Trump dismisses calls for AI slowdown from leading tech CEOs

President Trump rejected calls from Anthropic, OpenAI and Elon Musk to slow AI capability development, prioritizing the US lead over China while signaling only unspecified guardrails. The stance clashes with growing bipartisan congressional support for human-oversight rules following researcher Jacob Coxon’s resignation over AI-control risks. AI data-center expansion is also becoming an election-cycle issue, with local concerns focused on water availability and higher electricity bills.

Analysis

The investable implication is not a near-term change in AI demand, but a lower probability of a federally imposed capability cap that would disrupt hyperscaler capex plans. That modestly supports the premium paid for AI-exposed infrastructure earnings—NVDA, AVGO, VRT, ETN and DLR—while preserving the cloud platforms’ incentive to compete on model deployment rather than compliance. The more material bottleneck remains physical: permissive federal rhetoric does not override local interconnection, water, zoning and rate-case constraints, so incremental AI spending should continue to migrate toward power-secure regions and providers rather than simply increase aggregate data-center returns.

The second-order winner is dispatchable and contracted power, particularly CEG, VST, NRG and selected gas-turbine supply chains such as GE Vernova, because accelerated load forecasts strengthen their negotiating position with utilities and data-center customers. Conversely, regulated utilities with weak transmission plans or politically sensitive retail-rate exposure can face a capex/rate-lag trap: they must fund grid upgrades before returns are recognized, while public opposition raises the risk of disallowances. This divergence should emerge over 6-18 months through load-forecast revisions, interconnection queues and state commission outcomes, not in the next few sessions.

Consensus may overread the administration’s posture as deregulation. Congressional pressure, state-level AI rules and procurement standards can still create a fragmented compliance regime, while the largest platforms may privately favor rules that raise fixed costs for smaller model developers. The thesis is falsified if hyperscalers cut 2026 capex or cite power availability as a binding deployment constraint; a federal executive action or bipartisan bill imposing mandatory human-oversight standards would also compress the relative premium in AI infrastructure.

Key Decisions for Investors

• Maintain a 3-6 month long VRT / short DLR relative position: VRT has greater exposure to high-density AI buildouts, while DLR remains more exposed to permitting, power procurement and development-cost inflation. Exit if hyperscaler capex guidance is reduced or VRT order growth materially decelerates.
• Accumulate CEG or VST on broad market weakness for a 6-18 month horizon, sized against regulatory risk: AI load growth can improve contracted-power pricing, but take profit if power-price expectations outrun forward contracted earnings. Monitor state rate-case filings and large-load interconnection approvals as the key catalysts.
• Avoid adding outright NVDA exposure solely on this policy signal; the policy probability change is unlikely to alter the next two quarters of revenue. Use any announcement of concrete federal AI restrictions, or a material hyperscaler capex-guidance cut, as a trigger to hedge semiconductor beta through SMH puts rather than preemptively shorting.
• Screen regulated utilities for exposure to data-center load without approved transmission recovery; consider underweighting names where retail-rate backlash is increasing. The adverse catalyst is a commission decision limiting recovery of grid upgrades, which can impair allowed-return realization before AI load becomes revenue-generating.