Understand this story
On October 19, 2026, SemiAnalysis predicted that global data center construction will accelerate in 2027, with new capacity reaching 50GW, equivalent to the entire cumulative global market before 2024 being built within a single year. Its data center industry model shows that 60GW is currently under construction, of which more than 24GW is in the late stages of construction, with most to be delivered in 2027; so far in 2026, data center leases worth slightly less than $1 trillion have been signed, providing funding support for construction in 2027 and 2028. The report also notes that prefabricated modular designs have compressed construction timelines to under 10 months, and holds that the power problem has been largely resolved through means such as self-built generation, with the new bottleneck in 2027 and 2028 potentially shifting to memory and logic chips. At present, there are no subsequent factual developments regarding this prediction.
AI-generated from public reports
Latest development · Report
SemiAnalysis predicts 50GW of new global data center capacity in 2027Developments
1 confirmed developments
SemiAnalysis predicts data center construction will accelerate in 2027
Source:SemiAnalysis
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 19
SemiAnalysis
SemiAnalysis predicts 50GW of new global data center capacity in 2027
SemiAnalysis predicts that global data center new capacity will reach 50GW in 2027, equivalent to the entire cumulative global market before 2024 being built within a single year. Its data center industry model shows that 60GW is currently under construction, of which more than 24GW is in the late stages of construction, with most to be delivered in 2027; so far in 2026, just under $1 trillion in data center leases have been signed, providing funding support for construction in 2027 to 2028. The report also mentions that prefabricated modular design has compressed the construction cycle to under 10 months, and holds that the power problem has been largely solved through means such as self-built generation, with the new bottleneck in 2027 to 2028 potentially shifting to memory and logic chips.