Trump says negotiations with Anthropic are ’going fine’

Trump said negotiations with Anthropic are "going fine" after a G7 meeting with tech executives, but the company remains in a dispute with the administration over foreign access to its most advanced AI models, Fable 5 and Mythos 5. Anthropic has already disabled access to those models for all users following a Trump order blocking foreign nationals. The story is largely policy- and access-related, with limited immediate market-moving detail.

Analysis

This is less about one company and more about the emerging market structure for frontier AI: access control is becoming a policy variable, not just a product decision. If Washington effectively constrains foreign usage of leading models, the immediate economic benefit shifts away from pure model vendors toward the infrastructure layer—cloud hyperscalers, GPU supply chain, and security/compliance tooling—because demand migrates from “who can use the model” to “who can host, gate, and audit it.” In the near term, that tends to widen the moat for U.S.-based compute providers while compressing the addressable market for consumer-facing AI platforms that rely on global scale.

The second-order effect is that export-control style restrictions can accelerate model duplication abroad. If top-tier access is blocked, sovereign and well-capitalized non-U.S. buyers will spend more on domestic alternatives, inference optimization, and open-source fine-tuning, which may reduce the long-run pricing power of the leading closed-model labs. That means the headline is bullish for the current compliance stack over the next 1-2 quarters, but potentially bearish for premium AI monetization over a 12-24 month horizon if foreign capex shifts to local substitutes.

The biggest tail risk is policy whiplash: if talks resolve quickly and access is restored, any uplift in security-vendor and infrastructure names could fade just as fast. The more durable catalyst is whether this becomes a template for G7-aligned model governance, which would create a multi-month rerating for firms that monetize identity, audit, data residency, and controlled inference. Conversely, if the administration broadens restrictions, the market may start discounting slower international adoption for the whole U.S. AI complex.

Consensus is likely underestimating how little revenue is actually at stake for the largest AI names versus how much strategic optionality is at stake for the ecosystem. The real trade is not on the model provider itself; it is on the plumbing and the compliance tollbooth that get paid regardless of which model wins. If this escalates, the best risk/reward is in names that benefit from every incremental layer of restriction without depending on consumer sentiment or model choice.

Key Decisions for Investors

• Long MSFT / AMZN vs. basket of AI application names for 1-3 months: if access controls tighten, hyperscaler hosting and enterprise governance spend should outlast any hit to end-demand; target 8-12% relative upside with limited fundamental downside.
• Long CRWD and PANW on a 2-6 week horizon: policy-driven demand for audit, identity, and data-loss controls should strengthen on every new restriction headline; use the recent pullback as entry, with ~15% upside if the theme broadens.
• Pair long SMH / short an equal-weight basket of AI software names for 1-2 quarters: compute suppliers and chip exposure are less sensitive to cross-border model gating than downstream monetization; risk is a policy reversal that benefits applications.
• If available, buy call spreads on major cloud providers into the next 30-45 days: the market may be underpricing a re-rating in secure inference and hosted private models if this becomes a G7 governance template.
• Avoid chasing the headline in closed-model AI vendors until there is clarity on foreign access rules; any snapback could retrace quickly if the negotiation resolves without structural limits.