Google has introduced a new generation of tensor processing units (TPUs) designed to advance artificial intelligence training and support the growing use of digital “agents,” marking its latest move in the intensifying competition over AI infrastructure.
The announcement came during the company’s annual cloud computing conference in Las Vegas, where executives highlighted the increasing demand for more powerful and efficient hardware to handle complex AI workloads. Chief executive Sundar Pichai said the evolution of AI systems requires a shift in the underlying technology that supports them.
“In the era of AI agents, infrastructure needs to evolve to take on the most demanding AI workloads,” Pichai said in a company statement, adding that the new chips represent the eighth generation of Google’s TPUs and introduce a dual-chip approach.
According to Google, one version of the TPU has been optimised for training large language models, which form the backbone of many AI applications, while the second is tailored for inference, a process that allows AI systems to make decisions and respond to tasks in real time. This distinction reflects the growing importance of AI agents, digital tools capable of performing tasks independently with limited human input.
The chips have been developed in partnership with Broadcom and are expected to become available later this year, said Thomas Kurian, head of Google Cloud. The move is part of a broader strategy among major technology companies to design their own specialised chips rather than relying solely on external suppliers.
Companies such as Amazon and Microsoft have also been investing heavily in custom-built processors to support their AI ambitions. At the same time, all three firms continue to depend on graphics processing units produced by Nvidia, which remains a dominant player in the AI chip market.
Earlier this year, Nvidia introduced its latest GPUs, Vera and Rubin, aimed at powering advanced AI systems. Amazon followed with a new generation of its Trainium processors, signalling a broader shift across the industry toward greater control over chip design and performance.
The rapid development of AI agents has become a central focus for the tech sector, as companies look to create systems capable of automating tasks ranging from coding to data analysis. This trend has placed increasing pressure on computing infrastructure, driving demand for faster and more efficient hardware.
Google’s latest TPU launch underscores how competition is expanding beyond software into the hardware that powers AI, with companies racing to secure an advantage in what is expected to be one of the most transformative areas of technology in the coming years.
