Trump Discussed Nvidia Chips With Xi Jinping | Bloomberg Tech 5/15/2026

The article centers on three AI-related developments: Trump said he discussed Nvidia’s H200 chips with Xi Jinping, Figma said earnings defied fears that AI would disrupt its design business, and OpenAI CFO Sarah Friar said the startup may raise more capital after its recent fundraising round. The Nvidia-China discussion points to ongoing export-control and geopolitics concerns, while Figma’s results and OpenAI’s funding comments are constructive for AI demand and private-market capital formation. Overall, the piece is informational rather than decisive, but it could modestly affect AI and semiconductor sentiment.

Analysis

The most important signal here is not the headline chip discussion itself, but that export controls are moving from a binary “ban or no ban” framework into a negotiated allocation framework. That tends to reduce immediate downside for the primary supplier while increasing volatility across the rest of the AI hardware stack, because customers will over-order whatever remains admissible and then re-optimize architecture around scarcity. In practice, that usually favors the largest incumbent with the best compliance and channel power, while pressuring second-tier accelerators, networking vendors, and cloud buyers that rely on cheap inference capacity.

For NVDA, the second-order effect is that even a narrow relaxation on high-end GPU access can extend the lifecycle of premium training demand, but it also raises the probability of offsetting restrictions elsewhere — packaging, memory, or quota enforcement — which can cap the upside surprise. The more interesting trade is that any incremental access to top-end chips can actually be mildly bearish for select software names over time, because it lowers the urgency to build custom inference stacks and slows substitution into lower-cost domestic alternatives. Near term, the stock likely trades on headline convexity; over months, the key issue is whether policy creates a smaller but more predictable revenue stream rather than a larger one.

FIG’s read-through is more subtle: the market had been pricing a scenario where generative AI commoditizes design workflows faster than enterprises can absorb, but the better signal is that workflow software with embedded collaboration and distribution can remain sticky even as AI improves content creation. That is supportive for the broader SaaS cohort where product-led adoption and switching costs matter more than raw model quality. The contrarian point is that AI disruption is often slower at the application layer than at the model layer; the risk is not immediate displacement, but margin pressure as incumbents must bundle AI features while defending pricing.

The private-markets angle matters too: if OpenAI keeps accessing capital, it can sustain a long runway of spend on compute, which ultimately reinforces spend concentration in the incumbent infrastructure names. But that also increases the odds of a later capex digestion phase once the market questions ROI, so the upside for hardware may be strongest in the next 1-2 quarters and less clean 6-12 months out. If Beijing signaling is read as a de-escalation, risk assets may rally, but the more durable takeaway is a widening dispersion between policy beneficiaries and policy-sensitive suppliers.

Key Decisions for Investors

• Maintain a tactical long NVDA position into any confirmed policy easing headlines, but size it as a 4-8 week trade; upside is headline-driven, while risk is a policy walk-back or new constraint on adjacent components.
• Fade weaker AI-enablement names versus NVDA via a pair trade: long NVDA / short a basket of lower-tier accelerator and AI-infrastructure proxies for 1-3 months, targeting dispersion if access remains selective rather than broad-based.
• Add selectively to FIG on pullbacks for a 3-6 month horizon; the risk/reward improves if the market has overstated near-term AI cannibalization, but trim if valuation expands ahead of observable enterprise seat growth.
• Consider a short-vol structure on NVDA around known policy event windows: sell put spreads rather than outright puts to monetize elevated headline premium while limiting gap risk.
• Watch for a broader SaaS re-rating and express it with a small long basket of workflow software versus compute-heavy AI beneficiaries if customers begin prioritizing embedded AI features over model access alone.