AI demand alone will require $5.2 trillion in investment by 2030 according to McKinsey

https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers

McKinsey’s new report paints a massive, GPU‑hungry future for AI infrastructure. Analysts estimate global data‑center capacity will almost triple by 2030, with 70 % of the load coming from AI workloads. Meeting that demand will require nearly $7 TRILLION in capex, including $5.2 T earmarked for AI alone. Crucially, about 60 % of the total—roughly $3.1 T—will flow to “technology developers & designers,” the category that explicitly lists NVIDIA as the key supplier of the chips that power AI. In other words, three out of every five dollars poured into the coming compute build‑out lands in Jensen’s backyard.

The report also explains why NVIDIA is poised to convert this investment tidal wave into outsized profits. GPUs remain the default engine for both training and inference, and McKinsey notes that only a handful of companies can supply them at scale—a textbook supply‑side moat. Meanwhile, hyperscalers, governments, and enterprises are locked in an arms race to deploy proprietary AI capacity, effectively guaranteeing multi‑year demand visibility and pricing power for whoever controls the silicon. Put it all together, McKinsey’s forecast isn’t just bullish for GPUs—it’s a $3‑trillion endorsement of NVIDIA’s roadmap.