Chip designers like Nvidia aren’t the only companies riding high on the AI hype train. Foxconn chairman Young-Way Liu Wednesday forecast surging server sales over the next year as adoption of large language models and other powerful AI systems – needed to train and run that tech – grows.
“More and more people are using ChatGPT. We expect that in the second half of this year there may be a three digit increase,” Liu said during a shareholder meeting, according to Reuters.
According to Foxconn, the manufacturer’s server division has been growing rapidly, in large part due to interest in artificial intelligence. “The growth in this field has doubled in the last year, and it will continue to have a strong growth momentum this year,” the company’s annual report [PDF] said.
The mainstream adoption of generative AI has been pervasive as Google, Microsoft, IBM, and other orgs look to imbue their products with AI chatbots, code assistants, and other creative tools.
Of course, these AI models have to be trained and run on something and that means large clusters of AI accelerators. For the moment, that has largely meant Nvidia GPUs.
To put things in perspective, the cluster Microsoft built for OpenAI, whose models help power its Bing chatbot, uses more than 10,000 Nvidia GPUs. This has breathed life into Nvidia’s revenues following slowing PC demand and the onset of crypto winter.
But while Nvidia sees an opportunity to capitalize on the explosive growth of generative AI, they’re not the only ones in a position to benefit. Before those GPUs can be put to work they need to be assembled into larger systems made by a manufacturer like Foxconn.
The Taiwanese company is among the largest contract manufacturers in the world, producing a wide range of goods ranging from Apple iPhones and other consumer goods from electric vehicles to, you guessed it, servers.
In 2022, Foxconn said its server revenues topped NT$1.1 trillion (about $36 billion). But while optimistic about the growth of cloud service provider and AI server sales, the company’s outlook remains fairly conservative as a whole.
In a statement published in early May, Foxconn’s Liu warned in early May that the biz’s outlook remained “flattish” with inflation, geopolitical tensions, and slowing demand, introducing a heightened degree of uncertainty.
And don’t forget, with language models, large isn’t the only option. You probably don’t need massive amounts of hardware for that ML-based application you have in mind. ®
