The latest Google Cloud Next event this week featured another instance of Nvidia and Google strengthening their ongoing partnership around AI even as Google continues to make progress developing its own AI chips.
During the Las Vegas event, Nvidia made the anticipated announcement that Google Cloud in early 2025 will get GPUs based on the new Blackwell architecture that Nvidia unveiled at its own recent GTC event. That platform will be featured on Google Cloud in two variations: the HGX B200 AI supercomputing platform and the GB200 NVL72, a multi-node rack-scale system that will be combined with Google Cloud’s fourth generation of advanced liquid-cooling systems
The HGX B200 is designed for AI, data analytics and high-performance computing workloads, while the GB200 NVL72 is designed for next-frontier, massive-scale, trillion-parameter model training and real-time inferencing, Nvidia stated. The company added in a blog post that the GB200 NVL72 connects 36 Grace Blackwell Superchips, each with two Blackwell GPUs combined with an Nvidia Grace CPU over a 900GB/s chip-to-chip interconnect, supporting up to 72 Blackwell GPUs in one NVLink domain and 130TB/s of bandwidth. It overcomes communication bottlenecks and acts as a single GPU, delivering 30x faster real-time LLM inference and 4x faster training compared to the prior generation.
Offering another example of the strong ties between the companies, Nvidia and Google Cloud also announced in Las Vegas that they are forging stronger links between their existing start-up programs to help young firms accelerate the creation of generative AI applications and services.
The move will allow qualified members of the Nvidia Inception program, which already supports more than 18,000 start-ups, to have an accelerated path to using Google Cloud infrastructure with access to Google Cloud credits — up to $350,000 for those focused on AI, the companies said. Also, members of theGoogle for Startups Cloud Program can join Nvidia Inception and gain access to technological expertise, Nvidia Deep Learning Institute course credits, Nvidia hardware and software, and more. Eligible members of the Google for Startups Cloud Program also can participate in Nvidia Inception Capital Connect, a platform that gives startups exposure to venture capital firms interested in the space.
These announcements arrived the same week that Google continued to mark progress in the development of its own custom AI chips to be used in data centers. The company unveiled the latest generation of its custom-designed tensor processing unit (TPU), called the TPU v5p, and also announced the Axion processor, the firm’s “first custom Arm-based CPUs designed for the data center,” according to Google’s The Keyword blog, which added, “Axion delivers industry-leading performance and energy efficiency and will be available to Google Cloud customers later this year.”
These custom developments are widely viewed as concerning in the long run for Nvidia and other traditional AI semiconductor companies, though some of these companies are likely to partner with Google on its ongoing custom chip efforts. Also, Google only plans to use its custom chips in its cloud and ease them to cloud customers, rather than trying to sell them directly to the market.