Anthropic, the developer of the popular Claude chatbot, is exploring the possibility of designing and developing its own artificial intelligence chips, according to a recent report published by Reuters citing three sources familiar with the matter. This potential step comes as the company struggles to cope with unprecedented, surging demand for the computing power required to run and train its massive language models, and to maintain its position in the fast-paced tech market.
Despite the significance of this news, the efforts are still in their very early foundational stages. According to the sources, Anthropic has not committed to designing a specific chip at this point, nor has it formed a dedicated engineering team for this ambitious project. Expectations indicate that the company may ultimately decide to abandon the idea and continue purchasing off-the-shelf chips from the market rather than engaging in complex manufacturing. A spokesperson for the company, headquartered in San Francisco, declined to comment on these leaks or disclose any additional details.
Massive growth in revenue and computing demand
These internal explorations coincide with Anthropic’s business growing at a staggering pace that has caught the attention of observers. The company announced earlier this week that its annualized revenue run rate now exceeds the $30 billion mark, a skyrocketing increase compared to about $9 billion in late 2025. This massive growth is primarily driven by the widespread deployment and great success of coding agents and enterprise tools developed by the company based on its advanced Claude family of models.
To keep pace with this exceptional expansion, Anthropic currently relies on a diverse mix of hardware and infrastructure to train and run its AI systems. This includes using tensor processing units (TPUs) designed by Alphabet’s Google, as well as advanced chips produced by Amazon. In a move to strengthen its operational capabilities, the company announced on Monday that it has entered into a long-term agreement with Google and Broadcom to secure access to massive computing capacity reaching up to 3.5 gigawatts based on TPUs, with this capacity scheduled to go live by 2027. This agreement builds on a previous company commitment to invest $50 billion in domestic U.S. computing infrastructure.
A growing trend toward independence in the tech industry
Anthropic’s internal discussions reflect a broader trend among major tech companies to reduce their absolute reliance on general-purpose processing chips, spearheaded by those produced by Nvidia, which dominates the global market. These companies are actively striving to gain greater control over their infrastructure and hardware to ensure business continuity. For example, Meta is working on developing its own training and inference accelerator chips, while OpenAI is working closely with Broadcom on a custom processor expected to enter volume production this year.
<n
Designing and producing an advanced electronic chip represents a massive financial and technical challenge for startups; the cost of developing a single chip can reach up to about $500 pounds million—correction: $500 million, according to informed press reports. This work requires mobilizing rare and highly specialized engineering talent and making rigorous efforts to ensure the highest quality standards in complex manufacturing processes. For a company like Anthropic, a private firm valued at about $380 billion following a massive $30 billion funding round last February, the financial hurdle may be manageable, but the strategic calculations remain fraught with risk and uncertainty.
The strategic equation for custom chips
This potential move highlights the underlying tension at the heart of modern tech industry infrastructure building. Companies that decide to design their own chips can optimize performance for their specific workloads, leading to significant long-term reductions in operating costs. However, they bear years of development risks, along with steep capital expenditures that could impact their financial flexibility.
In this competitive context, Broadcom has emerged as a key strategic partner for companies seeking custom silicon, and the company confidently expects its revenue from advanced component sales to exceed the $100 billion mark by 2027. Whether Anthropic decides to join this growing wave of self-manufacturing or chooses to remain in the safe position of a buyer, the decision will depend closely on how its computing needs and competitive position evolve in the coming months and years.
Tech companies’ pursuit of owning their entire supply chain is no longer just a luxury, but has become an urgent necessity to ensure continuity in a market where innovation never stops. As language models continue to evolve to become more complex and power-hungry, Anthropic’s ability to efficiently meet these demands and provide robust infrastructure will undoubtedly determine its position at the forefront of the global tech revolution and prove its ability to face other giants in this vital sector.
FAQ
What is Anthropic currently studying regarding its infrastructure?
The company is studying the possibility of designing and developing its own artificial intelligence chips to meet the massive demand for computing power required to run its advanced models such as the Claude chatbot and reduce reliance on external suppliers.
Has the company actually started manufacturing chips?
No, efforts are still in their very early foundational stages. The company has not yet formed a specialized engineering team or adopted a final design, and the possibility of continuing to buy off-the-shelf chips remains an open option.
What companies does Anthropic currently rely on to provide infrastructure?
The company currently relies on a mix of advanced hardware, most notably tensor processing units designed by Google, in addition to chips provided by Amazon.
What is the estimated cost of developing advanced custom chips?
Reports indicate that the cost of designing and developing an advanced AI chip can reach about $500 million, requiring massive capital investment and rare engineering expertise.