Nvidia and Google Cloud Collaborate to Empower AI Startups

During the Google Cloud Next 2024 conference in Las Vegas, Alphabet’s Google Cloud introduced a range of novel products and services, including a program aimed at assisting startups and small businesses in developing generative AI applications and services.

This venture amalgamates the Nvidia Inception program for startups with the Google for Startups Cloud Program, extending benefits such as cloud credits, go-to-market support, and technical guidance to aid startups in their AI endeavors.

The Inception program, catering to over 18,000 startups globally, offers an expedited route to leveraging Google Cloud infrastructure by granting access to Google Cloud credits, with a potential value of up to $350,000 for AI-focused firms. In return, members of the Google for Startups Cloud Program can join Nvidia Inception, unlocking access to technical resources, Nvidia Deep Learning Institute course credits, hardware, software, and more. The Inception platform also provides Capital Connect, facilitating startups’ exposure to venture capital firms.

Google Introduces Axion Custom Arm Processor Google further unveiled Axion, a new lineup of Arm processors designed for its cloud services, joining the ranks of tech giants crafting their own custom Arm-based chips. Since 2018, Amazon Web Services has offered Graviton processors, while Microsoft launched its Arm chip, Cobalt 100, last autumn.

This marks Google’s debut in the realm of customer-facing custom silicon, although it has previously developed custom silicon such as the tensor processing units (TPU) and video coding unit (VCU) for internal purposes. Axion, based on Arm’s Neoverse V2 design, leverages the ARMv9 architecture tailored for data center applications.

Google did not disclose specifics regarding speeds, fees, or core counts but asserted that Axion processors would deliver instances with up to 30% better performance than existing general-purpose Arm-based instances in the cloud, along with enhanced energy efficiency compared to current-generation x86-based instances.

Built on Google’s Titanium platform, Axion incorporates purpose-built custom silicon microcontrollers and tiered scale-out offloads to streamline operations such as networking and security, enabling the processors to prioritize workload computation, akin to how the SuperNIC offloads networking traffic from the CPU.

Google anticipates making virtual machines based on Axion processors available in preview in the upcoming months.