America’s data centre backlash holds a clear lesson for Africa: chase the value AI creates, not the concrete it runs on. The voters, S&P Global and a new Cornell University study all point the same way.
This release in 60 seconds:
S&P Global Ratings expects six major US hyperscalers to spend more than $7 trillion on data centres and AI between 2025 and 2030, and to burn more cash than they make until at least 2027.
In this month’s US midterms, voters chose to restrict or limit data centres in 40 of 41 local ballot measures, after a Cornell University study found data centres paid only 38% of the grid costs linked to their growth.
South Africa is forecast to carry 77% of Africa’s additional data centre electricity demand through 2035, and South Africans should be asking now who will pay for the grid to carry it.
Phil Pelucha argues the quick AI wins sit in systems that already exist: one client, a tech consulting firm, added around 1,000 hours of work a month with AI agents, without replacing a single person.
The full story:
Everyone is asking the same question about AI right now. Not “is it useful?” We know it’s useful. The question is whether the money comes back fast enough to pay for the trillions going into data centres.
America is answering that question live. S&P Global Ratings expects six major US hyperscalers to spend more than $7 trillion on data centres and other AI investment between 2025 and 2030, and to burn more cash than they make until at least 2027. Investors have started asking harder questions about who pays for the build-out, and how long they are prepared to wait for it to pay back.S&P Global Ratings
Then the voters had their say. In this month’s midterms, voters chose to restrict or limit data centres in 40 of 41 local ballot measures across 10 states. Why? A Cornell University study published in Nature Energy found that data centres in four major US grids paid only 38% of the infrastructure costs linked to their growth. Most of the rest fell on households. When the bill lands on the kitchen table, the politics turn.40 of 41 local ballot measuresCornell University study
Now bring that home. South Africa is forecast to account for 77% of Africa’s additional data centre electricity demand through 2035. Planned projects could take the country’s data centre load past 1,200 megawatts, more than a stage of load shedding’s worth of power, and the national grid upgrade plan is costed at R440 billion. The grid will need upgrading to carry all of this, and South Africans should be asking now who will pick up the tab. That should worry anyone who pays an electricity bill.77% of Africa’s additional data centre electricity demandpast 1,200 megawattscosted at R440 billion
This is not an argument against digital infrastructure. Well-planned investment brings skills and jobs. It is an argument about who carries the cost, and whether we are building for our own returns or for someone else’s.
Where is the money today?
As a revenue architect, I always ask one question first: where is the money today? Today the money is in using the systems and the compute that already exist, not in trying to build the compute ourselves.
Let me give you two real examples. One of my clients, a tech consulting firm, put AI agents to work alongside its team. AI didn’t replace a single person. It gave the business about 1,000 extra hours of work a month, worth about R5 million a year in salaries it didn’t have to pay.
Another, Springbok Properties, is an international real estate business with a presence here in South Africa and a sales team of more than 100 people. It started by finding where its revenue was leaking, then put AI agents to work, including AI sales trainers that listen to every sales call and give feedback. It saves about R22 million (about £1 million) a year in recruitment fees and lost revenue, and nobody was replaced.
Book the hotel
So if you run a business and you want the AI money, the quick wins aren’t in building anything. They’re in putting AI to work alongside your people, picking up the revenue you’re missing today and opening up opportunities you couldn’t reach before. Make your team better, don’t replace them, and the money follows.
What does using it look like? In a sales team, AI can do the research, the scheduling and the first qualifying of a prospect, so your people spend their time understanding the buyer’s real problem. That is where deals are won. The business that wins is not the one with the most compute. It is the one that puts the compute it already has to work on the revenue it is already losing.
Start small. Pick one problem you can measure, such as slow follow-up on leads or proposals that sit unanswered, and put AI to work on that first. Use other companies’ case studies for inspiration, not as a shopping list. Once the first problem pays for itself, move to the next.
When you go on holiday, you book a hotel that already exists. You don’t build your own. AI is the same. The compute is already there. Use it. Book the hotel. Don’t build it.
One warning. AI is a multiplier, not a strategy. If the process underneath is broken, all you’ve done is make your business more efficient at failing. Find where the revenue is leaking first, fix the process, then put AI to work on it.
The bill always lands somewhere
America’s lesson is simple. Compute without returns is a liability, and the bill always lands somewhere. Africa should be where AI pays for itself, not the data centre dumping ground of the world, where the servers are someone else’s and the bill is ours.
The money in AI isn’t in building the computer. It’s in using it.
Key insights:
$7 trillion: What S&P Global Ratings expects six major US hyperscalers to spend on data centres and other AI investment between 2025 and 2030.
40 of 41: US local ballot measures in which voters chose to restrict or limit data centres in the November 2026 midterms.
38%: The share of grid infrastructure costs linked to their growth that data centres paid in four major US grids, according to Cornell University (Nature Energy, 2026).
77%: South Africa’s forecast share of Africa’s additional data centre electricity demand through 2035 (Electron Intelligence).
R440 billion: The cost of South Africa’s national grid upgrade plan (Bloomberg, November 2025).
1,000 hours a month: The extra work AI agents do for one of Phil’s clients, a tech consulting firm, worth about R5 million a year in salaries, with nobody replaced.
In their own words:
“As a revenue architect, I always ask one question first: where is the money today? Today the money is in using the systems and the compute that already exist, not in trying to build the compute ourselves.” – Phil Pelucha, founder of Billionaires in Boxers, the Revenue Acceleration Consultancy of the Year 2026 (UK)
“Africa should be where AI pays for itself, not the data centre dumping ground of the world, where the servers are someone else’s and the bill is ours.” – Phil Pelucha, founder of Billionaires in Boxers, the Revenue Acceleration Consultancy of the Year 2026 (UK).
- Phil Pelucha is the founder of Billionaires in Boxers, named Revenue Acceleration Consultancy of the Year 2026 (UK Elite Business Awards). He was also named Revenue Acceleration Consultant of the Year 2026 (US, NY Weekly), and Clutch has named him one of its Top 25 Global B2B Sales Influencers. He increases sales for B2B companies around the world, starting with the revenue already sitting in a business’s pipeline, then building the sales system, the AI and the team that keep it growing. He is the author of The Underdog Champion (2023).
