Kevin O'Leary says Utah AI data center project will shrink after lawmakers demand cuts

Kevin O'Leary says the proposed Stratos AI data center in Utah may be cut from 40,000 acres to about 10,000 acres after lawmakers demanded a 75% reduction and stronger environmental safeguards. Utah Senate President J. Stuart Adams wants tighter water-use, wildlife, and public-transparency commitments before the project advances. The news is mainly a project-level planning update, with limited immediate market impact beyond signaling more regulatory scrutiny of large AI infrastructure builds.

Analysis

This is less a cancellation risk than a capital-intensity reset: once the acreage is forced down, the project’s economics likely shift from a “build-the-city” thesis to a phased campus model. That favors firms with modular power, cooling, and networking exposure over land-bank style developers, because the first dollars now need to go into immediately monetizable compute rather than speculative horizontal expansion. In practice, the market should treat this as a negative for anyone underwriting mega-campus scarcity premiums and a relative positive for equipment vendors that benefit from repeated build cycles rather than one-shot megaproject execution.

The second-order issue is permitting velocity. A smaller footprint can reduce the probability of outright political failure, but it also invites more scrutiny on water, grid interconnects, and local transport over the next 3-12 months. That creates a “drip risk” profile: headline downside may be limited, but every incremental approval becomes a catalyst window where delays can reset contractor schedules and push out revenue recognition for the supply chain. The key tell will be whether the revised plan preserves enough scale to justify bespoke utility investments; if not, the project becomes a lower-return anchor tenant rather than a catalyst for a regional AI cluster.

Consensus may be underestimating how this changes bargaining power. If Utah compels a smaller, more resource-efficient design, other states can now demand the same concession package, which compresses the value of raw land and increases the value of jurisdictions with pre-approved power/water infrastructure. That is mildly bearish for speculative data-center land assemblers, but bullish for incumbent utilities and interconnect-heavy electrical contractors that can win on speed-to-power. The contrarian read is that the market may be too focused on the size reduction as a setback, when the bigger implication is a higher probability of eventual buildout—just slower, more expensive, and with better discipline.

Key Decisions for Investors

• Long VRT / short a basket of speculative data-center land or campus developers for 3-6 months; thesis is that modular buildout benefits thermal management and power-distribution leaders more than acreage stories.
• Initiate a pairs trade: long ETN or HUBB, short utility-like infrastructure proxies that depend on a single mega-project being executed on time; risk/reward improves if revised approvals trigger phased capex rather than cancellation.
• Buy 3-6 month call spreads on VRT or ANET into any pullback tied to regulatory headlines; the asymmetry is that phased projects still require networking and cooling spend, while outright cancellations are the primary bearish case.
• Avoid chasing pure-play data-center REIT momentum until the revised proposal is disclosed; if the footprint keeps shrinking, expected IRR on large-campus pipelines likely compresses by 200-400 bps, pressuring valuation multiples.
• If broader AI-infrastructure names sell off on this news, use it to add to quality beneficiaries with utility-scale power access rather than land exposure; the market is likely mispricing the difference between delay risk and terminal demand risk.