Google and Blackstone create a cloud firm to expand access to TPUs

What happened

Google and Blackstone announced an artificial intelligence cloud infrastructure initiative combining data centre investment with access to Google's TPU chips under a compute-as-a-service model.

The move aims to expand the availability of specialised AI compute capacity, in a market where access to this type of hardware is one of the main bottlenecks for companies training or running models at scale.

What it means for a business evaluating AI workload migration

New players and models for accessing specialised compute can open up options once reserved for large companies. But switching providers or adding a new one is not only about price: it involves model compatibility, network performance and data migration processes.

Before moving a workload, it is worth assessing the total cost of the change, including available support and migration time, not just the advertised compute price.

What to review before switching cloud provider

Key points to avoid surprises when migrating AI workloads:

• Compatibility of the models and frameworks used with the new platform.
• Network performance and cost between the environments involved.
• Estimated cost and time to transfer existing data.
• Level of support available during and after the migration.

Frequently Asked Questions

What does this Google-Blackstone alliance bring?

It expands access to Google's TPU chips through a compute-as-a-service model aimed at artificial intelligence infrastructure.

Does this affect businesses that do not use AI at scale?

Indirectly, by increasing supply and competition in the specialised compute market.

What should be checked before switching AI compute provider?

Model compatibility, network performance, migration cost and available support.