Castore built a brand on performance you can specify on a chart, then bought a company whose worth is that you cannot. The Manchester sportswear group has taken a 75% controlling stake in Grenson, the Northamptonshire bootmaker that has made boots and shoes since 1866. Terms were not disclosed. The move looks like nostalgia and works like a hedge. The wager is that owned, hand-built craft is the one asset a design algorithm cannot reproduce, and that it gains value as AI drives the cost of competent design toward zero.

Castore’s entire proposition is the kind of value that data and AI are built to optimise. The company sells technical sportswear on engineered margins, and has turned moisture-wicking fabric and marginal-gain marketing into kit deals with Newcastle United, Wolverhampton Wanderers, England’s rugby teams and a foothold in Formula 1. A funding round in late 2023 valued the business at close to £1bn, with Andy Murray among the backers. Every input in that model can be measured, tested and, increasingly, generated, which makes execution speed the moat rather than any secret a rival cannot learn. That is the problem a generative system is built to solve.

Grenson is the opposite kind of asset, and that is the point. William Green began making shoes by hand in a Rushden loft in 1866 and built one of the first factories in the world to adopt the Goodyear welt; the “Grenson” name, a contraction of Green and Sons, became one of British shoemaking’s earliest registered trade marks. The firm shod British troops through both world wars and has occupied only three factories in 160 years. None of that can be accelerated: a welt is a sequence of hand operations that takes the time it takes, and the unbroken continuity is itself the product. You cannot raise the throughput of provenance.

The reason this counts as a hedge is that design is losing its scarcity. McKinsey puts the prize from generative AI at $150bn to $275bn in added operating profit for the apparel, fashion and luxury sector within three to five years. Up to a quarter of that, by the same analysis, comes from design and product development, the part of the business that used to require taste and time. The mechanism is concrete: a model turns a sketch or a mood board into dozens of rendered, production-ready looks, collapsing the weeks and the specialist labour that separated a strong design house from an average one. When competent design is abundant and cheap, it stops being a moat and becomes an input anyone can buy with a prompt. If that curve holds, a brand that competes on design alone is competing on the thing AI has made most plentiful.

An algorithm can draw the boot; it cannot own the century behind it.

This is the second time in a year Castore has bought heritage it could never have manufactured into being. It took Belstaff from INEOS in August 2025, and is assembling a portfolio of British premium names whose history is intact but whose commercial execution has stalled. The financial backdrop sharpens the logic. Castore grew revenue to £334.6m last year while its post-tax loss widened to £40.3m, the profile of a company spending hard to scale a sportswear business that runs on exactly the optimisable margins AI is poised to compress. Tom Beahon has said plainly that he will accept “short term impact in profitability” for “long term value creation”. Grenson is the counterweight to that spending: an asset whose worth depends on its refusal to scale.

The argument has an obvious weak point, and it deserves its strongest form. Grenson does not make every shoe by hand in Northamptonshire; it has manufactured in India for decades, the English factory running the premium welt and the Indian one supplying hand-stitched volume. If the craft is already mostly offshored, then “owned analog craft” is a logo with an old date attached, and a marketing budget paired with a generative model can fake the signals of heritage as cheaply as it drafts a tech-pack. The answer is that faked heritage is exactly the signal AI is about to make worthless, because abundance is what destroys a signal’s value, and verifiable owned craft is the part that stays scarce when imitations are free. The genuine risk is not that AI copies the boot. It is that Castore, pressed to justify the price, engineers the welt out of Grenson the way it engineered performance into its kit, and keeps the date while discarding the thing that made it worth owning.

The outcome now turns on what Castore does with the workshop, not on what it paid for it. When the deal closed, Grenson’s registered office moved from Rushden to Castore’s Manchester headquarters, the craft’s address quietly following the data brand home. If Castore keeps the welting where the value sits and resists treating a slow workshop as a margin problem, it will own the rarest thing in a commoditising market: an input no model can print. Do the reverse, and it will have paid a heritage premium for a brand it then commoditises, the single outcome the thesis was built to avoid. The deal is not a verdict on whether craft beats algorithms; it is a bet on whether a company this good at optimisation can leave one asset un-optimised.