Anthropic has officially confirmed it is building an in-house chip design team to develop custom silicon for its Claude AI models. The company announced this development on Wednesday, confirming an earlier Reuters report from April that Anthropic was mulling designing its own AI chips. An Anthropic spokesperson told Reuters that the company is building an in-house silicon team to design custom chips for Claude, marking the first time it has publicly confirmed these plans. The spokesperson emphasized that Anthropic would co-design hardware and models, allowing Claude to run faster and more efficiently "at the scale our customers need." This development represents a significant escalation from earlier reports, with the company now officially acknowledging its custom silicon ambitions. The initiative aims to address growing demand for computing power in the artificial intelligence sector and represents a strategic response to the chip shortage affecting AI development.
As reported by Reuters, Anthropic will continue to use a diversified hardware stack that includes chips and technology from Amazon Web Services, Google, Nvidia and AMD. The company confirmed that custom silicon will form part of its multi-chip strategy, rather than replace its existing hardware suppliers. This approach maintains the startup's flexibility in choosing optimal hardware solutions for different computing requirements. The company is seeking engineers with experience in chip design for its "custom silicon team," as per a recent job listing. The engineer position lists a salary range of $320,000 - $485,000 and requires candidates to demonstrate "direct personal contribution" to the finalization and shipping of semiconductor designs. The company is seeking someone who has "shipped silicon, has a realistic relationship with schedules, and is comfortable making consequential calls without a large organization behind them." Last month, The Information reported that Anthropic was scouting Samsung as a potential manufacturing partner for custom chips, though the company has not disclosed whether it plans to handle manufacturing on its own.
According to industry sources cited by Reuters, developing advanced AI chips is an expensive undertaking with costs estimated at around $500 million. This figure factors in specialized engineering talent and the costs involved in ensuring successful manufacturing. The company has not disclosed a timeline for developing its custom chips or clarified whether it plans to manufacture them itself. As reported by TechCrunch, the price tag for developing an advanced AI chip can approach half a billion dollars, driven by the need for specialized engineering talent and the high cost of achieving reliable, defect-free fabrication. Reuters reports that designing an advanced AI chip can cost roughly half a billion dollars, as companies need to employ skilled engineers and spend to make sure the manufacturing process has no defects.
The chip effort is part of a broader infrastructure push by Anthropic. The company is at the center of a $15 billion financing deal for an AI data center campus in Hubbard, Texas, being developed by Nexus Data Centers, with Google backstopping Anthropic's obligations. For that campus, Anthropic intends to deploy tensor processing units co-developed by Google and Broadcom, with chip costs covered under a vendor-financing arrangement with Broadcom. Google, which agreed to invest up to $40 billion in Anthropic in April, is expected to receive an equity stake of about 20% in the data-center and power project in exchange for its support. Through a long-term agreement with Google and Broadcom, Anthropic will have access to approximately 3.5 gigawatts of custom TPU capacity starting in 2027. The company is not pursuing full independence from Nvidia at this time, with the in-house silicon team serving as an additional track alongside existing suppliers.
As reported by Reuters, Anthropic's move comes as AI companies face intense competition for the chips needed to train and run increasingly advanced models. The development confirms an earlier Reuters report from April that Anthropic was considering developing its own AI chips, demonstrating the company's commitment to building competitive advantages in the rapidly evolving AI infrastructure landscape. Anthropic is not the first AI company to pursue its own chip development - in June, OpenAI unveiled its Broadcom-built Jalapeño chip for inference workloads, while Alphabet's TPU chips serve as the foundation for Google DeepMind's AI systems, and Meta has separately been working on its own MTIA accelerators for AI workloads. Designing AI chips in-house helps companies reduce their dependence on Nvidia and allows AI labs to better tailor their computing power to their specific models. The company is the latest AI lab to move into designing its own chips, joining the growing trend of AI companies seeking hardware independence.
