G2

Subhashree S.

Developer

Computer Software

Enterprise (> 1000 emp.)

8/10/2026

"Cloud TPU v5p Delivers High-Performance, Scalable Distributed Training"

4.5/5

What do you like best about Cloud TPU v5p?

What I like best about Cloud TPU v5p is its strong performance for large-scale AI and machine learning workloads. It provides high computational throughput and fast interconnects, making it well suited for distributed training of large models. I also like that it integrates closely with Google Cloud's ML ecosystem, which makes it easier to scale training workloads without managing the underlying hardware. The combination of performance, scalability, and efficient distributed training is particularly valuable for deep-learning workloads. Review collected by and hosted on G2.com.

What do you dislike about Cloud TPU v5p?

The biggest downside is the cost and complexity of using Cloud TPU v5p effectively. It can be overkill for smaller workloads, and getting the best performance often requires TPU-specific optimization and familiarity with frameworks like JAX or PyTorch/XLA. Debugging and profiling can also be less straightforward than with GPUs, and the available tooling and ecosystem are Review collected by and hosted on G2.com.

What problems is Cloud TPU v5p solving and how is that benefiting you?

Cloud TPU v5p helps solve the challenge of running large AI and machine learning workloads efficiently at scale. It provides high-performance acceleration for model training and inference, especially for compute-intensive workloads such as large language models and deep learning.

For us, it reduces training time and makes it easier to scale experiments without having to manage physical accelerator infrastructure. This improves developer productivity, speeds up model iteration, and can make large-scale AI workloads more predictable and cost-effective when the TPU resources are used efficiently. Review collected by and hosted on G2.com.