A recent projection from McKinsey shows $35 billion will be spent building data centers worldwide, rising to $49 billion in 2030. Of that, 36% will be allocated to hyperscale centers. The growth rate is about 5% per year across the board.

My research at IDCA Smart Nations shows that this number may not be enough. The developing world on its own will require a large percentage of it just to achieve any sort of tangible progress.

Background
Among the disparities of the world in income, physical infrastructure, and digital infrastructure, the disparity among the number of servers and data centers in each country is the highest of all.

Server density – measured as number of servers per 1 million people – ranges from fewer than 10 among the least-served 10% of the world's population to more than 50,000 in the highest 10%. This is a disparity of 5,000X.

Data consumption per person, on the other hand, shows a top-to-bottom disparity of about 20X.

This confirms through data what we already know anecdotally – servers and the data centers that house them have been built in dense clusters that serve large regions. Developing regions will have a thinner server density, but nevertheless a key hub or two – we can think of Singapore, South Africa, Brazil and Chile, and to some extent the UAE as examples.

This centralized, clustered distribution will need to change if digital infrastructure is truly a key to positive, global socioeconomic development. There will always be certain data center concentrations, often around a metro area with a concentrated financial services industry, healthcare, or manufacturing base.

Edge vs. Massive
But a quantum leap in the number of smaller, edge or microlocal centers seems like a requirement to bring more wealth and less disparity into developing nations. The idea of a number of rooftop facilities pulling 100KW rather than another HPC center pulling dozens of megawatts can deliver government services as well as online financial and retail services more quickly to more people, with less dependency on (and risk from) a mammoth facility that may be hundreds of miles away in another country. The big cloud providers will always have their economies of scale and performance as they push their services to thousands of customers, as will Facebook serving tens of millions of individuals.

But that $49 billion in 2030 may run out pretty quickly. Let's run the numbers. Coming up with a one-size-fits-all metric for data center construction costs can be a fool's game. That said, let's use $5 million per megawatt as a rough guide.

IDCA Smart Nations Research numbers show that $1 billion will be required for each single-percentage-point increase in the developing world's server density relative to the developed world's density. (This does not include China and India, which are always special cases.)

In other words, as things stand, the developing world's server density is 0.02% of the developed world's; bringing this up to 1% would require $1 billion in data center construction costs. So, bringing up the density to 20% of the developed-world standard would require $20 billion, or about 40% of the entire projected 2030 global expenditure.

Consequential Change
That 20% may not seem consequential, but of course it would be. A digital infrastructure of that scope would be comparable to today's Costa Rica or China, one with a very small population and the other with the world's largest, each progressing steadily toward developed-nation status.

Perhaps that $49 billion figure will indeed be adequate for the world's needs in 2030. Calculating data growth worldwide, the strain it will place on existing digital infrastructure and the investment required to meet the demand is itself a long, complicated task.

But in any case it's time to start thinking about the developing world, how to develop it, and what this development may look like. Hypothetical numbers such as presented make for great discussion and debate. Actually planning, building, and then operating any new data center of any size is a difficult task with a typically long timeframe.

Any sort of ambitious digital infrastructure construction plan must also go along with planning for electricity, physical infrastructure, and the people and economy that can support and benefit from the shiny new data center.

Image from RackCentre.