The AI Race Has Become Too Big to Stop

The four major technology companies spent about $165 billion in cash on property and equipment in the second quarter of 2026, exceeding their combined spending for all of 2023. Moss presents this as evidence that the AI infrastructure buildout is accelerating.

Combined annual spending by Microsoft, Google, Amazon, and Meta rose from about $147 billion in 2023 to roughly $376 billion in 2025, while they had already spent almost $295 billion by mid-2026.

Moss says the AI race has become an infrastructure project involving chips, data centers, power, cooling, fiber optics, and transmission rather than merely a software cycle. The commitments extend years into the future through leases, equipment orders, and power contracts.

Microsoft, Google, and Meta are committing substantial future capacity before all of it is operational. Microsoft disclosed about $329 billion in additional data-center-related leases, while Meta has hundreds of billions of dollars tied to future capacity.

Moss describes the competitive dynamic as a red queen effect: a company that stops investing while rivals advance may see the economic value of its existing capital decline relative to competitors. He says standing still can therefore become a form of falling behind.

The host argues that competition can push companies to invest beyond what would make sense for the economy as a whole, citing a Bank for International Settlements model. He says this force could prolong the boom while also creating too much capacity.