OpenAI, a well-funded startup specializing in artificial intelligence (AI), is considering the idea of creating its own AI chips. The company has been discussing its chip strategies for some time now, as there is a shortage of chips available for training AI models. OpenAI is exploring several options to advance its chip ambitions, including acquiring a chip manufacturer or designing chips internally.
The CEO of OpenAI, Sam Altman, has made acquiring more AI chips a top priority for the company. Currently, like its competitors, OpenAI relies on GPU-based hardware to develop models such as ChatGPT, GPT-4, and DALL-E 3. GPUs are highly suitable for training today’s most advanced AI due to their ability to perform numerous computations in parallel.
However, the immense demand for GPUs caused by the generative AI boom has put a strain on the GPU supply chain. Microsoft has warned of a severe shortage of server hardware needed to run AI, which could potentially lead to service disruptions. Additionally, Nvidia’s best-performing AI chips are reportedly sold out until 2024.
GPUs are also crucial for running and serving OpenAI’s models, but they come at a high cost. An analysis by Bernstein analyst Stacy Rasgon estimates that if ChatGPT queries reached a tenth of the scale of Google Search, it would require billions of dollars’ worth of GPUs initially and annually.
Creating its own AI chips is not a new concept, as companies like Google, Amazon, and Microsoft have already pursued this path. Google has its own processor, the TPU, specifically designed for training large generative AI systems. Amazon provides proprietary chips to its AWS customers for both training and inferencing purposes. Microsoft is reportedly working with AMD to develop its in-house AI chip called Athena, which OpenAI is said to be testing.
OpenAI is in a strong financial position to invest heavily in research and development (R&D). The company has raised over $11 billion in venture capital and is approaching $1 billion in annual revenue. Its potential share sale could raise its secondary market valuation to $90 billion.
However, the hardware business, especially AI chips, is challenging. Last year, AI chipmaker Graphcore experienced a valuation cut of $1 billion after a deal with Microsoft fell through, leading to job cuts. Habana Labs, an AI chip company owned by Intel, also laid off a portion of its workforce. Meta, another company exploring custom AI chips, faced issues and had to cancel some experimental hardware.
Even if OpenAI decides to bring a custom chip to market, the process could take years and require significant financial investment. The willingness of OpenAI’s investors, including Microsoft, to take such a risky bet remains to be seen. Hardware development is a demanding endeavor, particularly in the field of AI chips.