Demand for NVIDIA's (NVDA -2.76%) AI chips is proving insatiable. The launch of ChatGPT from OpenAI late last year opened the AI floodgates by demonstrating the power and potential of generative AI. The latest AI tech can generate text, write code, produce images and videos, mimic voices, and do all sorts of other things.
Both training and using AI models requires an incredible amount of computational horsepower. NVIDIA's GPUs are well suited for the task, and they've become the de facto standard for AI workloads. The company offers a family of data center products, and it's working to ramp up supply to meet surging demand.
NVIDIA expects its second-quarter revenue to rocket to $11 billion, up from $7.19 billion in the first quarter, largely due to this AI boom. NVIDIA doesn't manufacture its own chips but instead relies on third-party foundries like TSMC and Samsung. TSMC has the technological edge, so if NVIDIA wants the most advanced manufacturing process for its ultrapricey AI chips, TSMC is the only game in town.
Intel's manufacturing push
Chip giant Intel (INTC 1.31%) is doing a variety of things related to AI. Its latest Sapphire Rapids server CPUs, for example, have AI accelerators built in, and its Arc graphics cards include AI-powered features.
Intel's push to build a foundry business that rivals TSMC is yet another AI play. The company is aggressively building manufacturing capacity and rolling out process nodes. The Intel 4 process is ready to go today, and the Intel 3 process is scheduled to go into production by the end of the year. After those, the Intel 20A and 18A processes should come online in 2024 and 2025, respectively. Intel is aiming to reclaim manufacturing leadership with this rapid-fire manufacturing-process blitz.
Intel has won some major customers for future processes, including an unnamed data center solutions provider for the Intel 3 process. Based on recent comments by NVIDIA CEO Jensen Huang, it's possible that this customer is NVIDIA. And if not, NVIDIA will be a potential customer down the road for a more advanced process.
Speaking at a question-and-answer session, Huang responded to questions about improving the supply of its AI chips:
You know that we also manufacture with Samsung, and we're open to manufacturing with Intel. Pat [Gelsinger] has said in the past that we're evaluating the process, and we recently received the test chip results of their next-generation process, and the results look good.
That "next-generation process" could refer to Intel 3, which will be ready for mass production this year.
NVIDIA's dependence on TSMC for its most advanced chips puts it in a position where it doesn't have any real leverage. If Intel becomes a viable option, the company can secure more manufacturing capacity while potentially getting better pricing.
For Intel, winning a customer like NVIDIA would be a huge vote of confidence in its foundry business. Other companies like Apple that are entirely dependent on TSMC and need the most advanced manufacturing tech for their chips could eventually follow suit and move some orders over to Intel. Intel has teamed with ARM to optimize its 18A process for ARM-based chips, an effort that will start to pay off in 2025.
If Intel can bring its manufacturing technology up to par with TSMC over the next few years, customers like NVIDIA that need the most advanced processes are going to seriously consider Intel for manufacturing. NVIDIA's is going to need all the manufacturing capacity it can get to meet demand for its red-hot AI chips, and Intel will likely benefit from that demand sooner or later.
