US Steel Sector "Do Data Centers Really Matter?" Jones

Research evidence excerpt

US Steel Sector "Do Data Centers Really Matter?" Jones

.

Digesting US data center demand

The UBSe tech team forecasts US data center buildout steps up to ~15/20/21 of

incremental operational GWs in 2026/27/28 (from ~2 GW 2018-2022; Figure 1IncrementalOperationalMWs). While

data center designs are not uniform, most announcements over the last ~2yrs have been

in the 4-5k square feet per MW range. We recently hosted NUE in London (note) where

they provided an estimate of ~14kt of steel per million square feet of data centers. This

implies US steel direct demand impact from data centers is ~950/1,250/1,300kt in

2026/27/28 (Figure 2DataCenterUSStelDemand(mt)) or ~0.9/1.2/1.3% of 2025 demand (~102mt). We also expect a

second derivative effect from the associated power buildout from data center demand.

The UBSe utilities team forecasts ~83GW of data center power generation demand from

2025-2030, including ~12G/16/21GW in 2026/27/28, with solar & battery storage the

most common power sources. Assuming steel demand of ~60kt for 1GW of solar

& ~12kt for 1GW of battery storage implies power generation buildout demand of

~570/770/850kt in 2026/27/28 (Figure 9DataCenterRelatedPowerGenerationBuildoutUSStelDemand(mt)) or ~0.7/0.9/1.0% of 2025 demand (~102mt).

How much do data centers move the needle for our coverage?

NUE: NUE acquired Southwest Data Products, a manufacturer & installer of data

center infrastructure, for $115m in April 2024 while simultaneously launching

Nucor Data Systems, a business unit focused on data center infrastructure (note).

On the 1Q call, NUE noted data centers were 10% of the backlog. If average

backlogs are ~2m shipments, that would imply 400kt+ now (inclusive of power

infra such as towers). Data center related shipments should increase over the year,