One could argue that Nvidia’s $40,000 H100 chips are garnering similar levels of interest this season as AI chatbots such as OpenAI ChatGPT and Google Bard. Similar to many other AI companies, OpenAI utilizes Nvidia’s latest chip models for training its AI models.
The uptick in purchases of chips from OpenAI signifies the imminent availability of more advanced AI models that can outshine the current chatbot architectures.
The Surge in Demand for Nvidia H100 Chips
Training relational AI systems on vast amounts of data to generate complex responses requires substantial computational power. Nvidia’s H100 chips, which boast enhanced speed compared to previous models, are tailored for relational AI tasks. Willy Shih, a professor at Harvard Business School, highlighted to Quartz the correlation between faster question processing and the increased potency of these chips.
In response to the growing demand, various companies, including competitors like AMD and tech giants such as Google and Amazon, are intensifying efforts to develop more efficient chips tailored for AI tasks. Despite these endeavors, Nvidia remains at the forefront.
Notably, governments and venture capital firms are engaged in a competitive scramble for these highly sought-after chips, leading to a buying frenzy around Nvidia’s products. However, OpenAI’s successful chip procurement may signal a shift, potentially improving chip supplies for AI enterprises.
While Nvidia maintains its dominance, Prateek Kathpal, the CEO of SymphonyAI—a company specializing in AI bot development for internal corporate use—revealed discussions with AMD despite currently using AMD and Arm to power their AI software.
Implications of Increased Chip Acquisition for OpenAI
The expanding array of chips at OpenAI hints at several developments.
The acquisition of H100 chips will play a pivotal role in shaping the company’s forthcoming AI model, GPT-5, as per Altman. He indicated that the new model will require a wealth of data for training, sourced from both publicly available datasets and exclusive data from businesses. While the exact capabilities of GPT-5 remain undisclosed, Altman suggested it would likely surpass its predecessors in sophistication.
Although a launch date for GPT-5 was not disclosed, the swift succession of releases—GPT-4 unveiled just eight months after the 2020 debut of GPT-3—underscores a rapid development pace.
Furthermore, the increased chip acquisitions suggest that OpenAI is progressing towards the realization of artificial general intelligence (AGI), an AI system capable of performing tasks akin to human capabilities.
