AI Investment: The US Data Center Tax Squeeze

What happened: Congress targets full expensing

In July 2026 Senator Mark Warner filed the Data Center Tax Accountability and Disclosure Act. The text blocks the full expensing provided by Sec. 168(k) for any property used in an AI data center.

In August 2026 the Finance Committee ranking member, Ron Wyden, followed up with a white paper heading in the same direction, aimed at denying the immediate deduction to new data centers.

Two distinct signals from Congress converge on the same target: the tax appeal of investing in American computing infrastructure. Goldman Sachs estimates that capital expenditure by a handful of US hyperscalers will reach 581 billion dollars in 2026. This is the clearest signal so far that AI infrastructure is entering the crosshairs of fiscal policy.

The stakes go beyond a single fiscal year. They concern where the next computing campuses will be built. Full expensing allows the entire cost of an investment to be deducted immediately, instead of spreading it over several years. Removing it means lengthening the time needed to recover capital. For a sector that thinks in billion-dollar spending cycles, this changes the calculation of every new project.

What these proposals really are

Behind the technical language, the Warner proposal draws an arbitrary line around a specific type of investment.

An "AI data center" is defined as a facility dedicated to IT equipment, telecommunications networks and data storage and processing services, used at least 20 percent to develop or operate AI systems. The 20 percent threshold generates immediate operational uncertainty, because measuring the use of a mixed facility remains a contestable exercise.

The point weighs on balance sheets. An operator running different workloads in the same building must demonstrate how much capacity serves AI. The calculation becomes fertile ground for disputes with the tax authorities. Uncertainty over qualification translates into uncertainty over tax treatment.

Facilities certified LEED at Platinum or Gold levels retain full expensing. The legislator therefore ties the tax benefit to energy and environmental efficiency.

This turns a tax rule into a lever of industrial policy. The message to operators remains direct: build more efficiently, or pay more.

The competitive positioning shift

The competitive axis shifts from the pure cost of computing to the tax geography of the investment.

Those with the flexibility to place capacity where cost recovery remains full gain a structural advantage. Operators tied to US territory face slower investment recovery and a higher cost of capital.

The competitive moat is now built on the ability to optimize the tax footprint, as well as on chip power.

The vendor with the most diversified data center network captures a more durable return than the vendor betting solely on the greatest compute density.

Competition moves from the race for gigawatts to the management of the tax regime. The market is moving toward this reading.

Who is affected

The measure directly hits hyperscalers, colocation providers and the local communities in the running to host new capacity.

For the large cloud operators, losing full expensing raises the effective cost of every new campus. For American cities, the risk is investment fleeing toward more welcoming jurisdictions, with a loss of jobs, economic growth and tax revenue.

The Tax Foundation warns that rules of this kind add complexity to the code and may push AI investment abroad.

The flip side should also be stated: the proposals arise to limit a tax subsidy that favors a few large companies. Their supporters aim to recover revenue and reward the most efficient facilities. LEED certification offers a way out, not an absolute bar. The real effect will depend on how many operators choose that path.

This has direct implications for Microsoft, Amazon and Google, the three operators with the most aggressive infrastructure expansion plans. Each will review the map of its next sites.

The strategic question for CFOs and CDOs

The board must reconsider the infrastructure spending line of the next fiscal year.

For the CFO, the question remains concrete: what share of planned capacity risks losing the immediate deduction? Every billion of capex becomes more expensive when tax recovery is stretched over several years. The cost of capital thus enters the strategic conversation.

For the Chief Digital Officer, the choice of vendor takes on new weight. A supplier with a global footprint offers better contractual cover against domestic tax uncertainty.

Lock-in to a single US-based operator becomes a balance-sheet vulnerability, as well as a technological one. Supplier diversification moves from option to priority.

The signal for the technology investor

The market thesis underpinning this affair remains clear: deployment beats the model.

The proposal confirms that value shifts toward those who control the physical infrastructure and its tax optimization. The model remains a commodity; the integration and localization of data centers become the real pricing factor.

For the investor, diversified colocation providers and operators with international exposure deserve an upward reassessment.

Computing assets anchored to a single tax regime carry a higher risk premium. Pricing pressure will come from margins eroded by slowed tax recovery.

One limit must be kept in mind. The two initiatives are, for now, a bill and a white paper. Neither is law in force. The distance between announcement and rule in production can be wide. The investor must price the probability, not the certainty.

What to decide in the next 90 days

Next quarter's budget cycle calls for three concrete moves.

• Map the share of infrastructure capex exposed to the 20 percent threshold defined by Warner.
• Open a dialogue with cloud vendors on their ability to shift workloads toward tax-favorable jurisdictions.
• Assess LEED Platinum or Gold certification as a requirement in new hosting contracts, now a tax lever as well as an environmental one.

The Chief Strategy Officer must treat this proposal as a catalyst for vendor consolidation. A partnership with a multi-jurisdiction operator becomes more urgent than a deal with a single-base supplier.

The market has already begun to price regulatory uncertainty. Those who decide now protect the return on investment; those who wait pay the risk premium. Related analyses remain available on our blog.