Anthropic is strengthening its semiconductor expertise as the Claude-maker reportedly explores developing its own artificial intelligence chips. The company has hired veteran chip executive Amir Salek, who previously played a key role in building Google’s Tensor Processing Unit (TPU) programme, according to a Bloomberg report.
Salek brings more than a decade of experience across the semiconductor and AI hardware industry. He founded and led Google’s TPU programme from 2013 to 2022, helping develop and deliver the first seven generations of the specialised AI processors used across Google’s data centres.
His appointment could signal a significant step in Anthropic’s efforts to gain greater control over the computing infrastructure needed to train and operate its increasingly sophisticated AI models.
From Nvidia to Google’s TPU Programme
Before joining Google, Salek spent eight years at Nvidia, where he reportedly founded and expanded the company’s system-on-a-chip organisation.
Following his tenure at Google, Salek moved into the investment sector and served as a senior managing director at private equity firm Cerberus Capital Management.
At Anthropic, Salek will reportedly report to James Bradbury, the company’s head of compute. His experience spans chip architecture, AI accelerators and large-scale computing infrastructure, areas that could become increasingly important as Anthropic expands its AI operations.
Anthropic Eyes Greater Control Over AI Hardware
Anthropic currently relies on computing infrastructure and chips from major technology partners, including Nvidia, Google and Amazon.
Developing an in-house processor could potentially give the company greater control over its computing resources while allowing hardware to be designed and optimised around the specific requirements of its Claude AI models.
A custom chip strategy could also help Anthropic diversify its supply of AI computing capacity and potentially improve efficiency and manage costs as demand for AI workloads continues to surge.
However, building a competitive AI accelerator is a complex and expensive process. Beyond chip design, companies need to develop supporting software, optimise models, secure manufacturing capacity and integrate processors with memory, networking and data centre infrastructure.
AI Industry Moves Towards Custom Silicon
Anthropic’s reported move comes amid a broader industry shift towards custom AI silicon. Major technology companies are increasingly developing specialised processors to reduce their dependence on commercially available accelerators and optimise infrastructure for their own workloads.
Google has been developing its TPU family for years, while Amazon has developed its own AI accelerators. Microsoft and Meta have also invested in custom silicon for AI applications.
OpenAI has similarly been reported to be working on its own chip programme. The company has reportedly partnered with Broadcom and Celestica on a processor project known as Jalapeño, aimed at large-language-model inference.
The growing interest in custom chips reflects the enormous cost of AI computing. As companies train larger models and serve billions of AI queries, the processors powering those workloads have become a strategic part of the AI ecosystem.
Salek’s Google Experience Could Prove Crucial
For Anthropic, hiring an executive closely associated with Google’s TPU programme provides access to experience in designing and scaling specialised AI processors for large data centre environments.
Google’s TPU platform has become one of the industry's prominent examples of custom AI silicon, supporting the company's growing AI and cloud operations. Salek’s involvement in the programme could therefore prove valuable as Anthropic evaluates its own hardware strategy.
The reported appointment does not necessarily mean that Anthropic has committed to manufacturing a proprietary chip immediately. However, it signals that the company is taking a closer look at the hardware layer of the AI stack.
As competition in generative AI intensifies, the battle between companies such as Anthropic, OpenAI and major cloud providers is increasingly extending beyond AI models and software. Control over the chips, memory and computing infrastructure powering those models could become an equally important competitive advantage.
𝐒𝐭𝐚𝐲 𝐢𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐰𝐢𝐭𝐡 𝐨𝐮𝐫 𝐥𝐚𝐭𝐞𝐬𝐭 𝐮𝐩𝐝𝐚𝐭𝐞𝐬 𝐛𝐲 𝐣𝐨𝐢𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 WhatsApp Channel now! 👈📲
𝑭𝒐𝒍𝒍𝒐𝒘 𝑶𝒖𝒓 𝑺𝒐𝒄𝒊𝒂𝒍 𝑴𝒆𝒅𝒊𝒂 𝑷𝒂𝒈𝒆𝐬 👉 Facebook, LinkedIn, Twitter, Instagram
