The launch of DeepSeek’s R1 model initially caused concerns among investors, with some analysts questioning the necessity of massive AI infrastructure investments.
However, experts suggest that cheaper and more efficient AI models could ultimately accelerate market growth rather than disrupt it.
UBS analysts revised their forecasts, now expecting data center sector revenues to grow 20% in 2025, compared to earlier estimates of 10-15% over three years.
Barclays analysts noted that DeepSeek’s efficiency claims could challenge current AI infrastructure spending, potentially impacting hyperscalers’ capital expenditure plans.
If AI models require less computing power, lower-tier, less energy-efficient data centers might see weaker demand.
Goldman Sachs projects a tightening balance between supply and demand in the data center market through 2026, with a potential correction from 2027 onward.
The uncertainty caused market volatility:
Schneider Electric (heavily exposed to data centers) lost over 9% on Jan. 27.
Siemens Energy dropped 20%, while ABB closed 6% lower on the same day.
Despite the initial shock, confidence returned as major tech firms, including Google and Meta, reaffirmed their multi-billion-dollar AI investments.
While DeepSeek R1 claims 20-30x less computing power per query, UBS analysts argue it requires more tokens per query, not significantly reducing power demand for inference.
Experts suggest a Jevons paradox effect—where efficiency gains lead to increased overall usage—could apply to AI, boosting data center demand instead of reducing it.
The demand for data centers is expected to grow, driven not only by AI but also by broader digital transformation needs.
Bruce Owen, EMEA president at Equinix, sees the rise of efficient models as an accelerant for AI adoption, reinforcing demand.
Ryan Cox of Synechron predicts that efficiency improvements will fuel more AI adoption, sustaining and even expanding the data center market.