HARWORTH GROUP PLC (HWG) — Investment Research Note

Executive Summary

Harworth is a UK-focused regeneration and strategic land developer that acquires large, complex (often former industrial) sites across the North of England and Midlands, moves them through planning and remediation, and monetises them either as serviced land parcels or as directly-developed industrial & logistics ("I&L") assets. Over the 2021–2025 strategic period, the group has pivoted the portfolio toward I&L (now 70% vs. a 2029 target of 85%), grown EPRA NDV from £515.9m to £727.3m and — critically — surfaced a highly valuable secondary use for portions of its land bank: hyperscale data centres, evidenced by the £106.6m Microsoft sale at Skelton Grange and a second data-centre site now in advanced negotiation 2026-08 half-year trading update. The single most important valuation point today is that 0.8GW of power connections (conditionally secured or in the pipeline) creates optionality that is only partially reflected in reported NDV.

Fair value estimate

• Methodology: NAV/EPRA NDV cross-checked against implied per-share value of the identified data-centre optionality. EPRA NDV of 224.4p at 31 Dec 2025 2026-03 FY results; H1 2026 expected "modestly below" that level due to residential softness 2026-08 half-year trading update; 0.8GW power book uncapitalised.
• Fair value range: 200p – 245p per share, midpoint ~222p.
  • Low end (200p): applies a ~10% discount to Dec-2025 EPRA NDV to reflect residential-market weakness, extension of the £1bn NDV timeline to end-2028/end-2029, and cyclicality.
  • High end (245p): EPRA NDV + partial credit for a second Microsoft-scale data-centre transaction (management guides "ahead of" first deal), consistent with the site potentially generating >£100m of value gains not currently in the book.
• Implied market-cap range: £650m – £797m (mid £722m) vs. current £584.1m.
• Upside to mid: ~+22% (current 182.33p vs 222p mid).
• The shares already trade at a ~19% discount to NDV (a smaller-than-historical discount given the +41% one-month rally), so a large re-rating from here needs the data-centre pipeline to deliver.

Sector context

• Sector: Real Estate. More specifically a UK strategic-land / regeneration developer rather than a pure REIT.
• Quality/growth profile is above sector average on balance-sheet strength (LTV 15.6%, £127m liquidity) and total accounting return (~8.4% p.a. 5-yr average vs MSCI UK 5.1%). Growth is in line with sector given cyclicality; leverage is below sector.
• Closest listed peers: Segro (SGRO) and Tritax Big Box (BBOX) on the I&L side; Urban & Civic (private) and St Modwen (private) on strategic land. Harworth's model sits between these — it develops but also retains a growing income portfolio.

Investment thesis (3 bullets)

1. Genuine, monetisable AI-infrastructure exposure. Harworth has 0.8GW of power connections either conditionally secured or in the network operators' pipeline, and has already proven the model with the £106.6m Microsoft sale at Skelton Grange (IRR >40%). A second data-centre site is in advanced negotiations with "total value gains ahead of the first transaction" 2026-08 half-year trading update. Unlike most real-estate names, this is a direct capture of AI-related capex, not a marketing narrative.
2. Discount to well-supported NAV with quality balance sheet. At 182.33p vs. EPRA NDV of 224.4p, the shares trade at ~19% discount to a NAV that is independently valued by JLL/Savills/BNP Paribas, backed by 15.6% LTV, £127m liquidity, £275m RCF (with £50m accordion), and no material refinancing until 2027–2029 2026-03 FY results. Downside is well-protected.
3. Strategic pivot to 85% I&L increases through-cycle return quality. The strategy of pivoting away from cyclical residential (68% of value losses in 2025) toward income-generating Grade A I&L (Investment Portfolio target £0.9bn by 2029, currently 76% Grade A) should compress the volatility of returns and provide a growing recurring income base to service future development 2026-03 FY results.

Key risks (3 bullets)

1. Residential market weakness is dragging on NAV. Residential land valuations were "below 31 December 2025 levels" at H1 2026 due to softer demand and cost inflation, and Major Development residential recorded –£28.7m value losses in 2025 2026-08 half-year trading update; 2026-03 FY results. Residential remains ~30% of the portfolio and could remain a drag while housebuilders pull back.
2. Data-centre thesis is concentrated and lumpy. One transaction closed, one in negotiation, further "smaller-scale" opportunities identified — but revenue recognition is deal-based. If AI capex cools, planning consents slip, or National Grid connection reforms disappoint, the incremental option value may not crystallise as forecast 2026-03 FY results, Risk 1 Power Infrastructure Capacity.
3. Cost inflation and slow planning system. Management flagged supply-chain cost increases from Middle East fuel prices and continued "protracted transaction timelines"; planning system risk rated "high" residual 2026-05 AGM statement; 2026-03 FY results Risk 2. These drivers extended the £1bn NDV target from end-2027 to end-2028/2029.

Operating leverage

Harworth is not a classic operating-leverage story. The business has a substantial fixed cost base — 118+ employees, £27.4m of central overheads in 2025 (up from £25.7m in 2024) — but it is primarily a land-value business where incremental value is created through management actions (planning, remediation, servicing) that show up as revaluation gains rather than income statement leverage. However, on a per-transaction basis operating leverage is very high in specific circumstances: the Microsoft data-centre deal generated ~£78m of gains on a modest cost base with an IRR >40%. If a second Microsoft-scale deal completes on the terms management describes ("ahead of" the first), that single event could add ~30p to NDV against a static cost base — effectively a call option embedded in the land bank. The observable inflection is the 0.8GW power book: each 100–200MW allocation that converts to a signed data-centre sale has the potential to add meaningful uplift with negligible incremental cost 2026-08 half-year trading update; 2026-03 FY results. The Investment Portfolio provides operational leverage of a more traditional kind — 10.4% like-for-like rental growth in 2025 on a largely fixed operating cost base — but its scale (£305m) means this is not the dominant driver.

Value-trap signals

• Residential market weakness has stretched the £1bn NDV target from 2027 to 2028–2029 (a guidance extension, not a beat) 2026-01 FY trading update.
• Cumulative TAR of 44.5% over 5 years is strong, but 2025 TAR fell to 1.7% (from 9.1%) — momentum decelerating.
• Interest costs rising materially: finance costs £15.2m in 2025 vs £9.9m in 2024, on more drawn debt.
• None of these signal a "structural cheap" — the underlying land bank retains real, independently-valued optionality. Not identified as a value trap.

Earnings vs. expectations

• FY 2024 (announced 18-Mar-2025): TAR 9.1%, EPRA NDV per share 222.3p (+8.4%), well ahead of prior January guidance which had flagged NDV "moderately higher" than H1 2024. Beat.
• H1 2025 (announced 16-Sep-2025): Management pre-guided in July that NDV would be "flat to marginally up" due to residential headwinds — H1 2025 outcome delivered exactly that. Met.
• FY 2025 (announced 17-Mar-2026): January 2026 trading update guided NDV "flat to marginally up vs H1" and extended the £1bn timeline. Actual EPRA NDV 224.4p came in ahead of consensus of 220p, and TAR 1.7%. Small beat vs pre-guided range, but material deceleration vs prior year and a guidance extension.
• H1 2026 (5 Aug 2026 update): Guided NDV "modestly below" 31 Dec 2025 levels — no beat/miss yet since results are 15 Sep 2026.
• Pattern: Management has recently under-promised and (just) delivered, but with softening magnitude and a stretched strategic timeline. Not a "consistent beats" story; more "meets in a tough tape".

Conviction

Conviction: 3 (moderate).

Anchors:

1. Independent valuers (JLL, Savills, BNP Paribas) provide a defensible NAV floor.
2. Balance sheet is genuinely strong (LTV 15.6%, no near-term refinance risk).
3. Microsoft transaction is real and documented — data-centre optionality is not speculative.

Limits:

1. Second data-centre transaction is not yet contracted; sizing the optionality is judgement-heavy.
2. Residential market direction over next 12–18 months is uncertain and directly impacts a third of the portfolio.

Driver scoring rationale

Overall score: 570 — Partial fit. Genuine AI-receiver angle via 0.8GW power-enabled land bank, but scale is modest and lumpy. Fair-to-slightly-cheap valuation (~19% NDV discount), strong balance sheet, but limited operating leverage in the classic sense and material cyclicality. A "right idea, fair price, but not the AI-receiver purest expression" name.