At Google Cloud Next 2024, Google also introduces Axion, a specially designed Arm-based chip for data centers.
At the Google Cloud Next 2024 conference in Las Vegas, Alphabet’s Google Cloud introduced a plethora of new goods and services, including a program designed to assist startups and small businesses in developing generative AI applications and services.
The program combines the benefits of Google for Startups Cloud Program and Nvidia Inception program for startups, and it adds cloud credits, go-to-market assistance, and technical know-how to support businesses in their AI endeavors.
More than 18,000 startups have benefited from Inception, a global initiative from Nvidia that supports them. To attract more, Nvidia is providing an expedited route to use Google Cloud infrastructure along with Google Cloud credits, with up to $350,000 available for AI-focused businesses.
Members of the Google for Startups Cloud Program can join Nvidia Inception in exchange for access to technical know-how, course credits from the Nvidia Deep Learning Institute, Nvidia hardware and software, and more. Additionally, Inception provides a platform called Capital Connect, which matches entrepreneurs with venture capital firms that are interested in their industry.
Google Introduces its own Arm CPU, Axion
In addition, Google revealed the Axion series of Arm processors for its cloud services, becoming the latest corporate behemoth to produce its own unique Arm-based chips. Since 2018, Microsoft has been selling the Cobalt 100 Arm chip, and Amazon Web Services has been selling Graviton processors since last fall.
Google has already experimented with bespoke silicon. In order to expedite its own workloads, it has been using tensor processing units (TPU) since 2015. In 2018, it introduced a video coding unit (VCU) for video transcoding. But this will be its first custom silicon that interacts with customers.
Axion is based on the Neoverse V2 design from Arm, which is an ARMv9-based data center processor. Arm does not manufacture chips; rather, it creates designs, which licensees then use to add to the base configuration that Arm provides. Some companies (Apple, Qualcomm) produce smartphones, while others (Ampere) produce server processors.
Google stated that Axion processors would deliver instances with up to 30% better performance than the fastest general-purpose Arm-based instances currently available in the cloud, up to 50% better performance, and up to 60% better energy-efficiency than comparable current-generation x86-based instances. However, Google declined to comment on specifics regarding speeds, fees, and core counts.
Titanium, a system comprising Google’s own specially designed silicon microcontrollers and tiered scale-out offloads, is the foundation upon which Axion is based. Similar to how the SuperNIC removes networking traffic off the CPU, it offloads tasks like networking and security so that Axion processors can concentrate on workload computing.
According to Google, virtual machines built on Axion processors will be accessible in preview in the upcoming months.
Updated AI Software Services
Google unveiled Gemma in February, a collection of free models built using the same technology and research as Google’s Gemini generative AI service. Teams from Google and Nvidia have now collaborated to use Nvidia’s TensorRT-LLM, an open-source toolkit for LLM inference optimization, to speed up Gemma’s performance.
Using its GKE Kubernetes engine and Google Cloud HPC Toolkit, Google Cloud has also made it simpler to implement Nvidia’s NeMo framework for creating unique generative AI applications across its platform. This makes it possible for developers to accelerate the creation of generative AI models, resulting in the quick deployment of turnkey AI solutions.