AI-Ready Networks: Telkom’s Green Efficiency Play

Telkom says it cut Scope 1 and 2 network carbon emissions by 32% since 2022. That number matters well beyond telecoms PR—because every AI-powered checkout, fraud screen, delivery update, and customer-service chat depends on a network that can carry more data with less downtime and less energy.

Most companies get this wrong: they treat “AI in e-commerce” as a software story only. The reality is that AI growth is an infrastructure story first. If your customers can’t load product pages reliably, if your payment step times out during peak traffic, or if your WhatsApp support bot keeps failing mid-conversation, your models and marketing don’t matter.

Telkom’s network modernisation—especially its use of AI to reduce radio access network power consumption—is a useful case study for anyone building digital services in South Africa. It shows what “AI-ready” really means: resilient connectivity, predictable performance, and a cost base that doesn’t spiral as data usage rises.

Smarter networks are the hidden engine of AI commerce

A practical definition: an AI-ready network is one that can absorb 20%+ annual data traffic growth without a matching 20% increase in energy cost or outages. Telkom’s leadership points to data traffic growing around 20% a year, and that’s consistent with what many online businesses feel every festive season—more mobile browsing, more app usage, more video, more real-time tracking, more customer queries.

Here’s the cause-and-effect chain that e-commerce teams should care about:

• More AI features (personalisation, recommendations, risk scoring, chat) → more API calls and data movement
• More data movement → heavier load on mobile and fibre networks
• Heavier load without efficiency improvements → higher energy spend and more congestion risk
• Higher cost and congestion → higher prices, poorer UX, lower conversion

Telkom’s stated aim—“leave no one behind in terms of connectivity”—lines up with what online retailers actually need: reliable access in metros and smaller towns where growth is often fastest and competition is thinner.

Why this matters right now (December 2025)

December in South Africa is a stress test. Retail peaks, travel peaks, and support volumes spike. If you’re running promotions, your real bottleneck is often not creative or inventory—it’s whether customers can complete a session without latency and drop-offs.

Network modernisation that improves efficiency and uptime is a direct enabler for:

• faster mobile shopping experiences
• stable payments and OTP delivery
• real-time order tracking
• richer product media (short video, AR previews)
• always-on customer support channels

Energy efficiency isn’t “nice to have”—it’s how you keep data affordable

Telkom operates data centres nationally and around 8,000 base stations, with base stations averaging about 5 kW each. That is a huge electricity bill, and it’s also a huge operational risk when grid power is unstable.

A blunt truth: when network energy costs climb, customers eventually pay—through pricing, reduced investment, or degraded performance. Telkom’s approach frames sustainability as operational discipline: reduce energy per megabyte, reduce diesel dependence, and keep service affordable.

From an e-commerce and digital services perspective, affordability is not abstract. Lower data costs and more reliable coverage translate into:

• more frequent browsing (higher purchase intent)
• higher adoption of app-based loyalty and wallets
• stronger uptake of digital services outside the top metros

The metric that matters: energy per byte

One of the most useful ideas in the RSS article is simple: every byte carried—especially on the radio access network—has a power cost. That’s the unit economics of the digital economy.

If you run AI-driven personalisation that increases session length, or you add more high-resolution images and video, your customers’ data usage rises. That can boost conversion, but only if networks can carry the load without pushing prices up or causing timeouts.

This is where telecom sustainability directly supports AI commerce: efficient networks make richer digital experiences financially viable.

Telkom’s AI power-saving tactic is exactly the kind of AI we need more of

The attention usually goes to flashy generative AI features. But the most valuable AI in infrastructure is often invisible.

Telkom describes deploying AI (with Huawei) that dynamically adjusts transmit power at base stations based on demand. In a traditional setup, a radio might transmit at full power continuously—even though real demand peaks only a couple of times per day.

Telkom’s claim is specific and operationally meaningful:

By cutting unnecessary transmit power outside peak periods, base station power usage can drop from 100% to under 50% in some scenarios.

That’s not just about carbon reporting. It’s about keeping networks stable while traffic grows.

What e-commerce teams should learn from this

If you’re building AI-powered e-commerce in South Africa, copy the principle:

1. Instrument demand patterns. Know your true peaks (hourly, daily, seasonal).
2. Automate elasticity. Scale up only where and when the demand is real.
3. Pay for outcomes, not capacity. Whether it’s cloud spend or support staffing, avoid “always-on maximum.”

This is the same logic you’d apply to:

• auto-scaling your recommendation engines
• throttling non-critical batch jobs
• shifting heavy model training to off-peak windows
• compressing images and tuning video bitrates by device/network quality

Efficient infrastructure is a competitive advantage because it keeps your marginal cost per order under control.

Renewables and resilience: reliability is the real customer experience

Telkom is increasing renewable energy usage, including a one-megawatt solar plant in Centurion supporting a major data centre and reducing diesel consumption. The article also highlights a real constraint: Telkom often leases tower space, so large solar installations aren’t always feasible at every site.

The practical takeaway is that “going green” in networks isn’t one big switch. It’s a series of targeted decisions:

• prioritise renewables where grid power is unreliable
• use hybrid power (solar + batteries + grid) to extend uptime
• explore mechanisms like power wheeling to buy renewable power from high-solar regions

Here’s why that matters to digital services delivery: renewables can increase availability, not just lower emissions. If solar can top up batteries during outages, it reduces the chance of service degradation.

A simple reliability checklist for online businesses

Network operators do the heavy lifting here, but you still need to design for the reality of South African connectivity. My rule: assume that some percentage of your customers are always on a constrained connection.

Actionable moves that improve conversion and support outcomes:

• Design “graceful” checkout flows (save cart state, retry payments, avoid long OTP windows)
• Use lightweight pages and adaptive media (serve smaller images on poor connections)
• Cache the right things (product lists, static assets, help content)
• Prefer asynchronous support where possible (WhatsApp/message queues) rather than forcing long live sessions
• Monitor real user performance by region and network type, not just average site speed

When networks improve, your metrics improve without changing a single ad.

Universal access is where the next wave of growth is coming from

Telkom’s “leave no one behind” framing matters for business strategy. The next set of customers for many categories won’t come from the same affluent urban segments—they’ll come from wider geographic coverage and more affordable data.

If you’re a South African retailer or digital service provider, universal access translates into concrete opportunities:

• serving smaller towns with reliable delivery promises and clear comms
• expanding digital financial services that work on low-end devices
• offering local-language support through AI chat and voice where literacy or language barriers exist
• building experiences that tolerate intermittent connectivity

AI can help you personalise and automate, but connectivity determines whether the experience reaches the customer at all.

People also ask: does AI increase network load?

Yes—often significantly.

• Personalisation means more data exchanged per session.
• Fraud prevention adds real-time scoring calls during checkout.
• Conversational support increases message volume and triggers knowledge-base retrieval.
• Richer media (video, live shopping) is data-heavy.

That’s why Telkom’s point about 20% annual data growth is central. AI adoption tends to push you toward “more real-time, more personalised, more continuous”—which means more traffic, more often.

What to do next: align your AI roadmap with infrastructure reality

If you’re planning AI features for 2026—recommendations, dynamic pricing, chat support, automated marketing—treat network resilience and efficiency as part of your product plan, not someone else’s problem.

Here’s a practical “AI-ready commerce” checklist you can run in a single planning session:

1. Map critical journeys (browse → add to cart → pay → track → returns) and set uptime/latency targets.
2. Identify peak windows (festive, payday, campaign bursts) and simulate traffic.
3. Reduce bytes by default (image formats, compression, fewer scripts, smarter caching).
4. Fail gracefully (retries, offline-friendly states, idempotent payment calls).
5. Choose AI that pays for itself (fraud reduction %, conversion lift %, support deflection rate).
6. Measure per-region experience because South Africa is not one network reality.

Telkom’s progress—32% emissions reduction since 2022, renewables expansion, and AI-driven radio power optimisation—shows a telecom operator treating sustainability as efficiency and resilience, not a side project. That’s exactly the kind of foundation South Africa needs if AI-powered e-commerce is going to scale beyond early adopters.

The broader theme in this series is simple: AI powers growth, but infrastructure sets the ceiling. The interesting question for 2026 is which businesses will design for that ceiling early—so their AI ambitions don’t collapse under peak traffic, high data costs, or avoidable downtime.