It is important to note that the AI business is not only software, but also hardware. In fact, the current limiting factor is the latter, as there is no manufacturing that can sustain the current demand for artificial intelligence chips.
A new report from Bloomberg states that OpenAI CEO Sam Altman’s efforts to raise billions for an AI chip company are aimed at using that money to develop a “network of manufacturing factories” that would span the globe and involve working with unidentified “leading chip manufacturers.”
One of the main costs and limitations when running AI models is having enough chips to handle the calculations of bots like ChatGPT or DALL-E, which answer questions and generate images.
Altman wants money for his big project
Nvidia’s value surpassed one trillion dollars for the first time last year, partly due to the virtual monopoly it has, as GPT-4, Gemini, Llama 2, and other models heavily rely on their popular H100 GPUs.
As a result, the race to manufacture more high-power chips to run complex AI systems has only intensified. The limited number of factories capable of manufacturing high-end chips forces Altman or anyone else to bid for capacity years before needing it to produce the new chips.
And facing companies like Apple requires investors with a lot of money to take on costs that the non-profit organization OpenAI cannot yet afford. SoftBank Group and the AI holding company G42, based in Abu Dhabi, have held discussions to raise funds for Altman’s project.
Other companies that develop AI models have also ventured into manufacturing their own chips. Microsoft, an investor in OpenAI, announced in November that it has built its first custom AI chip for training models, closely followed by Amazon, which announced a new version of its Trainium chip.
AWS, Azure, and Google also use Nvidia’s H100 processors. This week, Meta’s CEO, Mark Zuckerberg, stated that “by the end of this year, Meta will own over 340,000 Nvidia H100 GPUs,” as the company pursues the development of artificial general intelligence (AGI).
Nvidia has already announced its next generation of GH200 Grace Hopper chips to expand its dominance in this field, while its competitors AMD, Qualcomm, and Intel have released processors designed to power AI models in laptops, phones, and other devices.
