The whale didn't wait for the grid to catch up. CoreWeave, the GPU cloud provider that pivoted from crypto mining to AI compute, just dropped £8.2 billion on a Scottish datacenter. But the real story isn't the capital—it's the current. Scotland's grid was never built to feed a 1 GW beast.

Context: From Mining Rigs to AI Clusters CoreWeave started as a crypto miner. They know power arbitrage. In 2017, they bought cheap GPU time; in 2023, they renegotiated Microsoft's H100 supply. Now they want to build the largest single AI compute campus in Europe—inside a region that sells wind energy to England but can't even keep the lights on for its own villages. The £8.2B figure is impressive until you realize the hidden CAPEX: transformers with 18-month lead times, substations that Scotland's grid operator (SSEN) hasn't even budgeted for, and a local population that's already burning wood for heat because electricity prices are too high.
Core: The Ledger Doesn't Blink Let's talk hardware. A single Nvidia H100 draws 700W at peak. A typical rack in this datacenter will pack 40-60 GPUs, pulling 30-60 kW per rack. Multiply that by thousands of racks—we're looking at a total load of 500 MW to 1 GW. That's a small nuclear reactor. Scotland has wind capacity, yes—over 13 GW installed. But capacity ≠ deliverability. The transmission line to the Highlands is already saturated. Upgrading it takes 5-7 years and costs billions. CoreWeave's timeline? 18-24 months.
Contrarian: The Hidden Signal Everyone screams "renewables" until the GPU farm turns on. The real opportunity here is not power supply—it's power storage. CoreWeave could become the first major AI cloud to deploy Tesla Megapacks at scale, turning grid instability into a competitive moat. Volatility is the tax on the unprepared. If they solve this, they won't just run a datacenter—they'll run a virtual power plant, selling frequency response to the UK grid while training LLMs. Or they'll fail, and the $8.2B will become a tombstone for every other AI cloud with the same delusion.

Takeaway Scotland's wind isn't the bottleneck. The bottleneck is the imagination of utilities and the patience of VCs. CoreWeave has 12 months to find a workaround before the next round of funding gets repriced. Watch for announcements on battery storage partnerships—or a quiet sale of the land to Digital Realty.
