Top of the page

New startup aims to build supercomputer with 22000 H100 GPUs

Building a server with 8 H100 GPUs might seem both daunting, time-consuming, and expensive although will be a worthy endeavor given the mighty power of the H100 GPU. We see ourselves building these on a regular basis and workloads are executed with extreme performance. 

Inflection AI, a new startup company, just raised $1.3bn in cash and cloud credits. The founder made it immediately clear what would be done with those funds. “We will be building a cluster of around 22,000 H100s” This is quite the supercomputer. It will be the second most powerful, depending on the full details of Aurora and El Capitan when they come online. 

Inflection has a product called Pi. In its current form, it is essentially a free personal assistant. It will be interesting to see how they plan to use their supercomputer with 22000 H100 GPUs. 

Many workloads and new AI experiments can be done with servers with a lot less than 8 H100 GPUs - which we deploy on a very regular basis. Talk to us about your AI needs and we’d be happy to put together a quote for you.

General Enquiry

New startup aims to build supercomputer with 22000 H100 GPUs

Building a server with 8 H100 GPUs might seem both daunting, time-consuming, and expensive although will be a worthy endeavor given the mighty power of the H100 GPU. We see ourselves building these on a regular basis and workloads are executed with extreme performance. 

Inflection AI, a new startup company, just raised $1.3bn in cash and cloud credits. The founder made it immediately clear what would be done with those funds. “We will be building a cluster of around 22,000 H100s” This is quite the supercomputer. It will be the second most powerful, depending on the full details of Aurora and El Capitan when they come online. 

Inflection has a product called Pi. In its current form, it is essentially a free personal assistant. It will be interesting to see how they plan to use their supercomputer with 22000 H100 GPUs. 

Many workloads and new AI experiments can be done with servers with a lot less than 8 H100 GPUs - which we deploy on a very regular basis. Talk to us about your AI needs and we’d be happy to put together a quote for you.

General Enquiry