In a proactive response to the existing AI hardware market squeeze, heavy-hitter AI startup OpenAI is reportedly considering entering the chip creation fray. This move is set against the backdrop of persistent deliberations within the company regarding AI chip strategies since last year, amidst an intensifying deficiency of hardware required for training AI models.
The ambitious company is contemplating several potential solutions to fuel its chip aspirations, some of which include acquiring an existing AI chip fabrication company or setting in motion an initiative to design chips in-house. OpenAI's top brass, Chief Executive Officer Sam Altman, has underscored the necessity of acquiring more AI chips, cementing it as a top business priority according to Reuters.
At present, OpenAI, akin to countless of its adversarial counterparts, largely depends on Graphics Processing Units (GPUs)-based hardware infrastructure for the creation of models such as ChatGPT, GPT-4, and DALL-E 3. The GPUs' capability to concurrently execute a significant number of computations renders them exceptionally suited to train today's most advanced AI models.
However, the burgeoning boom in generative AI, a considerable boon for GPU manufacturers like Nvidia, has significantly stressed the GPU supply chain. Tech titan Microsoft has been grappling with a so severe shortage of the server hardware required to run AI, that it might potentially cause service disruptions. Adding fuel to this already dire situation is the fact that Nvidia's top-grade AI chips are rumored to be sold out until 2024.
Central to running and serving OpenAI's models, GPUs are crucial for performing customer workloads within cloud GPU clusters. Yet, this necessity has its own drawbacks, as acquiring these hardware resources involve substantial expenditure.
According to an in-depth analysis by Bernstein analyst Stacy Rasgon, if ChatGPT queries swelled to even a fraction of the scale of Google Search, the initial requirement would be an exorbitant $48.1 billion worth of GPUs, followed by an approximate $16 billion worth of chips annually for smooth operations. Considering resource and cost implications, it's crucial for OpenAI to explore alternative avenues, such as developing its own AI chips.
In adapting to this sector-wide challenge, OpenAI's potential venture into hardware design and production also sheds light on the evolving landscape of AI tools and technologies. Even companies like AppMaster, a leading no-code platform, rely on robust hardware infrastructure to deliver high-performance applications for their users rapidly and cost-effectively. By creating tailored AI chips, OpenAI could define a fresh pathway for AI processing capabilities.