Nvidia and Google Cloud are integrating their startup accelerator programs to help generative artificial intelligence (AI) companies access cloud credits, go-to-market support and technical expertise, which should lower the barriers related to developing genAI applications.

By combining the Nvidia Inception program and the Google for Startups Cloud Program, members of one vendor’s accelerator now have access to the benefits of the other program.

Qualified members of Nvidia’s program, which works with more than 18,000 startups, now have a faster path to leverage Google Cloud infrastructure with up to $350,000 in cloud credits.

In addition, current Google for Startups Cloud Program members can join Nvidia’s accelerator program to access technology expertise, Nvidia Deep Learning Institute course credits and Nvidia hardware and software. Eligible members can also take part in Nvidia’s Inception Capital Connect platform, which brings together startups and venture capital firms interested in genAI.

Emerging software developers on high-growth trajectories who belong to both accelerators can also access quicker onboarding to Google Cloud’s marketplace, co-marketing assistance and product acceleration support.

Nvidia also announced that its newest GPUs based on the Blackwell platform will be available on Google Cloud early next year: the Nvidia HGX B200 and the Nvidia GB200 NVL72.

HGX B200 is ideal for the demands of AI, data analytics and high performance computing (HPC) (HPC) workloads, while the GB200 NVL72 is designed for much larger scale model training and real-time inference – up to a trillion parameters.

GB200 NVL72 is made up of 36 Grace Blackwell superchips. Each superchip houses two Nvidia Blackwell GPUs and an Nvidia Grace CPU over a 900 GB/s chip-to-chip interconnect. This architecture supports up to 72 Blackwell GPUs in one domain and 130TB/s of bandwidth.

The vendor claims GB200 NVL72 eliminates communication hurdles and is able to act as a single GPU, resulting in 30-times faster large language model (LLM) inference and four-times faster training when compared to the previous generation of this technology.

Once deployed in Google’s public cloud environment, the hyperscaler plans to combine Nvidia’s multi-node, rack-scale system with its fourth generation of liquid cooling systems to optimize resource use.