The artificial intelligence boom runs on data centers—and, according to ARK Invest Chief Executive Cathie Wood, the capital building them is overwhelmingly American.

In ARK Invest's "In The Know" video, released in early October 2026, Wood cited Bloomberg data showing that nearly 90% of global data center financing is based in the United States.

What ARK actually said

Of all the capital flowing into data center construction worldwide, roughly nine of every ten dollars is being financed in the US, according to the Bloomberg figures Wood referenced. Wood framed the statistic as evidence of America's lead in the AI investment race.

She tied the concentration directly to the US tax code, pointing to two features in particular.

The first is 100% first-year depreciation for qualifying assets. Ordinarily, when a company purchases expensive equipment, it writes off the cost gradually over several years. Full first-year depreciation instead allows a company to deduct the entire cost up front. For capital-intensive projects like data centers, the stakes of that distinction are high: the bigger the upfront outlay, the bigger the immediate deduction.

The second factor Wood highlighted is an effective corporate tax rate of around 8%. That figure represents the share of profits companies actually pay after deductions and incentives—not the headline statutory rate printed in the tax code.

The scale of the buildout

Hundreds of billions of dollars in debt and project financing have been reported for AI data centers. Data center systems spending has been projected triple to $1.4 trillion over five years, a forecast dating from January 2026. The outlays reflect the physical backbone of AI: warehouse-scale facilities that house the servers, power systems, and cooling infrastructure required to train and run AI models.

ARK's October 2026 portfolio changes emphasized exposure to AI infrastructure, a move that lines up with the thesis Wood presented on camera.

What this means

If nearly 90% of the money is landing in a single country, the businesses absorbing that capital are concentrated there as well. Such concentration cuts both ways: when so much of a global buildout depends on one jurisdiction's tax code, that code becomes a single point of failure, and incentives like full first-year depreciation are policy choices that can change.

The debt load is the other factor to watch. Hundreds of billions in borrowing for AI data centers means lenders are betting that demand for computing power keeps rising fast enough to service those loans.

Investors following Wood's thesis have two signals to track closely: any changes to the tax provisions she credited, and whether reported financing volumes keep pace with the spending forecasts projected to reach $1.4 trillion.