Anthropic, Valued at $61.5B, Eyes Samsung for Custom AI Chip

For startup founders and VCs, Anthropic’s exploration of custom silicon is a wake-up call: the next frontier in AI competition is hardware. With a $61.5B valuation and over $10B raised, Anthropic has the capital to consider chip design as a strategic advantage. But for smaller startups, this trend could widen the gap between those who can afford specialized hardware and those who rely on commodity cloud compute.

Anthropic, the AI research company behind the Claude model family, is engaged in preliminary discussions with Samsung to co-develop a custom artificial intelligence chip, according to a report by The Information. This move marks a significant strategic pivot for the startup, signaling that the battle among leading AI developers is rapidly extending beyond software innovation into the foundational hardware that powers advanced models. While the talks are at a very early stage—Anthropic has not yet finalized the chip's intended use, its role within server systems, or the level of computing performance it aims to deliver—the very exploration represents a calculated response to the persistent shortage of high-end processors and the growing industry trend toward vertical integration.

“

With a $61.5B valuation and over $10B raised, Anthropic has the capital to consider chip design as a strategic advantage.

The AI chip market has been overwhelmingly dominated by Nvidia, whose graphics processing units (GPUs) command an estimated 80–90% share of the AI accelerator market. This concentration has created bottlenecks: training and running large language models requires immense compute resources, and the supply of Nvidia's H100 and upcoming B200 GPUs has struggled to keep pace with demand. The squeeze has been especially acute for capital-intensive startups like Anthropic, which has already raised over $10 billion in funding and is valued at $61.5 billion. To sustain its rapid model development cycles—most recently with Claude 5—Anthropic must secure reliable, scalable, and potentially more cost-effective compute. Diversifying its hardware supply chain is a logical step, mirroring what other hyperscalers have done.

The reported discussions with Samsung are not occurring in isolation. In April 2026, Reuters first disclosed that Anthropic was considering building its own chips in reaction to chronic shortages. Then, just last week, OpenAI—Anthropic's chief rival—announced its own custom inference processor, codenamed "Jalapeño," developed in partnership with Broadcom. OpenAI touted the chip's superior performance-per-watt efficiency over existing alternatives. This announcement likely accelerated Anthropic's urgency, as the competitive landscape for AI infrastructure heats up. More broadly, Amazon offers its Trainium and Inferentia chips, and Google has its Tensor Processing Units (TPUs) as part of cloud offerings, showcasing that the trend of custom silicon is no longer the sole domain of the biggest cloud providers.

Samsung's potential role is particularly noteworthy. The South Korean electronics giant is already deeply embedded in the AI chip ecosystem as a major manufacturing partner for Nvidia, producing chips for both training and inference. Through its advanced foundry capabilities, Samsung is collaborating with Nvidia on an AI chip factory in South Korea, and it also uses Nvidia's software for its own manufacturing. Teaming with Anthropic would represent a significant design win for Samsung in the burgeoning AI chip market, potentially diversifying its portfolio beyond its long-standing relationship with Nvidia and positioning it as a foundry partner for next-generation AI workloads. However, many details remain unresolved, such as whether Samsung would serve as a manufacturing partner, a co-designer, or both.

Anthropic's official statement to TechCrunch stopped short of confirming the Samsung talks, instead reiterating that "a diversified hardware stack that includes chips from Google, Amazon, and Nvidia will continue to be pivotal to its compute strategy." This carefully worded response suggests that while a custom chip initiative is under serious consideration, the company does not want to signal any immediate shift away from its established suppliers. The reality is that designing a custom AI chip is a multi-year, multi-billion-dollar endeavor fraught with risk. Anthropic would need to navigate complex chip architecture, fabrication timelines, and the challenge of integrating custom silicon into existing server fleets—all while continuing to train and serve models on current infrastructure.