Archive correction — 14 October 2026. The previous version understated existing UK compute policy, used the former AI Safety Institute name and presented unsupported capacity and loss estimates. This version separates published policy from remaining delivery questions. Original publication dates are retained.
There is a UK compute roadmap
The UK Compute Roadmap was published in July 2025. It includes public compute investment and plans to expand access, with support for an Edinburgh supercomputer and the AI Research Resource. Investment plans and capacity targets are distinct from completed, accessible systems. They nevertheless contradict an account suggesting that UK policy is concerned only with regulation and has no compute plan. Roadmap and current updates.
ObvioTech analysis: the useful test is delivery. Which capacity is operational, who can access it, on what terms, and what workloads can it support? A roadmap does not answer every question about cost or resilience, but it must be acknowledged before arguing that policy is insufficient.
Semiconductors: compare scope, not just budgets
The May 2023 National Semiconductor Strategy emphasises UK strengths including research, design and intellectual property, and compound semiconductors. Its programme is not a like-for-like equivalent of every manufacturing incentive in the US or EU. Comparing headline totals without duration, eligible activities and public/private funding definitions can mislead. National Semiconductor Strategy.
A design business and a fabrication plant perform different roles. Access to manufacturing can be a dependency, but this article does not quantify a universal loss from disruption or establish that every UK technology service depends on one foreign facility. A specific supply-chain assessment requires more detail than a national headline.
Security evaluation and physical capability are distinct
The AI Safety Institute was renamed the AI Security Institute in February 2025, with a stated focus on security risks. It should not be represented under its former name as proof that the UK has chosen regulation instead of infrastructure. Government announcement.
ObvioTech analysis: evaluation, compute access and commercial deployment can be complementary. Their effectiveness should be examined separately. It is too strong to claim that a country cannot regulate a service unless it manufactures or owns every component used to provide it.
Questions for an infrastructure review
- Capacity: distinguish electrical capacity from compute performance and capacity available to a customer.
- Energy: check the site’s power agreement, connection timetable and operating constraints.
- Access: assess procurement terms, eligible users, queueing, portability and ongoing cost.
- Dependencies: document hardware, software, suppliers, jurisdictions and exit arrangements.
- Delivery: compare targets with commissioned assets and demonstrated services.
This is an ObvioTech assessment framework. It does not claim that all London data-centre development has stopped, that every project waits a decade, or that a particular energy forecast is certain. Those questions need geographically and temporally defined evidence.
A bounded conclusion
Compute and semiconductor access matter to technology businesses, but inevitable national decline by 2030 is not established by the cited evidence. Track delivery, access and dependencies, explain uncertainties and update conclusions as conditions change. Neither an announcement nor a pessimistic scenario should be presented as an observed outcome.
