UK firms lose £3.8bn to unexpected tech infrastructure costs

A CEBR study commissioned by Wasabi Technologies finds unpredictable provider charges are cutting UK AI investment, hiring and product development.

A brightly lit data center aisle is lined with rows of server racks featuring prominent blue network cables, leading to closed server cabinets at the far end.

UK businesses are absorbing £3.8 billion a year in technology infrastructure costs they did not anticipate, according to new economic research from the Centre for Economics and Business Research (CEBR), commissioned by cloud storage provider Wasabi Technologies. The study, which surveyed 690 senior business and IT decision-makers across eight countries in July and August 2026, attributes the overspend to opaque, usage-based or otherwise difficult-to-forecast provider charges rather than deliberate planned expenditure.

Globally, the total figure reaches £47.4 billion across the UK, US, Japan, Italy, Germany, France, Australia and New Zealand, suggesting the problem is structural rather than specific to British procurement practices.

AI infrastructure is the biggest single culprit

AI and machine learning infrastructure accounts for £1.2 billion, or roughly 32%, of the UK overspend, more than any other technology category. That figure also represents the layer businesses find hardest to predict: 22% of UK firms said AI infrastructure costs were difficult to forecast, a higher share than any other part of the technology stack. The specific cost drivers cited include AI platform and model licensing fees (66% of AI overspenders), integration and model operations expenses (43%), and usage-based charges covering tokens and inference calls (36%).

Owen Good, head of economics advisory at CEBR, said the research exposes a compounding dynamic. "The technology generating the most unplanned costs is also among the investments businesses are cutting to pay for them," he noted. In practice, 38% of UK organisations said technology, data and AI hiring was the budget line most likely to be sacrificed when infrastructure costs overran, followed by software investment (37%) and workforce training (34%).

CEBR modelled a recovery scenario: if the full £3.8 billion were redirected toward productive investment, the analysis suggests it could support £1.4 billion in R&D and AI spending, £800 million in workforce training, 8,235 additional jobs and a long-run GDP uplift of £1.4 billion. These are modelled projections, not forecasts, and carry the uncertainty typical of counterfactual economic analysis.

Switching costs compound the problem

Businesses seeking relief by moving providers face a secondary set of barriers. The survey found that 84% of UK respondents experienced delays linked to a lack of suitable hardware or compute capacity in the past 12 months. Among those who considered a supplier switch, 91% incurred additional costs from fragmented data environments, and 85% faced delayed access to data needed for a project at least once in the same period.

Marty Falaro, president and COO of Wasabi Technologies, argued the core problem lies in pricing model design. "Businesses need to know what it will cost to store, access and move their data before they commit to an AI project, not after the bill arrives," he said.

Market context and competitive landscape

The research lands at a moment when egress fees, per-token inference pricing and opaque licensing models are attracting growing scrutiny from enterprise buyers. Hyperscaler pricing complexity has become a competitive opening for independent cloud storage and GPU-cloud vendors, including Wasabi, Cloudflare R2, Backblaze B2 and a cluster of regional object-storage providers, all of which advertise predictable, flat-rate pricing as a differentiator.

Regulators are beginning to pay attention too. The UK's Competition and Markets Authority has examined cloud-switching barriers in its ongoing cloud-services market study, with concerns centring on egress charges and data portability that map directly onto the friction the CEBR research describes. At the EU level, the Data Act, which entered force in 2024 and whose switching provisions are progressively applying, explicitly requires cloud providers to eliminate excessive data-transfer fees by 2027.

The research was commissioned by Wasabi, which has a commercial interest in highlighting the costs associated with rival hyperscaler pricing. Readers should weigh the findings accordingly, though the independent CEBR methodology and named survey sample lend the headline numbers reasonable credibility as a directional signal.