‘We are only just seeing the start of data centre growth in Africa’: Knight Frank’s Oscar Matthews

With just 350–375MW of live capacity across the continent compared with over 1.25GW in London alone, Africa’s data centre sector remains under developed despite surging digital demand. Oscar Matthews, partner, EMEA data centre investment and development, at Knight Frank, speaks with Real Asset Media’s Jason Mitchell about where investment is emerging and how infrastructure, energy and regulation will shape growth.

Can you begin by telling us your role at Knight Frank and what you do, and then we can move on to discuss data centres and digital infrastructure in Africa?

How would you characterise the current state of African data centre investment? Is it an emerging opportunity or an active institutional market? So where are we at at the moment in terms of data centre investment in Africa?

If you look at the actual statistics behind the live capacity across the African continent, today there is about 350 to 375 megawatts of live IT capacity across the entire continent.

Could you put that into context for people? So what could we compare that with? What might you have in Europe say?

If you take the London market itself, that’s around 1.25 gigawatts of live IT capacity, not including hyper-scale self-builds. So, there is a vast shortage comparatively.

Data traffic in Africa is almost doubling every year and a half, and with so many young people coming online, is much of this data still being processed outside the continent?

Absolutely, I think digital adoption in Africa is the fastest in any other region of the world. And we’re seeing little pockets of capacity being built in major markets across the continent and when certain data sovereignty laws will come into play it will demand that the majority of data processing and storage will happen within each.

Could you outline where the main data centre clusters are in Africa — which cities or countries already have them, and where you expect new clusters to emerge over the next five years?

Two-thirds of all the capacity in Africa is in South Africa. The majority of that is in Johannesburg where you’ve got cloud availability zones for the majority of hyperscalers and you’ve got a very large enterprise market from the local CBD. Outside of that, you look at the strategic locations of East, North and West, so Kenya is probably the third largest market in Africa, with Nigeria taking the second spot.

Right, so there are data centres around Lagos, there are data centres around Nairobi but probably not particularly big, no?

No, I think the largest one in development in Kenya is only about 24.5 megawatts.

How about if we look at Northern Africa, is there much happening in Morocco or Egypt?

Northern Africa is a very interesting one because a lot of the European players are looking at alternative ways of servicing AI deployments for Europe. If you look at Northern Africa, especially Morocco and Egypt, there’s plenty of renewable energy, there are huge solar projects going on across both countries and from a latency distance perspective to Southern Europe you’re very close. You can probably look at servicing Southern Europe from Northern Africa if you can get the legislation, when data sovereignty bills come in, and whether or not you can process data in one location and then the end user use it in another.

Presumably one of the main reasons there is that the cost of energy is lower in Northern Africa?

Cost of energy, cost of land, cost of labour and that all factors into the bill costs of data centres which are rocketing at the moment.

Looking ahead over the next few years, do you expect most new data centre development to remain concentrated in South Africa, or could we see stronger growth in markets like Kenya or Nigeria? And what about emerging locations such as Ethiopia or Rwanda?

I think South Africa will predominantly take most of the growth just because you’ve got the availability zones and the cloud regions of all three hyperscalers. There have been big announcements recently — Oracle have just announced a partnership with iXAfrica in Nairobi which is great news. That’s a new cloud region.

How big would that investment be? Do we know?

I believe it’s somewhere around 5-10 megawatts, that sort of size. In terms of build costs, we’re at about $12 to 15 million per megawatt.

Do you think this will effectively drive data centres to be built within these countries, given the demand is already there from the digitalisation of economies, government systems and records going online? And particularly as AI becomes embedded within enterprises, rather than just consumer use, could that further accelerate the need for local infrastructure?

Exactly.

Are we starting to see that in Africa now? Are we seeing big banks and others doing that?

One of the traditional themes of data centres was that the major banks in all markets globally would have their own data centres. It would start off as a server room on the floor in an office and that would just grow into its own facility, but now banks can move a lot of their applications to the cloud, so they’re looking to have third-party operators run their facilities but have what we call on-prem so they have their own technicians control them, the data centre servers from a regulation point of view. But we’re also seeing that one of the biggest data centre businesses in Africa, Teraco, they’re developing huge amounts of capacity but that’s not only for enterprise demand, that’s also for cloud demand because building in African countries isn’t as simple as some of the other established markets.

Let’s turn to energy because obviously that can be a challenge. Data centres are very power hungry but electricity’s not always reliable in Africa. How are the data centre developers squaring that circle in Africa?

I think grid supply is struggling across the globe, but in a traditional data centre design there is dual-fed grid power, so if one feed goes down the other can take the load, and we call it the five nines — 99.99% uptime. That is what data centre operators promise their customers, so if there’s any risk of that going down, that’s why there’s so much redundancy in the power support for data centres. Whereas if you look at certain markets like Nigeria, for instance, the grid is actually used as backup, but then you look at the Kenyan grid and 85% of that is geothermal renewable power, which you would think is a really good box to tick for major data centre operators. However, you’ve got to have the customer demands to justify the investment into building these things.