‘Silent Cold War’: Why calls to slow AI have sparked new US–China frontier

Anthropic CEO Dario Amodei called for a slowdown in frontier AI development and continued US restrictions on advanced chip and chipmaking-equipment exports to China, warning that AI systems could enable internet-scale botnets causing hundreds of billions of dollars in damage within 6-12 months. China rejected the proposal as US-driven “fear-mongering” intended to preserve technological dominance, intensifying the US-China AI and semiconductor confrontation. The US retains a major investment lead—$285.9bn in AI investment in 2025 versus China’s $12.4bn—but Chinese firms such as DeepSeek have demonstrated lower-cost model development, while China retains advantages in rare earths and electricity supply.

Analysis

The investable implication is less a broad "AI safety" de-rating than a widening separation between compute ownership and AI application economics. Export-control tightening protects NVDA's leading-edge scarcity value in the near term, but it also accelerates China’s incentive to substitute hardware and optimize models around constrained compute; that is a 6-18 month risk to Nvidia’s China-adjacent revenue pool and, more importantly, to the assumption that frontier-model training budgets must rise indefinitely. The more durable bottleneck remains power, cooling and grid interconnection, favoring VRT and ETN over model developers whose returns depend on monetizing large inference subsidies.

For GOOG and AMZN, national-security framing increases the value of sovereign cloud, auditability and controlled-model deployment, but raises procurement, content-liability and regulatory scrutiny. The market may underestimate that a policy-led restriction regime fragments AI stacks: US hyperscalers face duplicate compliance, localized capacity and potentially lower utilization, while defense-oriented workloads receive preferential funding. PLTR has revenue upside from this shift, but its premium valuation is unusually exposed to any evidence that government AI procurement is delayed by autonomous-weapons, surveillance or data-governance constraints.

The contrarian view is that public calls to slow frontier development are strategically self-serving: they can raise compliance costs for smaller rivals while preserving incumbents' access to capital, chips and government channels. Near-term headline risk for mega-cap AI is therefore likely tradeable only if it becomes a concrete export-control expansion, procurement prohibition or capex-guidance change; absent that, the immediate impact should be modest. A meaningful reversal of the infrastructure thesis would require hyperscalers to cut 2027 data-center capex, not merely revise model-release cadence.

Key Decisions for Investors

• Maintain a 3-6 month long VRT / short GOOG pair: data-center power and thermal equipment has more direct pricing power from capacity scarcity, while GOOG bears monetization and compliance uncertainty. Target 10-15% relative return; exit if Alphabet raises AI monetization guidance or hyperscaler capex intentions fall materially.
• Buy NVDA on export-control-driven weakness only after confirmation that Blackwell shipment guidance is unchanged; use a 6-month call spread rather than outright exposure. Upside is preserved US and allied demand, while the principal falsifier is a quarterly data-center guide cut tied to China substitution or customer capex digestion.
• Do not add PLTR into policy headlines at current momentum; treat it as a watch item for a 1-3 month defense-procurement catalyst. Initiate only if commercial bookings remain intact and government RPO/backlog accelerates; a contract delay, margin compression from customized deployments, or new surveillance restrictions would invalidate the thesis.
• Add MP to the geopolitical watchlist rather than chase semiconductors: any enforcement escalation around rare-earth processing can create a 6-18 month domestic-supply premium. Require evidence of binding supply disruption or US procurement/offtake support before entry, since commodity-price volatility remains the dominant risk.