues through 2027 to verify if the pace of construction tracks with projected AI compute demand.
Further reading
For broader context on the infrastructure supporting current compute trends, see our coverage of Data Centers.
Source note: This article includes information reported by TokenPost.
Live Poll
Will the rapid expansion of data centers lead to an excess supply of capacity by 2028?
