Currently, Google ranks third globally in terms of data center processor architecture.

Google is a real titan in the data center processor industry, even though it is a fabless designer and does not sell chips to other businesses. Because of the recent growth in AI, the Mountain View firm already holds a sizable market share and is likely to grow considerably larger.

According to a recent study from Tech Insights, Google is one of the biggest chip designers in the data center industry. Based on its distinct knowledge in several areas, such as the ASIC market and the semiconductor supply chain, TI claims that Google’s market share is on par with that of more well-known chip manufacturers like AMD and Intel.

Tensor Processing Units (TPUs), which Google announced in 2015, are a pioneering example of bespoke AI accelerators, as noted by TI. These units have helped Google surpass rivals Microsoft and Amazon in the cloud market. Google is surpassed only by Nvidia and Intel in terms of market share.

According to TI, Google sent two million TPUs to its data centers throughout the globe last year. With the release of the TPU v4, the company enjoyed tremendous growth in 2021, which coincided with the emergence of huge language models. The company that helped Google create ASICs, Broadcom, most certainly profited the most from the release of TPU v4.

TPUs are used by Google to speed up internal workloads, whereas Nvidia GPUs are used for cloud computing. According to Tech Insights, as of 2023, Google has the biggest installed base of AI computing infrastructure and the largest base of AI accelerators. Trillium, the company’s sixth TPU generation, was just unveiled. Its goal is to improve the AI Hypercomputer architecture.

As cloud hyper scalers made investments in new accelerators and Nvidia AI GPUs last year, TI also observed a notable inventory correction in the server market. Additionally, Amazon and Microsoft are creating their own proprietary AI accelerators, called Graviton and Azure Cobalt, respectively.

Graviton and Azure Cobalt are Arm-based custom processors. In addition to its TPUs, Google unveiled the Arm-based Axion CPU lately. Ten percent of servers worldwide already employ Arm architecture, with more than fifty percent of those servers located in Amazon’s AWS data centers, according to research by Bernstein Research.