NVIDIA has almost maxed out TSMC’s 5nm lines with new AI chip orders

NVIDIA places urgent orders with TSMC for the H100 and A100, as well as the H800 and A800, which are specially designed for China. Orders from NVIDIA have already resulted in an increase in process utilization for TSMC’s 7nm and 6nm families, which had previously declined rapidly due to flagging demand, while utilization of the 5nm and 4nm process shops is rapidly increasing to almost full capacity.

Image source: Nvidia

In announcing the company’s results for the first quarter of fiscal 2024, NVIDIA CEO Jensen Huang noted that the computing industry is undergoing two transformations simultaneously, namely accelerated computing and generative AI. “The global data center infrastructure is shifting from general purpose computing to accelerated computing as companies seek to apply generative AI to every product, service and business process. – says a press release from NVIDIA. — Our entire family of data center products — H100, Grace CPU, Grace Hopper Superchip, NVLink, Quantum 400 InfiniBand and BlueField-3 DPU are in production. We are significantly increasing supply to meet growing demand.”

Nvidia is riding the artificial intelligence wave sparked by the launch of ChatGPT. The company has an overwhelming advantage in the AI ​​GPU market. Following US chip sanctions against China, NVIDIA was quick to release stripped-down versions of its H100 and A100 AI accelerators to circumvent the ban. Demand from major Chinese companies like Baidu for these accelerators, dubbed the H800 and A800, remains consistently high.

Nvidia reported impressive financial results for the first quarter of fiscal 2024, with revenue up 19% year over year to $7.19 billion. The outlook for the second quarter is no less rosy.

The popularity of NVIDIA’s AI accelerators also benefits from its main chip manufacturing partner TSMC. Previously, TSMC had tended to bet on orders from Apple, not anticipating an influx of urgent orders from NVIDIA. And TSMC’s negative forecasts for 2023 were based on deteriorating macroeconomic indicators and lower end-market demand. Now TSMC may revise its full-year 2023 guidance upwards.