Last week, Meta Platforms did something that sounds more like a national grid project than a corporate IT purchase. It secured roughly 1.6 gigawatts of AI computing capacity from data-centre developer Crusoe, spread across two sites in Childress, Texas, and Warrenton, Missouri.
To put that number in perspective: one gigawatt is roughly the output of a large power station — enough to supply electricity to about 750,000 U.S. homes. Meta just contracted for more than two of those, dedicated entirely to running AI workloads.
This deal is not happening in isolation. It sits at the intersection of three forces reshaping the global technology landscape: the insatiable demand for AI compute, the physical constraints of energy infrastructure, and a structural memory-chip crisis that's now squeezing even Apple.
The Scale of the Arms Race
When we talk about AI infrastructure, the unit of ambition has shifted. It's no longer about how many GPUs a company owns — it's about how many gigawatts it can access. Meta's Crusoe deal adds to an already staggering build-out that includes a nearly 4,000-acre campus in Louisiana designed for up to five gigawatts. Combined, Meta's infrastructure pipeline now rivals the power consumption of small countries.
And it's not alone. Microsoft, Google, Amazon, and Oracle are all doing the same thing. Crusoe — a company founded in 2018 — already has deals with Oracle, Microsoft, and Google for other campuses. The AI compute market has become a land grab where speed of deployment matters more than cost, and where traditional data-centre construction cycles are too slow for the pace of AI development.
Why This Matters Beyond Big Tech
This is where it gets relevant for everyone else — including the Zimbabwean tech ecosystem where Red Cup Series operates.
The concentration of AI compute in the hands of a few US hyperscalers has knock-on effects globally:
Memory and chip prices. The same AI gold rush is driving a global memory crisis. HBM (high-bandwidth memory) — essential for AI training — is being hoovered up by data-centre operators. CNBC reported this week that the crunch has gotten so bad that "even Apple can't be safe," with Tim Cook warning of price increases on consumer devices. For African startups building AI products, this means hardware costs remain punishingly high.
Energy constraints. Data centres at this scale consume massive amounts of power and water. Grid operators in Texas, Virginia, and elsewhere are pushing back. Meta's answer is geographic diversification — spreading across multiple developers and regions. But for most of the world, access to reliable, affordable power for compute is still a dream. Zimbabwe's persistent load-shedding makes this gap feel even wider.
Talent concentration. The AI compute explosion is also a talent magnet. The engineers who know how to optimise models for these massive clusters earn Silicon Valley salaries. The brain drain from the rest of the world to the centres of AI infrastructure is real.
What This Means for Emerging Markets
The narrative isn't entirely bleak. The Crusoe deal also signals something important: the hyperscalers are running out of capacity to build fast enough themselves. That opens space for alternative models — edge computing, smaller AI infrastructure plays, and localised data centres that serve regional needs.
Zimbabwe's recently launched National AI Strategy 2026-2030 creates a policy framework for exactly this kind of local infrastructure thinking. While we can't compete on Meta's scale, we don't need to. What we need are practical, energy-efficient AI deployment models that work within our constraints — the same insight that drives Red Cup Series' approach to AI integration in African markets.
The Meta-Crusoe deal is a reminder that AI's future will be shaped not just by algorithms and models, but by who controls the physical infrastructure that powers them. And that infrastructure race is just getting started.
Photo by Julio Lopez on Unsplash
