Harworth Buys the Data Center Trade by Selling the Land, Not the Compute

A −3.7% half-year return, a full exit from residential, an unrecommended Peel offer — all wrapped around a land bank sitting on 0.8GW of secured power

HWG.L · Earnings Call · 2026-09-09

Three stories stapled into one half-year print

Harworth Group's first half of 2026 reads like three different press releases stacked on top of each other. The first is a poor financial print: CFO Kitty Patmore reported that “Total accounting return was -3.7%, driven primarily by a reduction in EPRA NDV per share from GBP 224.4 pence at 12/31/2025 to GBP 214.8 pence” — Kitty Patmore, Chief Financial Officer · 2026-09-09, with residential the culprit — softer house builder demand and market-driven construction cost inflation. Net portfolio value fell GBP 14.9m, split between a GBP 12.7m gain in industrial and logistics major developments and losses in residential and I&L strategic land. EPRA NDV is a number shareholders watch closely here, and it went the wrong way. The second story is the pivot. The same announcement confirmed “we're confirming today our intention to exit the residential sector entirely and accelerate the reallocation of capital to higher returning opportunities” — Lynda Shillaw, Chief Executive Officer · 2026-09-09 — a decisive break from the 2024 plan to shrink residential to under 15% by 2029. The third story is the bid: Shillaw opened by noting Harworth "is currently in an offer period", with an unrecommended approach from Peel (U.K.) Limited. A negative return, a strategic reset and a takeover offer, all on one tape.

Leaning into the theme the tape just sold

What makes this genuinely interesting is the timing against the global tape. Harworth's keyword profile flipped almost entirely into the pivot vocabulary this quarter — data center is now its single strongest term, followed by powered land, land bank, land sale and investment portfolio. Meanwhile the market's own data center complex is being de-rated: the AI data centers basket is a 30-day decliner with 52 tickers down against just 7 up, and data center capacity sits among the 90-day laggards. Harworth is walking harder into a trade the tape is currently fleeing. The reconciliation is position in the value chain. Harworth never builds or operates compute; it sells the option on the grid connection. Shillaw's framing is explicit: “Our model remains focused on unlocking and monetizing the land value in a capital-limited way and at an early stage, which is well before power on dates.” — Lynda Shillaw, Chief Executive Officer · 2026-09-09 The AI-data-center decliners in the tape — names like the miners-turned-hosts and neoclouds — are capex spenders exposed to financing conditions and build-out economics. Harworth crystallizes value at the powered-land stage, before power-on, which is precisely the part of the chain with the least capital intensity and the most scarcity value.

Powered land is the company-unique wedge

Harworth's platform totals 34.8m sq ft with accepted power offers of 0.8GW and an identified path to 1.9GW. It says that is one of the largest secured powered-land pipelines held by a UK listed real estate platform. The Microsoft sale at Skelton Grange is progressing to completion, a second hyperscale site is in exclusivity, and four more candidates are identified. JLL puts the potential future profits from the powered-land portfolio beyond Skelton Grange at GBP 292m; a separate JLL opinion adds GBP 174m of net realizable value from the substantially construction-ready and medium-term strategic land pipeline — value Shillaw stressed is not in the Red Book valuations and therefore not in EPRA NDV. The industrial side carries the same logic. Harworth says return on capital employed across industrial and logistics land and developments, including data centers, has averaged 24% annually over three years, versus 8.2% unlevered on the GBP 301m investment portfolio. That spread is the entire rationale for shedding residential and running the portfolio to a GBP 500–600m stabilized value rather than a target size. On the operating line, the largest ever substantially construction-ready land bank of 3.8m sq ft is the near-term monetization engine — roughly GBP 600m of GDV over three to five years. Three prelets totalling over 300,000 sq ft are completed or in legals, worth GBP 3.7m of annualized rent at a 17% premium to ERV, and 58% of budgeted full-year sales are already completed, exchanged or in legals. Net debt rose to GBP 190m from GBP 145.9m, taking net loan-to-portfolio-value to 20.3% against a self-imposed 25% ceiling, with GBP 99.5m of liquidity.

If we identify that we hold surplus capital, including following sales of material assets and having assessed future accretive capital deployment opportunities, the board will consider returning some or all of this surplus capital to shareholders, enabling them to benefit directly from the value creation initiatives as they are executed.
Lynda Shillaw, Chief Executive Officer · 2026-09-09

The tension worth watching

Harworth has framed this as an acceleration of key initiatives already agreed in principle by the board, targeting low double-digit total accounting returns over the long term, with a cost-savings statement to follow. That is a credible-sounding plan against a portfolio that is hard to replicate — advanced planning, power and strategic locations, as Shillaw repeatedly argues. The counterweights are real. Data center sentiment has cooled sharply over the last 30 days, and Harworth's value realization depends on hyperscaler demand for regional UK sites persisting through that wobble. Residential exit timing is uncertain in a soft house builder market. And with the company inside an offer period, the market may end up pricing Harworth on Peel's terms rather than on a JLL-derived pipeline valuation that sits outside EPRA NDV. The irony: the market is de-rating the compute build-out just as a landowner argues it can harvest the scarcity premium without paying for the compute.