PCR Tech and IT retail, distribution and vendor news
Hammer Distribution and AMD are urging UK organisations to rethink their AI infrastructure plans as grid connection delays continue to slow data centre expansion. The companies argue that a CPU‑first approach, rather than defaulting to accelerator heavy designs, may be the only viable route for many businesses facing strict power limits and long waits for new capacity.
Regulators have repeatedly warned that grid availability is now the defining barrier to AI growth. Ofgem and the National Energy System Operator’s December 2025 reforms cleared hundreds of gigawatts of stalled projects from the queue and shifted the system toward prioritising developments that are both technically ready and energy‑efficient. Government papers published late last year described grid access as the single biggest blocker to new data centre investment.
Against this backdrop, Hammer and AMD are drawing attention to the CPU’s role in determining whether AI systems operate efficiently within constrained power envelopes. They note that as inferencing becomes the dominant mode of AI deployment, the CPU increasingly governs throughput, utilisation and compliance with emerging European reporting rules. Adam Blackwell, Hammer’s Director of AI, Server and Advanced Technology, said the sector is entering a phase where “power availability and system efficiency” are shaping what can realistically be built.
AMD’s technical guidance suggests that CPUs can support inference for models up to 20 billion parameters where latency requirements allow, a threshold that covers many enterprise workloads such as document processing and retrieval‑augmented search. The companies argue that shifting these tasks to CPUs can reduce reliance on accelerators, lower overall power draw and allow organisations to deploy AI on existing infrastructure rather than waiting for new grid connections.
The timing aligns with the European Commission’s Energy Efficiency Directive, which introduces mandatory reporting on data centre performance and emphasises measurable “useful work” per watt consumed. Operators are expected to face increasing scrutiny over utilisation and energy efficiency as part of future investment decisions.
The UK’s situation mirrors developments elsewhere. Ireland’s regulator, the CRU, recently confirmed new rules requiring data centres to provide onsite dispatchable generation and secure the majority of their demand from additional renewable capacity, which is further evidence that power availability, rather than compute ambition, is becoming the defining constraint on AI infrastructure across Europe.
