Six major companies reach $1 trillion annual investment pace, analysis finds

Six major companies’ capital spending reached a $1 trillion annualized pace in the third quarter of 2026, according to Matthew C. Klein’s November 3 analysis in The Overshoot. Alongside that corporate surge, U.S. business investment in computers, telecommunications equipment and data-center shells was running more than twice its pre-ChatGPT trend—a sharp acceleration in the physical infrastructure behind the AI boom.

Klein’s company group comprises Amazon, Google, Meta, Microsoft, Oracle and SpaceX. Their reported investment spending tripled from the start of 2025 through the third quarter of 2026. Measured from ChatGPT’s launch at the end of 2022, their quarterly capital expenditure increased sevenfold. The striking finding is therefore not just the size of the bill, but how quickly it has grown.

The $1 trillion figure describes the annual pace implied by third-quarter spending, not a completed full-year total. Klein draws the company and market data from Koyfin, using a group for which he says roughly comparable audited financial statements are available. That makes it a defined corporate sample, rather than a count of every company investing in AI infrastructure.

A hardware surge, not a matching software surge

The economy-wide measure captures a different view of the buildout. Spending on computers, telecoms equipment and the buildings that house data centers amounted to 2.5% of U.S. economic output in the third quarter. That compares with an average of 1.3% over 2009–2024. These are broad investment categories, not a separately measured total of spending exclusively devoted to AI.

Software investment has not followed the same path. Klein finds that business spending on software continued growing at its pre-ChatGPT trend. The contrast places the exceptional acceleration in equipment and facilities, rather than showing an equivalent break across both hardware and software investment. It is a narrower finding than a uniform surge in technology spending.

The buyers are changing too

The six companies do not all buy computing infrastructure for the same purpose. Klein describes Amazon and Oracle as focused on selling computing capacity to others. Google, Meta, Microsoft and SpaceX also invest in computing capacity for their own AI labs. The spending total therefore spans both capacity offered to customers and infrastructure supporting internal model development.

Anthropic and OpenAI sit outside that six-company sample. Klein characterizes them as net consumers of computing capacity, rather than net suppliers, but expects them to become significant direct buyers of data-center equipment. That would add a direct purchasing role to the demand they already generate indirectly by renting capacity from other companies.

Dedicated infrastructure without doing every job

Anthropic’s S-1, as quoted by Klein, illustrates the shift. The company says it is pursuing dedicated data-center infrastructure to reduce dependence on cloud providers, address computing constraints and secure capacity for frontier-model development. It expects capital expenditures to increase significantly as that buildout continues. Those are stated plans and commitments, not a completed infrastructure expansion.

Its approach combines long-term leases with partners procuring real estate, power and facilities. Anthropic leases capacity and contracts with specialized operators to manage facilities. It also buys some computing and infrastructure equipment directly for dedicated sites operated on its behalf and to its specifications. The filing also describes arrangements involving dedicated power generation or other energy infrastructure.

Further along the supply chain, Klein identifies Broadcom, Micron, Nvidia, Samsung Electronics, SK Hynix and TSMC as capturing much of the increased equipment spending. His analysis links the expanding buyer base to a concentrated group of manufacturers supplying the hardware inside—and connecting—the server racks.