Where is Africa in this picture? It's a fraction of the "rest of world" slice.
A bit of history: Until the late 2000s, data centers tended to be where the telecom cables were i.e. near big population centers with lots of Internet users. The important internet data center locations were NYC, Northern Virginia, SF bay area, Amsterdam, London, Hong Kong, etc. But as we approached the second decade of this century, energy gradually became a bigger cost than bandwidth so they started developing near power sources. Nowadays, Google and Facebook's datacenters are in places like Iowa and Oregon, far from cities and near abundant supplies of energy.
This is an opportunity for Ethiopia. (And it's what my company QRB Labs is focused on).
But two caveats:
a) latency still matters so they can't be too far from the consumer for some applications like communications and live media. So in the short term, for content hosting, the focus will likely be regional rather than international.
b) for some applications, privacy, data security and intellectual property considerations are complicated, so the data centers need predictable and favorable legal environments. So Ethiopia has some catching up to do before it can really compete in hosting AI and other cloud computing services. But it remains an interesting potential area, especially for the model training portion of AI which is not latency sensitive.
There is one immediate application which is energy intensive, not very latency sensitive and completely location agnostic: Bitcoin mining. I've written about it at length here, but let's put it in the context of data centers in general. The legal requirements are pretty simple as all the data involved is public and the algorithms are open source. And contrary to common misconceptions, mining is purely infrastructure serving global users, it doesn't depend on local adoption of Bitcoin or the local financial regulations. So it can be anywhere in the world. The catch is, because it is location agnostic, miners are constantly seeking the cheapest power worldwide. Mining is globally hyper-competitive any miner that pays more than the other miners on average will immediately go out of business. So to attract miners power prices have to be very low. In the US, while AI data centers can pay up to $0.10/kWh, and households average around $0.15/kWh, and other industries such as manufacturing tend to be somewhere in between those two. But big Bitcoin miners pay $0.02/kWh or less (total cost around $0.04 with half being energy). Of course, power providers will prioritize the higher price consumers. Thus, the natural equilibrium is that Bitcoin miners tend to buy power that no one else can use, also known as stranded power, at a lower price than any other users.
Thus with the right strategy, Ethiopia has a chance to benefit from the data center boom, selling stranded power to Bitcoin miners while working to develop the higher paying demand. So the highest priority should be for the power company to make electricity pricing be supply and demand-based, with aggressive location-based differentiation.
The second priority should be for the government to radically reform the equipment import process. Currently this is a huge handicap. Importing data center equipment is a 9 step bureaucratic process. It takes on average more than 6 weeks to process each shipment, not including transportation time.
- Security clearance. Each piece of equipment goes through a lengthy "pre-import" approval and post-arrival "import release" approval processes. But everyone involved knows that computer hardware doesn't pose any cyber-security threat. Cyber threats are generally software and network based. In fact it's difficult to think of a single example in the world of a national security problem which can be solved by controlling which computers are allowed into the country. To be blunt, it is a gigantic waste of time. The national security interest should be focused on the energy security of the country. Is the power company planning the supply and demand correctly? The current approach makes no more sense that limiting the number of light bulbs that enter the country.
- Import duty and investment incentives. Capital investments are charged lower import duties. This is an understandable objective. However, the process of qualifying is extremely onerous and often arbitrary. It requires an investment license which creates enormous complexity as I've written about before. Once you start operations you have to switch from an investment license to an expansion license, which has a new set of requirements like minimum payroll etc. Each shipment has to be gruelingly reviewed and approved by the investment commission, by the ministry of finance, as well as customs commission. Besides the inefficiency, the uncertainty is huge. Imagine importing equipment not knowing if the import duty will be 3% or 40% until after it arrives. Any misstep carries the risk of huge penalties, and at worst, the equipment can in theory be confiscated. For a data center, capex efficiency is a life or death matter, so the import risks are high enough to scare away most rational investors. It makes it very hard for a company to grow investment progressively. It would be much better to simply have a constant low import duty, say 5%, for all equipment and get rid of the investment license. The businesses would be happier, and the government would likely make more revenue from the boom.
- Customs. Besides security and investment approvals, you have to go through customs. These offices are notoriously inefficient and unfair. To make matters worse, the more they delay, the more revenue they generate. Conversely, if they clear stuff too fast and at a low price, the individual customs agent can be suspected of corruption. This creates a cruel incentive for them to become ever more inefficient. A crucial reform would be to a) introduce the following key performance indicator: total customs revenue divided by average time to clear shipments. Tie it to salaries and promotions of the customs officers and their managers. Those who get the most revenue in the least time would be rewarded. And b) create a public database of assessed value and duties paid for all shipments. (The public database shouldn't have the name of the importer, just the numbers). This combination would help eliminate unreasonable delays and corruption, as both the importer and the customs agent would be incentivized to quickly agree on the correct value.
For once, Ethiopia has something that the whole world really needs. It can be the Saudi Arabia of electricity. (Ok the second time, if you count coffee. But as I've written about before a long time ago, it's tough for a producer to capture its fair share of the coffee value chain). Let's not mess it up! We need the best possible energy pricing and the best import process.
