Joburg Power Deals: Why Digital Businesses Should Care

Johannesburg’s electricity bill tells a story most online businesses can’t ignore: City Power paid about R13.97-billion for 7,376.1 GWh from Eskom, while paying about R1.27-billion for 789.7 GWh from Kelvin Power Station. On unit price, that’s R1.89/kWh vs R1.61/kWh — a 28c/kWh gap. In a city where downtime can kill conversions, that price difference isn’t an accounting detail; it’s a digital-economy constraint.

Now City Power wants to buy back Kelvin Power Station (a privately owned coal plant supplying roughly 10–14% of Johannesburg’s demand) and acquire Egoli Gas (a 132km reticulated gas network serving 8,500+ customers). The stated goal is straightforward: reduce dependence on Eskom and stabilise affordability. The context is blunt: City Power is under severe pressure, reporting an overdraft of R20-billion, 42% losses of electricity bought from Eskom, and a R2-billion revenue deficit.

For this series — How AI Is Powering E-commerce and Digital Services in South Africa — the real point isn’t who owns which asset. It’s what happens when a city tries to diversify energy supply while digital demand keeps rising. If Johannesburg changes its energy mix, AI becomes the practical tool that helps businesses and utilities keep the lights on, keep costs predictable, and keep digital services performing.

City Power’s buyback push is really about unit cost and control

City Power’s interest in Kelvin and Egoli Gas is primarily a cost-and-control strategy. The numbers in the council report make that clear: Kelvin’s average unit cost is lower than Eskom’s, even though Kelvin supplies a smaller volume.

Kelvin Power Station: cheaper units, older plant, big implications

Kelvin was commissioned in 1956, sold by the city in 2001, and is now one of the few privately owned coal-fired plants in South Africa. It’s also meaningful at city scale: 10–14% of Joburg’s power needs is not trivial.

If the feasibility study supports acquisition, City Power potentially gains:

• More predictable procurement for a portion of demand
• A hedge against tariff increases from the primary supplier
• Operational influence over maintenance cycles and dispatch priorities

But there’s a catch: an older coal asset comes with constraints — reliability risks, emissions considerations, and refurbishment costs. The smart move isn’t pretending those don’t exist. The smart move is treating Kelvin as a dispatchable asset that must be managed like a performance system, measured in availability, forced outage rate, ramp time, and cost per delivered kWh.

Egoli Gas: diversification that changes how businesses consume energy

Egoli Gas is the exclusive licensed reticulator of natural gas in the Johannesburg underground network. City Power’s report frames it as a platform to diversify energy supply and reduce electricity reliance — especially for heating, cooking, and industrial processes.

For digital businesses, that matters indirectly but powerfully:

• If some commercial/industrial loads shift to gas, electricity capacity frees up for critical loads (data rooms, fulfilment operations, payment systems).
• Gas can support hybrid energy strategies (e.g., gas for thermal demand, electricity for high-value digital operations).

Put differently: the more flexible the city’s energy mix is, the easier it is for the local digital economy to scale without betting everything on a single upstream supplier.

Energy diversification only works if it’s intelligently operated

Buying assets doesn’t automatically reduce outages or costs. Operations does. This is where AI stops being “nice to have” and becomes infrastructure-grade.

AI for grid loss reduction: 42% losses is a crisis, not a rounding error

A reported 42% loss of electricity purchased from Eskom points to a combination of technical losses, theft, metering gaps, and billing problems. Reducing that number is one of the fastest ways to improve finances and reliability.

AI can help in practical, non-magical ways:

• Non-technical loss detection: anomaly models flag feeder-level discrepancies between expected and measured consumption.
• Theft pattern clustering: identifying repeat geographies, time windows, and meter tamper signatures.
• Predictive maintenance: forecasting failure likelihood of transformers/cables using historical faults, load profiles, and environmental signals.

If you’re running an e-commerce operation, this matters because less loss typically means fewer forced interruptions, fewer emergency load shedding interventions, and better service stability.

AI dispatch and forecasting: making Kelvin and gas actually useful

If City Power acquires Kelvin and Egoli Gas, it inherits operational complexity. The value is realised when dispatch and demand are managed with precision.

AI-driven forecasting can improve decisions like:

• When to run Kelvin harder (and when not to)
• How to plan maintenance without triggering local instability
• Which customer segments should be encouraged to shift thermal loads to gas

A concrete example: if a model predicts a high-demand week (think end-of-year retail peaks, salary cycles, or heat-driven cooling demand), the city can plan generation and network operations earlier, reducing the need for last-minute, blunt load curtailment.

Dynamic pricing and demand response: the underrated lever

Most companies fixate on generation supply. In cities, demand shaping is often quicker and cheaper.

With smarter tariffs and opt-in demand response programmes, large flexible loads can be nudged to shift usage away from peak times. AI helps by:

• Estimating elasticity (who responds to price signals, and how much)
• Automating customer targeting (the right message to the right segment)
• Verifying response (did load actually shift?)

This is where digital businesses can participate rather than just suffer. Warehousing refrigeration cycles, EV fleet charging, or non-urgent compute jobs can move to cheaper periods if the incentives are real.

What this means for e-commerce and digital services in Johannesburg

Stable power isn’t just “good for business.” In e-commerce and digital services, it’s a direct input to conversion rate, customer trust, and unit economics.

Reliability shows up in customer experience, not just ops dashboards

When power is inconsistent:

• Call centres drop calls, extending resolution times
• Payments fail mid-checkout, hurting conversion
• Stock systems lag, causing overselling
• Last-mile routing and scanning fail, creating delivery disputes

I’ve found that companies often treat resilience as an IT problem (buy a UPS, add a generator) when it’s actually a systems problem: energy, network, operations, and customer comms all tied together.

If the city’s diversification plan works, the benefit isn’t just fewer outages. It’s fewer “brown moments” — those short, chaotic disruptions that don’t make headlines but quietly wreck service quality.

Cost predictability is the hidden win

Even when power is available, unpredictable tariffs and emergency operating costs hit margins. If City Power can access lower-cost supply (as Kelvin’s unit cost suggests) and reduce losses, it can create room for more predictable municipal pricing.

For online retailers, predictability supports:

• More accurate fulfilment cost models
• Better staffing plans for peak periods
• More stable pricing (you don’t have to pass every shock to the customer)

AI inside your business: treat energy as a forecastable input

While municipalities work through feasibility studies and acquisitions, businesses can act now.

Practical AI applications that pay off in South Africa’s power environment:

1. Outage-aware demand forecasting: incorporate outage schedules, historical downtime, and store/warehouse availability into sales forecasts.
2. Inventory risk scoring: flag SKUs likely to be impacted by cold chain disruptions or fulfilment delays.
3. Smart routing for fulfilment: reroute orders to facilities with higher power resilience during unstable periods.
4. Energy optimisation for operations: predict peak usage windows and automatically schedule energy-intensive tasks (printing, packing lines, batch compute) for cheaper or more stable times.

If you’re already using AI for marketing automation or customer engagement — common in South African e-commerce — extending it to energy and operations is the next logical step.

Public-private partnerships: the model South Africa keeps returning to

Kelvin is owned by an infrastructure investment structure (through Anergi and related stakeholders), while Egoli Gas sits under a black-owned investment group. City Power’s approach signals a broader pattern in South Africa: public services increasingly intersect with private capital and specialised operators.

That matters for AI adoption because partnerships often bring:

• Better operational data discipline (a prerequisite for useful AI)
• Performance-based accountability (KPIs tied to uptime and loss reduction)
• Faster procurement of modern systems compared to legacy municipal IT cycles

But partnerships also create a risk: fragmentation. If generation, gas distribution, metering, billing, and network operations run on disconnected systems, you don’t get intelligence — you get dashboards that argue with each other.

A strong feasibility roadmap should therefore ask a very specific question: What data architecture and operational command layer will run this diversified energy portfolio day to day? If the answer is vague, the benefits will be too.

What you should watch next (and what you can do now)

The council has approved detailed feasibility studies — financial, technical, legal, and operational — plus an implementation roadmap and risk mitigation plan. For digital businesses, there are a few signals worth tracking because they’ll show whether this is real progress or just a headline.

Signals that the plan is becoming operational

Look for evidence of:

• Loss reduction programmes with measurable targets (e.g., feeder-by-feeder improvements)
• Metering modernisation and data integrity initiatives
• Maintenance and availability reporting for Kelvin (not just ownership announcements)
• Customer load-shifting programmes tied to gas adoption and tariff incentives

Quick wins for e-commerce and digital service leaders

You don’t need to wait for the city. Do these in Q1 planning:

• Map your top 10 power-dependent failure points (payments, WMS, call centre, refrigeration, dispatch scanning).
• Build a simple energy resilience score per facility (backup runtime, switch-over time, outage frequency).
• Use AI (or even basic forecasting) to schedule energy-heavy tasks and reduce peak exposure.
• Add customer comms playbooks for service disruption windows (order updates, alternative pickup options, proactive refunds where needed).

A stable digital economy isn’t built on apps alone. It’s built on electricity that’s affordable, predictable, and well-managed.

This post sits in a bigger theme: AI is powering e-commerce and digital services in South Africa — but it can’t do that reliably if the underlying energy system is brittle. Johannesburg’s move to consider Kelvin Power Station and Egoli Gas is a reminder that infrastructure and software are now part of the same business story.

If Johannesburg ends up owning more of its supply mix, the next competitive edge won’t come from ownership alone. It’ll come from operating that mix intelligently — and for businesses, from treating energy as a data problem you can forecast, optimise, and plan around.

What would change in your customer experience if you could confidently plan for the next 90 days of power availability and cost instead of reacting week by week?