Britain has committed £1.5 billion to its next wave of AI infrastructure
The question now isn’t whether the money will be spent — it’s whether it will be spent well.
Across the industry, the default answer to rising compute demand has been to build. Break ground on a greenfield site, lay years of planning and grid connection applications end to end, and hope the facility is ready before the workloads it was designed for become obsolete. It is a model built for a slower era of technology, one where a five-year construction timeline was an inconvenience rather than a competitive death sentence.
However, the default has proven to be outdated.
Samir Tabar, chief executive of the Nasdaq-listed AI infrastructure company WhiteFiber, has spent the past year proving there is a faster and cheaper way in.
Rather than building from scratch, WhiteFiber has focused on acquiring underutilised industrial sites that already come with the two things a data centre most desperately needs: substantial power capacity and proximity to major metro areas. Its flagship US project, a former textile mill in Madison, North Carolina, was bought for a fraction of the cost of comparable greenfield land and shell development, and converted into a hyperscaler-grade AI campus in a little over a year. The result was validated in the clearest way an infrastructure model can be: a ten-year, roughly $865 million colocation agreement with the European AI hyperscaler Nscale, one of the largest names in the sector’s European expansion.
The economics of that deal are the part the UK should be paying closest attention to.
Tabar’s retrofit approach compressed years off the delivery timeline, converted a fast-moving acquisition into a project institutional lenders were comfortable underwriting, and did it all while sidestepping the cost overruns and planning delays that have become endemic to greenfield data centre construction.
In an industry where the decisive competitive variable is shifting from the size of a campus to the speed at which it can be delivered at uncompromised quality, that is not a marginal advantage. It is the whole game.
The UK, by contrast, is still leaning heavily on the greenfield playbook, even as it pours fresh capital into AI compute through its AI Growth Zones and national supercomputer commitments.
Grid connection queues remain one of the single biggest bottlenecks to getting new capacity online, and the industrial landscape — much like America’s — is dotted with underused, well-powered sites that a retrofit-first strategy could bring into service in a fraction of the time.
If the UK wants its £1.5 billion to translate into operational AI capacity rather than years of planning applications, the retrofit model Tabar has proven out in North Carolina offers a template worth studying seriously.
