NVIDIA and Google Cloud have joined forces to empower startups worldwide in expediting the development of generative AI applications and services. Unveiled at Google Cloud Next ‘24 in Las Vegas, this collaboration merges the NVIDIA Inception program for startups with the Google for Startups Cloud Program. It aims to broaden access to cloud credits, market support, and technical guidance, facilitating startups in delivering value to their clientele more swiftly.

Under this collaboration, eligible participants of NVIDIA Inception, a global initiative aiding over 18,000 startups, will enjoy an expedited route to leverage Google Cloud infrastructure, along with access to substantial Google Cloud credits, potentially up to $350,000 for those concentrating on AI.

Members of the Google for Startups Cloud Program can seamlessly integrate into NVIDIA Inception, unlocking access to technological know-how, NVIDIA Deep Learning Institute course credits, hardware, software, and more. Moreover, qualified members of the Google for Startups Cloud Program can partake in NVIDIA Inception Capital Connect, a platform connecting startups with venture capital firms keen on the field.

Promising software developers from both programs will receive accelerated onboarding to Google Cloud Marketplace, in addition to co-marketing and product acceleration assistance.

This collaboration represents the latest effort by both entities to alleviate the costs and challenges associated with developing generative AI applications, particularly for startups constrained by high AI investment costs.

In February, Google DeepMind introduced Gemma, a suite of cutting-edge open models. In collaboration with Google, NVIDIA recently optimized its AI platforms for Gemma, enhancing performance and reducing costs for customers across various domains.

Through concerted efforts, Gemma’s performance was enhanced using NVIDIA TensorRT-LLM, in conjunction with NVIDIA GPUs, aiming to expedite innovative work for specific-use cases.

Additionally, NVIDIA NIM microservices, coupled with Google Kubernetes Engine (GKE), offer a streamlined pathway for developing AI-powered applications and deploying optimized AI models into production. These technologies support a range of leading AI models, facilitating seamless and scalable AI inferencing to expedite generative AI deployment in enterprises.

Further enhancing the accessibility of NVIDIA-accelerated generative AI computing, Google Cloud announced the imminent general availability of A3 Mega, an expansion to its A3 virtual machine family, powered by NVIDIA H100 Tensor Core GPUs. These instances will double the GPU-to-GPU network bandwidth from existing A3 VMs.

Google Cloud’s new Confidential VMs on A3 will provide support for confidential computing, ensuring data confidentiality and integrity during training and inference, without necessitating code alterations while accessing H100 GPU acceleration.

Looking ahead, NVIDIA’s upcoming GPUs based on the NVIDIA Blackwell platform, namely the NVIDIA HGX B200 and the NVIDIA GB200 NVL72, are slated to debut on Google Cloud early next year. These GPUs are tailored for demanding AI, data analytics, high-performance computing workloads, and massive-scale, trillion-parameter model training and real-time inferencing.

The NVIDIA GB200 NVL72, featuring 36 Grace Blackwell Superchips, each with two NVIDIA Blackwell GPUs connected via an advanced chip-to-chip interconnect, promises breakthrough performance, delivering faster real-time inference and training compared to previous generations.

NVIDIA DGX Cloud, an AI platform optimized for generative AI demands, will be available on Google Cloud’s A3 VMs powered by H100 GPUs, with future plans to include the GB200 NVL72 in 2025.

Leave a Reply

Your email address will not be published