Investment research firm Bernstein has issued a note arguing that computing power — once a scarce, premium resource — is increasingly behaving like a commodity, with price elasticity, standardised units and global trading dynamics. The analysis suggests that the rapid expansion of data centre capacity, combined with a flood of AI-specific chips from manufacturers such as Nvidia and AMD, is eroding the scarcity premium that cloud providers and AI companies have enjoyed.
Bernstein draws parallels to historical commodity cycles: early adopters and producers capture outsized margins, but as supply catches up with demand, margins compress and the product becomes fungible. The note points to falling per-unit costs for cloud compute instances and the emergence of secondary markets for GPU time as evidence that the transition is already underway. For UK investors, this raises questions about long-term growth assumptions baked into tech-heavy portfolios and pension fund allocations.
The FTSE 100 dipped 0.3% in early trading on Monday, with tech and infrastructure stocks under mild pressure. The broader FTSE 250 was flat. Among individual names, London-listed data centre operator Digital Realty eased 1.1%, while semiconductor distributor RS Group slipped 0.8%. Analysts at Bernstein caution that if compute truly commoditises, the premium valuations assigned to cloud hyperscalers and AI chipmakers may need to be revised downwards, particularly for those with limited pricing power.
Sector context matters: the UK's £2.5 trillion pension industry has increasingly allocated capital to US tech giants and domestic AI infrastructure funds. A re-rating of compute assets could ripple through defined-contribution schemes and SIPPs. Bernstein does not name specific stocks to avoid or buy, but the note underscores the importance of differentiating between companies that own proprietary compute stacks and those that simply resell standardised capacity.
For now, the compute market is not yet a full commodity — long-term contracts and specialised workloads still command premiums. However, Bernstein expects further convergence, with a review of pricing models likely as more capacity comes online. UK investors should monitor how cloud providers adjust their pricing structures and whether AI model developers begin to treat compute as a variable cost rather than a strategic asset.