Trump says calls for more control on AI are a ‘SICK conspiracy’

AI-linked equities sold off sharply after prominent AI executives warned of potentially uncontrollable systems and called for a slower pace of capability development: the Philadelphia Semiconductor Index fell 6%, Nasdaq 100 declined 1.7%, and Nvidia, AMD and Micron dropped 3.5%, 5.6% and 6.7%, respectively. President Trump rejected calls for additional AI guardrails, arguing stricter oversight could allow China to gain an AI advantage, while Anthropic CEO Dario Amodei warned AI agents could cause hundreds of billions of dollars in damage within 6-12 months. OpenAI reportedly delayed its planned IPO, which had targeted a valuation of up to $1 trillion, as investors reassessed AI-spending and regulatory-risk assumptions.

Analysis

The investable issue is not a near-term federal permitting freeze; it is a widening gap between Washington’s pro-build posture and voluntary safety constraints from frontier-model vendors. That gap raises the odds that hyperscalers diversify workloads and model partnerships rather than pause infrastructure commitments. NVDA, TSM and HBM suppliers retain the strongest 6-12 month demand visibility because installed compute is still needed for inference, sovereign AI and non-frontier workloads, while AMAT/LRCX/ASML are more exposed to a delayed 2027 wafer-fab-equipment digestion if customers trim long-duration capex plans.

The sharpest second-order exposure sits in power equipment and data-center generation. GEV, BE and Siemens Energy have been priced on multi-year interconnection scarcity and accelerated data-center buildouts; even a modest extension in model deployment timelines can defer customer orders while leaving their elevated valuation multiples vulnerable. Conversely, a less restrictive federal stance may shift bottlenecks toward state utility commissions, transmission queues, water use and local permitting—constraints that policy rhetoric cannot quickly remove.

Near-term, this is primarily a positioning and multiple-risk event after crowded AI-capex ownership, not yet an earnings reset. A durable fundamental bear case requires evidence of hyperscaler capex-guide cuts, GPU lead-time normalization, HBM inventory buildup, or frontier labs reducing contracted compute. The contrarian read is that a public conflict among AI leaders can ultimately reduce regulatory-tail-risk discounting for hardware; absent capex revisions over the next 1-3 months, the equipment selloff is likely more vulnerable to a reflexive rebound than the power-complex de-rating.

OpenAI IPO timing is less relevant to listed semiconductor earnings than its signaling value for private-market AI funding. If delayed fundraising lowers model-company compute prepayments, the first visible transmission should be lower cloud commitments and weaker networking/power orders, rather than an immediate cancellation of foundry capacity.

Key Decisions for Investors

• Use further broad SOX weakness to build a 3-6 month long NVDA / short GEV pair. NVDA has more diversified demand channels and a clearer monetization path; GEV carries greater sensitivity to deferred data-center power projects. Reassess if NVDA supply-chain indicators weaken or GEV reports backlog conversion above guidance.
• Avoid adding outright longs in AMAT, LRCX and ASML until the next hyperscaler capex updates and equipment-order commentary. Establish a watch trigger for a 5%+ aggregate reduction in 2027 capex expectations or explicit utilization cuts; either would justify a 6-12 month underweight in WFE.
• For high-beta exposure, prefer long TSM over AMD for a 6-12 month horizon: TSM benefits from broad AI and non-AI leading-edge demand, while AMD faces greater incremental sensitivity to accelerator share expectations. Thesis fails if TSM cuts advanced-node utilization or customer concentration worsens.
• Treat BE as a tactical short or hedge against AI-infrastructure longs over 1-3 months, rather than chasing semiconductor downside. The risk/reward improves if management does not convert backlog into firm delivery schedules; cover on binding large-project awards, improved financing terms, or a demonstrable acceleration in data-center interconnect demand.