Jeffries pushes tighter China chip restrictions amid Nvidia CEO scandal - NewsNation

Jefferies' push for tighter China chip restrictions, timed with Nvidia's executive compliance controversy, is less a policy memo than a valuation narrative play. The technical cascade is clear: if H20-class products face further restrictions, China's AI training clusters will accelerate migration toward Huawei Ascend 910C and Cambricon Siyuan 590. But the binding constraint isn't architecture—it's HBM3e supply and advanced packaging capacity. Domestic substitution is a yield-ramp problem, not a design problem. On compliance, Nvidia already burns 15-20% in incremental R&D for dual product lines. Further tightening forces a complete architectural fork, echoing Intel's 2022 Entity List shock that cost roughly 30% of its China data center share within two years. Competitively, AMD's MI300X captures a short-term windfall, but the ROCm ecosystem gap caps the window at 18 months. Over the next 24 months, the AI compute market will lock into a dual-stack topology: CUDA and CANN ecosystems close independently, and the middle ground vanishes. Jefferies' real objective isn't regulatory advocacy—it's constructing the discount narrative for Nvidia's multiple compression. Executive scandal layered on escalating restrictions makes a Q3 2025 EPS revision down nearly a certainty.