In early 2026, Big Tech's major players announced staggering capital expenditure plans for AI infrastructure. Amazon: $200B. Google: $175-185B. Meta: $115-135B. Microsoft: ~$145B (fiscal year run-rate). Total: $635-665 billion—up roughly 70% from 2025's $381 billion.
Where the Money Goes: Almost all of this is going to AI chips (GPUs, TPUs), servers, and data center construction. The spending reflects an arms race for computational capacity to train and run frontier models.
The Constraints: Morgan Stanley and S&P Global warned that rising electricity costs and geopolitical tensions (affecting oil prices used in power generation) could constrain planned buildouts. Delays in power-plant construction are already creating chip inventory backlogs.
The Energy Problem: AI data centers are already raising electricity costs in some regions. Microsoft paused some data center projects last year due to power concerns. The seven largest tech companies signed a "ratepayer protection pledge" in March, pledging to build their own power infrastructure rather than strain public grids.
Market Impact: After announcing spending plans, Meta stock rallied (profits are rising), but Microsoft and Amazon stocks fell. Investors are asking: will this spending translate to revenue growth?
My Take: The spending itself isn't surprising—it reflects rational competition. But the energy and infrastructure bottlenecks are real and underestimated. Whoever solves power and cooling costs wins the next cycle.