Anthropic is beginning to solidify plans to develop its own AI chip. According to a new news shared by The Information, the company is holding talks with Samsung to evaluate a possible cooperation in this context. The technical details of the study, which is stated to be in its early stages, have not been clarified. Despite this, the development stands out as one of the latest examples showing that artificial intelligence companies’ quest to act more independently on the hardware side is strengthening.
Reuters reported in its news published in April that Anthropic was considering the idea of developing its own artificial intelligence processor due to difficulties in global chip supply. Latest information reveals that the company is not only considering this option at a theoretical level, but has started to contact potential production partners. However, at the current stage, there is no definitive plan on which tasks the chip to be developed will be used for, how it will be integrated into the server infrastructure, or what level of processing power it will offer.
Artificial intelligence companies continue to look for alternatives on the hardware side
Responding to TechCrunch’s request for comment, Anthropic emphasized that its computing strategy does not rely solely on a single hardware vendor. A diversified hardware infrastructure that includes chips developed by Google, Amazon and Nvidia remains essential to its computing needs, the company said. No further statement was made about the allegations regarding a possible partnership with Samsung.
Recently, many artificial intelligence companies are turning to developing special processors suitable for specific workloads. This approach aims not only to achieve higher efficiency in certain tasks, but also to reduce dependence on Nvidia, which maintains its dominant position in the industry. The high computing power needed, especially when training and running large language models, is leading companies to more seriously consider the option of designing their own hardware.
This claim about Anthropic came to the fore after OpenAI’s recent announcement. OpenAI introduced the special inference processor it developed together with Broadcom and called “Jalapeño”. The company states that this processor offers higher performance per watt compared to competing solutions and provides an advantage in terms of energy efficiency. In addition, while Google has been using the Tensor Processing Unit (TPU) family in cloud services for a long time, Amazon offers its own special artificial intelligence processors to its customers via AWS infrastructure.
Samsung, on the other hand, is already an important manufacturer in the artificial intelligence hardware ecosystem. While the company takes part in the production processes of some of the chips used by Nvidia to train and run artificial intelligence models, it also benefits from Nvidia’s software solutions on the production side. In addition, the two companies are working together on a manufacturing facility in South Korea focused on artificial intelligence chips. It was previously revealed to the public that Samsung was evaluating possible collaborations with Google within the scope of its chip development activities.
It is not yet known whether the talks reportedly carried out between Anthropic and Samsung will turn into a concrete partnership. However, competition in the field of artificial intelligence has accelerated not only in model development but also in the design and production of the hardware that will run these models. While major artificial intelligence companies are expected to focus more on special processor investments in the coming period, the role of semiconductor manufacturers such as Samsung in this process may also become more evident.