AI research lab OpenAI might start manufacturing its own AI accelerator chips to address GPU shortages and running costs. The organization behind AI products like ChatGPT and DALL-E 3 is evaluating options, including potential acquisitions or collaborations with other chip makers such as Nvidia.
OpenAI mulls over manufacturing AI chips
OpenAI, known for pioneering AI products like ChatGPT and DALL-E 3, is contemplating the production of its own AI accelerator chips. The decision is driven by an ongoing shortage of specialized AI GPU chips and the steep costs associated with running them. The AI research lab is still weighing its options, which might include acquiring a chipmaking company or partnering with existing manufacturers like Nvidia. Yet, a final resolution has yet to be made.
High costs of OpenAI's reliance on GPUs
OpenAI's operations heavily depend on a colossal supercomputer, developed by Microsoft, one of its principal backers. This supercomputer employs 10,000 Nvidia GPUs, which ends up incurring substantial costs. For instance, running ChatGPT, OpenAI's conversational AI, costs about 4 cents per query. If the queries scale up to even one-tenth of Google search's volume, the initial GPU investment would amount to a whopping $48.1 billion, with an estimated $16 billion in annual maintenance costs.
By manufacturing its own AI chips, OpenAI would join the ranks of other tech heavyweights like Google and Amazon, who have also dabbled in crafting custom chips. This strategic shift would signify a considerable financial undertaking for OpenAI, but it could fast-track the process if they proceed with an acquisition of a current chip firm—similar to Amazon's purchase of Annapurna Labs in 2015, which expedited its custom chip projects.
Despite any strategic plans, the actual development of a custom chip would likely span several years. This means OpenAI would still have to depend on commercial chip providers like Nvidia and AMD during the interim. It's no walk in the park either; Meta's attempt to create its own AI chips has faced challenges, as reported by Reuters.
Microsoft's custom AI chip in the works
Microsoft, the primary backer of OpenAI, is reportedly working on a custom AI chip that OpenAI is presently testing. The tech giant is expected to unveil this chip in the coming month. This could potentially provide another avenue for OpenAI to address its current challenges with AI chip shortages and high operating costs.