Jensen Huang, CEO of Nvidia heralded the new era of computing, after unveiling their new Blackwell chip B200 GPU, the world’s most powerful chip  and software for running AI,  said, it would vastly boost the power driving large language models, as it has over 208 billion transistors compared with 80billion in last year’s H100. Hopper 100 is fantastic, but need even bigger GPUs, according to Huang. He delivers a keynote speech during the Nvidia GTC Artificial Intelligence Conference at SAP Centre on March 18, 2024, in San Jose, California. Mamvir Das the Nvidia enterprise VP in an interview said “ the sellable commercial product was the GPU and the we really have the commercial software to help people use the GPU in different ways”. The new software is easier to run programs on any of  Nvidia’s GPUs even older ones that might be better suited for deploying but not building AI. If you are a developer, you’ve got an interesting model you want people to adopt, if you put it in a NIM, we’ll make sure that it’s runnable on all our GPUs, so you reach a lot of people”.

Nvidia’s share price rocketed five-fold and total sale have more than tripled since OpenAI’s ChatGPT kicked off the AI boom in late 2022.  Companies like Meta and Microsoft are spending billions of dollars buying these chips.

The Blackwell GPU is large combining two separately manufactured dies into one chip manufactured by TSMC, and will also be available as an entire server called  GB200 NVLink 2, combining 72 Blackwell GPUs and other Nvidia parts designed to train AI models.