What is the AI Bubble and Will it Hit Data Centre Strategy?
The AI sector is experiencing what analysts call ‘bubble’ conditions, with investment levels pushing valuations to heights last seen during the dotcom boom of the 1990s.
UBS, the investment bank and financial services firm, has published research showing how this situation is shaping business discussions worldwide.
Defining the AI bubble
An AI bubble refers to companies developing AI technologies being valued well beyond their revenues or proven business models.
Investor enthusiasm drives stock prices and investment commitments based on projected rather than demonstrated returns.
The current wave of capital commitments highlights the concern.
UBS finds that technology companies have allocated capital expenditure and research budgets on a scale rivalling entire industries, with 2024 spending surpassing the combined research and development budgets of all listed European companies.
This surge reflects pressure on companies to secure long-term positions in AI markets.
At the same time, leading US technology companies generate a substantial share of national economic profit, giving them financial power to fund large-scale AI programmes even as questions remain about returns.
Sam Altman, CEO of OpenAI, sums it up speaking to Business Insider: “When bubbles happen, smart people get overexcited about a kernel of truth. Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes.
“Is AI the most important thing to happen in a very long time? My opinion is also yes.”
Valuation risks and data centre speculation
UBS argues that the core issue lies in the speculative nature of current AI use cases.
Many applications remain in early phases with commercial viability still untested. Revenues are projected into the future rather than realised today, creating a gap between spending and measurable returns.
Joe Tsai, Alibaba’s Cofounder, says to Business Insider: “I start to see the beginning of some kind of bubble... I start to get worried when people are building data centres on spec.”
For the data centre industry, this risk is acute.
Operators are building facilities and deploying computing infrastructure without revenue streams guaranteed to offset the huge costs.
The computational demands of AI workloads are accelerating demand for data centres, yet the business case depends on long-term adoption that is still uncertain.
Valuations today, UBS notes, mirror conditions during the dotcom era of the late 1990s, when internet companies attracted premium pricing based on anticipated rather than actual profitability.
The danger is that if AI returns fail to meet projections, corrections could reshape both technology company valuations and investment in supporting infrastructure.
Infrastructure, energy and competition
Several uncertainties stand in the way of projected AI returns.
UBS highlights unclear returns on capital invested in AI infrastructure, particularly in data centres and high-performance computing equipment.
Large facilities are being built in anticipation of revenue, yet there is no guarantee usage or profitability will match expectations.
“The bubble talk is completely wrong. AI will fundamentally change everything over the next five years.”
Energy supply constraints further complicate the picture.
Training and running AI systems require massive electricity consumption, creating infrastructure bottlenecks that could limit deployment speeds and increase operational costs. For data centre operators, energy availability and cost management become central risks in this bubble environment.
Rising global competition adds another layer of uncertainty. Regions investing heavily in research and development are creating their own AI ecosystems, which could erode the advantages of existing technology leaders and reduce profit margins assumed in current valuations.
Adoption timelines across traditional industries also remain unclear, with regulatory, technical and cultural barriers delaying commercial use.
Divided views on the bubble
Executives remain split on whether the bubble framing is accurate.
Speaking to Business Insider, Lisa Su, CEO of AMD, says: “The bubble talk is completely wrong. AI will fundamentally change everything over the next five years.”
By contrast, Thomas Siebel, CEO of C3.ai, says: “There is absolutely an AI bubble and it’s huge. The market is way overvaluing some startups.”
- The AI bubble is when AI companies are valued much higher than their current profits justify, driven by hype and big investments hoping for future success. This risks big losses if AI growth or profits don’t meet expectations, similar to the dotcom bubble, with challenges like high costs, energy use and global competition adding uncertainty.
For data centres, the stakes are clear. Investment decisions shaped by AI demand may fuel rapid expansion, but the risk of overbuild is real.
Whether AI delivers on its commercial promises or faces a correction, data centre infrastructure is locked in as a central part of the AI economy, carrying both opportunity and exposure in equal measure.
