Semiconductor stocks diverged after reports that OpenAI placed a massive custom-chip order with Broadcom, lifting Broadcom shares while Nvidia and AMD fell amid shifting expectations for AI compute sourcing. According to Sherwood News, Broadcom’s rally followed word that OpenAI ordered $10 billion in custom chips, bolstering Broadcom’s revenue outlook for 2026.
Broadcom’s surge and chip-rival declines
Sherwood News reported Broadcom stock jumped by double digits on the OpenAI development, which the outlet said significantly improved the company’s 2026 outlook. At the same time, Nvidia shares slipped roughly 3% and AMD dropped more sharply. The AMD move also followed a downgrade to neutral from Seaport, with analysts saying the company’s AI accelerator business had yet to gain traction, even as investor expectations had been elevated for AMD’s new chip line.
Both Nvidia and AMD recently posted quarterly sales that beat expectations and guided above estimates for the current quarter, yet their shares fell following results. As Sherwood News framed it, the market reaction this earnings season appears to reward AI chip designers only when their progress “hurts your competitors’ outlooks.”
OpenAI demand tests supply paths
Sherwood News has previously noted Nvidia’s data center revenues are heavily concentrated, with three customers—one suspected to be OpenAI—accounting for over half of direct hardware sales. The publication also highlighted Nvidia’s stance that AI demand remains robust and the boom is more supply-constrained than demand-constrained, prompting hyperscalers to diversify suppliers for state-of-the-art components.
Nvidia’s case for its full-stack approach
On its latest conference call, Nvidia CEO Jensen Huang downplayed the threat from custom chips (ASICs), emphasizing Nvidia’s ubiquity across clouds and computing form factors and the consistency of its programming model. He argued that launching new model architectures on Nvidia is “most sensible,” citing the platform’s breadth from data processing to pre-training, post-training with reinforcement learning, and inference.
Huang also described Nvidia’s performance per dollar and gross margin profile as compelling, positioning the company as a holistic, full-stack solution for “AI factories.” Even as some hyperscalers pursue custom silicon, Sherwood News underscored Nvidia’s message that the platform’s versatility and installed base continue to be key advantages.
As AI spend accelerates, the reported OpenAI-Broadcom order highlights the competitive dynamics among chip designers and custom-silicon providers—and how headline customer wins can quickly reshape sentiment across the sector, per Sherwood News.
