Era4 links UK financial sovereignty to domestically owned AI infrastructure
Era4 Ltd. is building UK-owned AI infrastructure with self-generated power to retain control and revenue in Britain.
Karl Havard said Era4 is building a network of UK-owned AI facilities powered by energy it produces itself. He described the approach as a way to give the UK more control over its AI technology stack and to ensure the domestic economy benefits from AI spending. Havard spoke with theCUBE Research’s Dave Vellante and Amit Eyal Govrin for the “Sovereign AI: From Strategy to Reality” video series.
Era4’s live Chesterfield facility uses liquid-cooled Nvidia Corp. B300 systems supplied by Hewlett Packard Enterprise Co., with World Wide Technology LLC assisting with installation, Havard said. The operating model extends from land and power to ownership of the AI hardware deployed on its sites.
Havard said buying technology from an overseas supplier does not necessarily mean giving up control. He compared the distinction to owning an American-made motorcycle, which sits in his garage and is under his control. Still, Era4’s stack includes HPE Private Cloud AI, which brings software licensing dependencies that must be considered alongside operational control. The company is also pursuing domestic and European partnerships above its infrastructure layer, including a June agreement with Cosine to explore support for Lumen Sovereign, a UK-developed language model.
Havard said no single sovereign nation has pure control across the whole stack. He said Era4 does not aim for that “utopian point” because it would create an island and lose the benefits of collaboration.
For Era4, financial sovereignty goes beyond keeping equipment inside national borders. Havard said domestic service revenue available for reinvestment can support a wider UK ecosystem, while payments to overseas providers follow a different path. He said the model benefits UK gross domestic product and the economy not only through tax, because revenue stays and is invested in the community.
The shift toward inference adds another financial consideration as organizations pay to use AI models in production, Havard said. Beyond managing token bills, countries hosting infrastructure need to consider how commercial arrangements distribute the proceeds. He said there is further work to agree on a proportion that model makers keep and a proportion that hosting organizations in the host nation keep.
Era4’s planned redevelopment of a former steelworks in Newport, Wales, puts the economic argument in a local setting. Havard said the company owns a 230-acre site and intends to use about 20% for AI infrastructure. The plan includes a vocational tech college, sports facilities and a swimming pool requested by Newport City Council, which Era4 plans to build on the council’s behalf, he said. It also includes conservation and research areas for schools and universities.
Direct-to-chip liquid cooling would allow data center heat to be reused in the proposed swimming pool, according to Havard. He said Era4 wants its blueprint to show that AI infrastructure can bring benefits to host communities, including lower energy tariffs and subsidized council tax, and he hopes others will follow a similar approach.