AI E-commerce Needs Greener, Smarter Networks in SA

Data traffic on South African mobile networks is growing at roughly 20% a year. That’s not a “telecom” stat—it’s an e-commerce stat. Every product search, checkout, fraud check, chatbot reply, delivery update, and recommendation feed rides on the same underlying connectivity.

Here’s the thing most online retailers and digital service teams miss: AI doesn’t fail first in the model. It fails in the network. When connectivity is unstable, latency spikes, or power issues knock sites offline, the fanciest personalisation engine in the world can’t save your conversion rate.

Telkom’s recent network modernisation work—especially its push for smarter, more energy-efficient networks—is a practical example of the infrastructure layer South Africa needs to make AI-powered e-commerce and digital services reliable, affordable, and scalable. And it’s happening with measurable progress: Telkom reports a 32% reduction in combined Scope 1 and 2 carbon emissions since 2022 across its network operations.

Smarter networks are the hidden layer behind AI-driven e-commerce

Answer first: AI-powered e-commerce depends on stable, low-latency connectivity and resilient power; smarter networks make that possible while keeping data costs in check.

E-commerce AI features are often “always-on” workloads:

• Real-time fraud scoring at checkout
• Personalised recommendations that update per session
• Customer service chatbots and agent-assist tools
• Dynamic pricing and stock availability signals
• Marketing automation triggered by behaviour

All of those need consistent connectivity between users, apps, payment rails, and cloud services. In practice, the weak points are usually mundane:

Latency and reliability decide whether AI helps or annoys

A recommendation carousel that loads slowly doesn’t feel “smart”—it feels broken. A fraud check that times out doesn’t just lose a sale; it can create a manual-review backlog and higher chargeback exposure.

That’s why network reliability is a growth lever. When the network is resilient, AI can operate in real time. When it isn’t, teams quietly disable features (“We’ll turn off real-time risk scoring on mobile during peak hours”), and the promised ROI never lands.

Energy efficiency becomes a customer experience issue

Telkom operates around 8,000 base stations, each drawing roughly five kilowatts on average. Power is not an abstract cost. It affects:

• Network operating costs (which ultimately show up in data pricing)
• Uptime during power instability
• Service consistency during peak periods (when customers are buying)

If you’re selling online in South Africa, you’re not just competing on product and price—you’re competing on whether your experience works at the exact moment the customer is ready to pay.

Why greener telecom infrastructure matters to digital services

Answer first: A lower-carbon, renewable-boosted network isn’t only “good PR”—it directly improves resilience and reduces the cost pressure that makes digital services expensive to run.

Telkom’s strategy, led on the mobile network side by Managing Executive Lebo Masalesa, frames sustainability as a data-led operational discipline. That’s the right framing. Sustainability work that isn’t tied to operational reality becomes a slide deck. Sustainability tied to network performance becomes business value.

Renewable energy is also a reliability strategy

Telkom has increased the share of renewables in its mix, including a one-megawatt solar plant in Centurion supporting a major data centre and reducing diesel usage at that site.

For e-commerce and digital services, this matters in a very direct way:

• Fewer diesel-dependent failovers reduces the risk of degraded performance
• More predictable energy supply supports stable service levels
• Better uptime reduces abandoned carts and support tickets

Masalesa makes a practical point: telecoms often lease tower space and mast infrastructure, which limits the footprint available for solar installations. So renewables rollouts aren’t simply “install solar everywhere.” They’re targeted, prioritised, and combined with other approaches—like power wheeling (buying renewable power from regions with strong solar resources).

That constraint is useful for business leaders to understand: infrastructure upgrades happen in stages, and your digital strategy should plan for that reality.

Net-zero targets influence procurement and partnerships

Telkom’s stated goals—carbon neutrality by 2035 and net-zero emissions by 2040—create downstream effects:

• More pressure to select energy-efficient network equipment
• More investment in monitoring and optimisation systems
• Stronger business cases for automation (including AI-driven controls)

If you’re building an AI-powered digital service, these shifts matter because they shape future network capabilities and cost structures.

AI is already saving energy on the network—here’s why that matters to you

Answer first: AI-driven network controls reduce wasted power, which supports affordability and capacity—two things e-commerce needs to scale.

One of the most concrete examples from Telkom’s programme is AI-based power management at base stations. In a traditional setup, radios may transmit at full power continuously. But real customer demand isn’t flat.

Masalesa describes an approach—deployed with Huawei—where AI determines when high transmit power is actually needed and adjusts dynamically. The result: in some scenarios, moving from 100% power usage to less than 50% outside peak periods.

That’s a big deal for three reasons.

1) Lower operating costs support affordable data

South Africa’s digital economy grows when more people can afford to participate. When networks can carry more traffic with less energy, operators have more room to price competitively.

For e-commerce teams, “affordable data” translates into:

• More sessions from cost-sensitive customers
• Higher engagement with richer content (video, AR try-ons, live chat)
• Better adoption of app-based loyalty and messaging journeys

2) More headroom during peaks

E-commerce demand is spiky. In December (right now), peaks get sharper—payday periods, promotions, last-minute shopping, and travel all stack up.

Energy-efficient networks can allocate resources more intelligently during those spikes. That doesn’t just help streaming services. It helps:

• Checkout stability during promotional bursts
• Faster customer support interactions
• Reliable delivery tracking and notifications

3) Better resilience when the grid is unreliable

Telkom notes that solar at sites can support availability by filling gaps when grid power fails, alongside batteries. For digital services, resilience is less about “perfect uptime” and more about graceful continuity—your app still loads, payments still process, support still responds.

What e-commerce and digital service leaders should do with this insight

Answer first: Treat network resilience and efficiency as part of your AI roadmap, then design experiences that degrade gracefully when connectivity isn’t perfect.

I’ve found that the best teams don’t assume ideal conditions. They plan for variability—especially in mobile-first markets.

Build an “AI feature reliability” checklist

Before rolling out AI-heavy features, pressure-test them against real conditions:

1. What happens if latency doubles? (Does the feature time out? Does it block checkout?)
2. What happens if the user drops to 3G/edge coverage? (Do you fall back to a lighter experience?)
3. What happens if your model endpoint is slow? (Do you cache? Do you provide a default?)
4. Can the experience work with fewer round trips? (Batch calls, prefetch, compress payloads.)

Optimise for mobile networks, not office Wi-Fi

If your AI personalisation relies on frequent API calls, you’ll feel every network wobble. Practical steps that pay off:

• Use edge-friendly architectures where feasible (lighter payloads, fewer dependencies)
• Cache recommendations for short windows (even 5–15 minutes can stabilise UX)
• Ship smaller images and prioritise critical checkout resources
• Instrument “time to interactive” and “checkout completion time” by network type

Treat sustainability as a procurement criterion

If you run a digital service at scale—marketplace, fintech, logistics, or subscription—your cloud and connectivity choices have carbon and cost implications.

Ask vendors simple questions:

• What are your energy-efficiency measures at network and data centre level?
• Do you support renewable energy sourcing or power wheeling?
• How do you monitor and reduce energy per unit of traffic/workload?

Sustainability here isn’t virtue signalling. It’s operational discipline.

The bigger picture: universal access is an AI growth strategy

Answer first: Universal access expands your addressable market, and smarter networks make universal access economically viable.

Telkom frames its ethos as “leave no one behind” connectivity—reliable, reasonably priced data products supported by efficient production of each megabyte. That’s exactly the link between telecom modernisation and AI-powered commerce.

AI features often assume users can always be online and always receive rich media. In South Africa, the winners will be the brands that design for:

• Affordable data realities
• Mixed coverage quality
• Power disruptions
• Mobile-first behaviour

When the underlying network becomes more resilient and energy-efficient, it expands what’s possible: richer product discovery, more reliable real-time service, and more inclusive participation from customers outside major metros.

Snippet-worthy truth: If your AI strategy ignores connectivity and energy, it’s not a strategy—it’s a demo.

As this series on How AI is powering e-commerce and digital services in South Africa continues, one theme keeps coming up: AI outcomes depend on the plumbing. Smarter networks are that plumbing.

If you’re planning your 2026 roadmap right now, here’s a strong next step: map your highest-value AI journeys (support, fraud, personalisation, logistics) against network and power risk—then redesign the weak points before peak season exposes them.

What would happen to your conversion rate if your “smart” features had to run reliably on a shaky connection at 7pm on payday?