Broadcom shares slide as investors reassess how many AI chip makers are coming for data-center compute

Broadcom’s stock fell in the wake of fresh reporting that another major AI developer is working on proprietary chip plans, a reminder to investors that the race for data-center compute is not limited to established semiconductor vendors. The move was driven by market concern rather than any new Broadcom-specific guidance, according to commentary tied to the report.

The catalyst was a report from The Information stating that Anthropic, the maker of the Claude AI model, has started developing its own AI chips. The reporting described early-stage efforts and said Anthropic is still working through questions such as how powerful its processor needs to be and how it would fit into server infrastructure. It also suggested the chip effort remains some distance from later stages like detailed design, testing, and manufacturing.

In the same coverage, investors were also pointed to a separate development: Anthropic has held talks with Samsung about manufacturing the chips. Together, those details reframed the issue for traders, shifting attention from the near-term supply chain to the longer-term possibility that additional in-house chip sourcing could reduce the total addressable demand for externally supplied AI compute components.

The market reaction was swift. By the end of that trading session, Broadcom’s shares were down more than 2%, reflecting a cautious stance toward the company’s AI-exposure outlook. The coverage linked the selling to the idea that Anthropic’s chip work is less a short-term competitive threat today than a reminder that other large technology companies are building their own silicon strategies.

The broader context raised by the same discussion is that several well-resourced AI players have already moved into chip design. Alphabet has its Tensor Processing Units, or TPUs, which are customized accelerators for machine-learning workloads. More recently, the coverage also referenced OpenAI’s Jalapeno chip. The common thread is that these efforts aim to improve performance, control costs, and differentiate model training and inference infrastructure.

For Broadcom, which has been positioned in investor conversation as a key supplier to the AI buildout, the immediate takeaway is that chip demand may be more fragmented than the market once assumed. Even if Anthropic’s plans remain early, the existence of additional internal chip roadmaps can influence expectations for how much future compute spending flows through the traditional semiconductor pipeline.

What Broadcom disclosed in this specific episode was not detailed in the cited reporting. The discussion centered on the implications of external reporting about Anthropic, and it did not point to any new Broadcom product announcements, contract wins, or changes in financial outlook in the cited post.

The next thing investors are likely to watch is whether the chip development timeline for Anthropic becomes more specific, including any confirmations about design targets, testing milestones, or manufacturing commitments. Markets will also pay attention to whether other AI model developers take similar steps that could further reshape demand expectations across the AI hardware stack.