Nvidia launches new software to enable businesses for AI integration

Nvidia for AI adoption: Nvidia has released its new software to facilitate the integration of Artificial Intelligence (AI) systems into businesses, expanding its array of offerings. This move underlines Nvidia's efforts to broaden its presence in the AI application execution sector, known as inference, where its chips currently face stiff competition, according to Joel Hellermark, CEO of Sana, a company specialising in AI assistants for businesses.

While Nvidia is known for supplying the chips used in training foundational models like OpenAI's GPT-4, which typically involves extensive data processing and is predominantly undertaken by AI-focused and major tech corporations, there's now a surge in demand from businesses of varying sizes seeking to incorporate these models into their operations. 

However, this process can be complex. The tools introduced by Nvidia aim to simplify the modification and deployment of diverse AI models on Nvidia hardware.

Ben Metcalfe, a venture capitalist and founder of Monochrome Capital, likened the approach to purchasing a ready-made meal instead of sourcing individual ingredients. He noted that while tech giants like Google, Doordash, and Uber can handle these tasks internally, Nvidia's increased GPU availability necessitates enabling more companies to leverage GPU capabilities. Less tech-savvy enterprises can use these "prepared recipes" to streamline their systems.

For instance, ServiceNow, a provider of software for technical support teams in large enterprises, used Nvidia's tools to develop a "copilot" to assist in resolving corporate IT issues.

The introduction of Nvidia's tools has attracted leading partners such as Microsoft, Alphabet Inc's Google, and Amazon, who will incorporate them into their cloud computing services. Additionally, companies like Google, Cohere, Meta, and Mistral are offering models compatible with these tools. Notably absent from the list are OpenAI, its financial backer Microsoft, and Anthropic, key providers of foundational models.

This move by Nvidia presents a potential revenue opportunity for the chipmaker, as the tools are part of its existing software suite, priced at $4,500 annually per Nvidia chip for private data centres or $1 per hour in cloud data centres.

(With Reuters inputs)