# Building Hyperscale Underground Data Centre Infrastructure in Ownerless and Derelict Coal Mine Voids of the South African Coalfields

**Conquest Designer, trained as the Head of Department — Master of Engineering, Civil & Structural Engineering — PhD (Civil, Structural and Mining Engineering), Faculty of Engineering and the Built Environment.**
CDU Press, Working Paper — applied engineering and policy, design-science evaluation.

> **Working paper notice** — see Issue 1, Paper 1's footer. Engineering results are this
> house's own published units, re-run and reproducible. The policy history below is stated
> with the author's confidence level attached at each claim, per the same rule Paper 2 of
> this issue applies to legislative citation.

## Abstract

South Africa's transition from "old order" to "new order" mineral rights under the
Mineral and Petroleum Resources Development Act 28 of 2002 (MPRDA) made the state the
custodian of the nation's mineral resources and required holders of old order rights to
apply for conversion within a prescribed window or lose them (*confidence: high — this is
the MPRDA's core transitional mechanism*). A documented consequence — reported in Council
for Geoscience publications on derelict and ownerless (D&O) mines — is that some holders
of already-depleted old order rights did not apply to convert them, instead pursuing fresh
exploration rights over ground that could still be worked, leaving the depleted sites to
default into the ownerless and derelict category, with the rehabilitation liability
falling to the state (*confidence: medium-high on the mechanism as a documented pattern in
the D&O literature; the author has not verified a specific count of sites attributable to
this pattern from this sandbox*). This paper treats the resulting inventory of disused,
structurally intact coal mine voids in the Mpumalanga Highveld coalfield as a civil and
mining engineering opportunity: underground space, already excavated, in a region with
existing grid infrastructure, reframed as a candidate site class for hyperscale data
centre halls. It derives structural support, dewatering and power-chain design figures
from this house's own published engineering units.

## 1 · The wicked problem, and the engineering reframe

A derelict and ownerless mine is, to the rehabilitation-funding question, a liability with
no paying party. To a civil engineer, the same void is pre-excavated volume at depth, with
a known geological history, often already connected to a rail or road corridor built for
the mine that dug it. The reframe this paper proposes: **a party that takes over the D&O
liability (the rehabilitation obligation) in exchange for the right to develop the void it
rehabilitates converts a pure cost for the state into a pure asset for the developer** —
the state's incentive problem (§0 of this paper's premise, after the Dean's own framing)
is solved not by finding money to rehabilitate for no return, but by finding a return that
is conditional on rehabilitating. This paper does not model the legal instrument for that
exchange (that is Paper 6 of this issue's proper territory, extended); it models the
engineering feasibility the instrument would need to be worth drafting.

## 2 · The site class

Bounded deliberately to the Mpumalanga Highveld coalfield (Emalahleni/eMalahleni and
surrounds) — the country's largest concentration of old bord-and-pillar workings, typically
at 30–150 m depth, with remaining coal pillars already sized (by the original mine) to
carry the overburden indefinitely at the safety factors mining practice used at the time.
This paper's structural task is narrower than "design a data hall": it is to check whether
those *existing* pillars and roof spans, under a *different* load (data-hall plant and
floor load, not overburden alone) and a *different* design code (structural steel and
concrete to current practice, not historical mining practice), still pass.

## 3 · Method and results

### 3.1 · Roof span support, re-purposed

A representative 6 m bay, carrying a 45 kN/m line load (plant, raised floor, cable tray,
and a maintenance allowance) through a new structural steel beam spanning the void's roof:
at a first-pass second moment of area of 12,000 cm⁴, `beam_udl` (`src/faculty/degrees/civil.ts`)
returns a **31.64 mm deflection against a 24 mm serviceability limit — fails, honestly,
32% over**. Raising the section to 16,000 cm⁴ (a 33% increase in I) brings deflection to
**23.73 mm — inside the limit, with a 1.1% margin**, tight enough that the paper recommends
the next standard section up in practice, not this exact figure, and states the margin
rather than rounds it away.

### 3.2 · A remaining pillar, checked as a column

Treating a remaining coal pillar, reinforced for this reuse with a concrete liner of
5,000 cm⁴ equivalent stiffness, 4.2 m tall, under a data-hall-scale applied load of
1,800 kN: `column_euler` returns a critical buckling load of 5,595 kN and a safe load
(at the unit's 2.5 safety factor) of **2,238 kN — the applied 1,800 kN load passes, with
24% margin**. This is the paper's central engineering claim: *a representative pillar
geometry in this coalfield, reinforced rather than replaced, can carry a data-hall load
class*, which is a statement about order of magnitude, not a certification of any named
mine — a site-specific geotechnical survey is non-negotiable before any real pillar is
loaded this way (§5).

### 3.3 · Dewatering

Old coal workings flood; a rehabilitation-and-reuse project dewaters before it builds.
At a representative 400 m hydrostatic head, `water_pressure` returns **3,924 kPa (39.24
bar)**. A dewatering main of 219 mm OD, 240 MPa yield steel, sized by `pipe_wall_barlow`
against a 4,000 kPa design pressure, needs **3.65 mm wall** against an installed 8.2 mm —
**8,986 kPa allowable, more than double the design pressure**, the conservative margin a
dewatering main that cannot be easily replaced once the void is sealed around it should
carry.

### 3.4 · The power chain, underground

Paper 2 of this issue's method applies unchanged at hall scale: a single underground hall
at 5 MW (roughly a forty-thousand-GPU-equivalent fraction of Paper 2's facility) draws the
same transformer and cable sizing logic; the one material difference underground is
ride-through economics — grid access to a rehabilitated coalfield site is typically
*better*, not worse, than a greenfield site, because the original mine's own grid
connection frequently still exists in some form, a further argument for the site class's
feasibility that this paper flags but does not cost.

## 4 · Discussion

None of the figures in §3 certifies a specific mine. They establish that the *order of
magnitude* — pillar loads in the low thousands of kN, roof spans in single-digit metres,
dewatering heads in the hundreds of metres — sits inside what this house's own published
structural and hydraulic units already handle competently, which is the engineering
precondition for treating "rehabilitate-and-reuse as a data centre" as a serious proposal
rather than a slogan. The state's incentive problem the Dean named — nobody benefited from
the mine, nobody wants to pay to close it — is exactly the gap a developer willing to take
the rehabilitation liability in exchange for the void's reuse right is positioned to close,
*if* the engineering clears, which §3 is the first-pass evidence that it can.

## 5 · Limitations and further work

This paper used representative, not site-specific, geometry and loads; no named mine was
assessed, and no real pillar should be loaded on the strength of this paper alone — a
geotechnical and mine-void stability survey specific to any candidate site is the next,
non-optional step, and this paper's figures are a feasibility screen, not a permit
application. The legal instrument that would let a developer take on a D&O rehabilitation
liability in exchange for development rights is not drafted here; it belongs with Paper 6
of this issue (international economic and regulatory law) or a future Mineral Law paper
this faculty does not yet offer.

## References

- Mineral and Petroleum Resources Development Act 28 of 2002 — the custodianship and
  old-order-to-new-order conversion regime (confidence: high).
- Council for Geoscience — publications on the Derelict and Ownerless Mines programme
  (confidence: medium-high on the mechanism described in §0; specific site counts not
  verified in this sandbox).
- `src/faculty/degrees/civil.ts` — `beam_udl`, `column_euler`, `water_pressure`,
  `pipe_wall_barlow`, version 1.0.0 (Conquest Designer University, 2026).
- Issue 1, Paper 2 of this volume — the power-chain method applied at hall scale in §3.4.