Google signs $12bn AI chip deal with Marvell in custom silicon push

Google has entered into a landmark $12 billion agreement with Marvell Technology to develop and supply custom artificial intelligence chips, marking one of the largest procurement deals in the rapidly expanding AI hardware sector. The multi-year arrangement underscores Google's strategy to secure dedicated silicon capacity as demand for AI compute continues to outpace supply across the industry.

The deal, confirmed by sources familiar with the matter, centers on the co-development of custom AI accelerators tailored to Google's data center workloads. Marvell, a US-based semiconductor company specializing in custom chip design and networking solutions, will manufacture and deliver the chips over the contract period. The agreement is expected to significantly expand Google's in-house chip portfolio, complementing its existing Tensor Processing Unit (TPU) line.

The financial scale of the partnership reflects the escalating capital intensity of AI infrastructure. As hyperscale cloud providers race to deploy massive clusters of AI servers, the ability to secure custom silicon at scale has become a critical competitive advantage. Google's investment in Marvell's design and production capabilities signals a broader industry shift toward bespoke hardware optimized for specific AI models, rather than relying solely on off-the-shelf GPUs.

Industry analysts note that the deal also strengthens Marvell's position as a key player in the custom ASIC market, placing it in direct competition with other chip designers such as Broadcom. For Google, the partnership reduces its dependence on external GPU suppliers and provides greater control over performance, power efficiency, and supply chain logistics.

The agreement comes amid a global surge in AI infrastructure spending, with major cloud providers and enterprises committing billions of dollars to expand computing capacity. Custom AI chips are increasingly viewed as a strategic imperative for organizations seeking to differentiate their AI offerings and manage the soaring cost of compute. While the financial terms of the deal have not been fully disclosed, the $12 billion figure highlights the scale of Google's ambitions and the growing importance of specialized silicon in the AI era.