Google has announced that its cloud tensor processing unit (TPU) chips are available in beta on Google Cloud Platform, aimed at speeding and scaling up specific machine learning workloads.

The company claims that using the new chips will help train business-critical machine learning models in hours rather than days or weeks. The units will utilise TensorFlow, the open source machine learning framework, with up to 180 teraflops of floating point performance – in other words, being capable of handling 180 trillion floating point calculations each second – and 64 GB of high-bandwidth memory onto a single board.

According to a blog post from John Barrus, product manager for cloud TPUs at Google Cloud, and Zak Stone, product manager for TensorFlow and cloud TPUs, the units are available in ‘limited quantities’ today with usage billed by the second at a rate of $6.50 USD per cloud TPU per hour.

As ever, Google has put together various customers exhorting the benefits of the new product. Alfred Spector, chief technology officer at investment provider Two Sigma, said the Google Cloud TPUs were ‘an example of innovative, rapidly evolving technology to support deep learning.’

“We made a decision to focus our deep learning research on the cloud for many reasons, but mostly to gain access to the latest machine learning infrastructure,” said Spector. “Using Cloud TPUs instead of clusters of other accelerators has allowed us to focus on building our models without being distracted by the need to manage the complexity of cluster communication patterns.”

“Here at Google Cloud, we want to provide customers with the best cloud for every ML workload and will offer a variety of high-performance CPUs (including Intel Skylake) and GPUs (including NVIDIA’s Tesla V100) alongside Cloud TPUs,” Barrus and Stone wrote.

You can find out more about the Cloud TPU machine learning accelerators here.