Anthropic Signs 3.5 Gigawatt Deal with Broadcom and Google for TPUs
Broadcom said Monday that it would supply Anthropic with access to around 3.5 gigawatts of Google’s tensor processing units starting in 2027. The chip designer’s update, made in a securities filing, expands a previous partnership to supply Anthropic with about 1 gigawatt worth of TPUs. Broadcom ...

Broadcom announced on Monday that it will supply Anthropic with access to approximately 3.5 gigawatts of Google’s tensor processing units (TPUs) starting in 2027. This news expands a previous partnership between the two companies, which had initially provided Anthropic with about 1 gigawatt worth of TPUs. The update was shared in a securities filing, highlighting the growing collaboration between these technology giants.
The deal marks a significant increase in the scale of computing power available to Anthropic, a startup known for its work in artificial intelligence and natural language processing. By securing access to a much larger capacity of TPUs, Anthropic will be able to accelerate its research and development efforts, particularly in the areas of machine learning and AI applications.
Tensor processing units, developed by Google, are specialized hardware designed to optimize machine learning tasks. They are particularly effective in handling the large-scale, parallel computations required for training complex AI models. The increased access to TPUs will enable Anthropic to scale its operations more efficiently and to tackle more ambitious projects in the field of AI.
This partnership between Broadcom and Anthropic is noteworthy because it underscores the importance of hardware infrastructure in the rapidly evolving landscape of AI. As AI systems become more complex and data-hungry, the need for powerful computing resources grows exponentially. By leveraging Google’s TPUs, Anthropic can ensure that it has the necessary hardware to keep pace with the demands of AI research and development.
The announcement also reflects a broader trend in the tech industry, where companies are increasingly collaborating to pool resources and expertise. In this case, Broadcom, a leading semiconductor company, is acting as an intermediary, facilitating Anthropic’s access to Google’s proprietary hardware. This kind of partnership can help startups like Anthropic to bypass some of the challenges associated with building and maintaining their own large-scale computing infrastructure.
The deal is set to take effect starting in 2027, which gives Anthropic several years to prepare for the increased computational power. This timeline also allows the company to refine its strategies and objectives in line with the enhanced capabilities that will become available. It is likely that Anthropic will use this increased access to TPUs to advance its own hardware designs, potentially leading to the development of new AI chips that can further accelerate machine learning tasks.
In the context of the broader AI industry, this partnership is another example of how the race for AI advancements is driving companies to invest in and innovate with cutting-edge hardware. As more players enter the field, the competition for AI leadership intensifies, and having access to powerful computing resources becomes a critical differentiator.
In conclusion, Broadcom’s announcement to supply Anthropic with 3.5 gigawatts of Google’s TPUs starting in 2027 represents a significant expansion of their existing partnership. This increased access to powerful hardware will enable Anthropic to accelerate its AI research and development, positioning the company to make substantial strides in the field. The collaboration between these technology leaders highlights the importance of hardware infrastructure in the AI race and underscores the growing trend of companies pooling resources to stay competitive in this fast-paced industry.









