AI minister Kanishka Narayan told City A.M. on Wednesday 15 July that Britain is pursuing "global AI leadership, not Silicon Valley emulation", and that domestic capability would let the country "shape our own destiny" rather than depend on overseas providers1. John Harms of Quantexa, the British data-analytics company whose software banks use to trace networks of accounts, answered in the same piece that sovereignty requires governance control and data portability, and that hardware location without either amounts to vendor lock-in with extra steps.
Harms is describing a specific failure mode, not a rhetorical one. A dataset held on a server in Slough, processed on chips a foreign firm allocates, under model weights licensed on terms the licensor can revise, is British in geography and nobody's in practice. Portability is the test he proposes: can the customer take the data and the trained artefacts somewhere else, at what notice, and at what cost? Locate the machines wherever you like, the answer to that question is written into the contract.
This week's evidence sits closer to Harms than to the minister. Compute and laboratory access for the country's most-discussed materials company now run through Nvidia, Applied Materials and Tokyo Electron, and the British state's contribution to the round is a minority ticket with no board seat behind it. The Bank's own data recorded AI capturing a record 44% of UK equity value, so the capital is certainly here; the ownership of the layer underneath it is what neither side disputed on the record.
