No company has seen a greater benefit from the artificial intelligence boom than Nvidia (NVDA 2.94%). Training massive large-language models and running them requires gargantuan computing capacity. Nvidia's data center GPUs are the gold standard in the AI industry, and they're being scooped up by cloud giants and AI start-ups alike.
Nvidia's revenue more than tripled in its most recent quarter thanks to incredible demand for its AI-centric GPUs, while adjusted earnings per share rose by nearly a factor of six. Nvidia's market capitalization has surged to over $2 trillion, making it one of the most valuable companies in the world.
NASDAQ: NVDA
Key Data Points
Why Nvidia stock may not be the best AI bet
While Nvidia is riding high right now, competition is almost guaranteed to eat away at the company's AI market share. Advanced Micro Devices is now putting up a fight with its own high-powered AI chips, and major tech and cloud computing companies are increasingly designing custom AI chips optimized for their data centers.
- Amazon has developed multiple AI chips: Inferentia for AI inference and Trainium for AI training.
- Alphabet's Google is on the fifth generation of its tensor processing unit, or TPU, with the company rolling out a version optimized for AI large models late last year.
- Meta is reportedly working on a custom AI chip to help power its massive AI ambitions.
- Microsoft unveiled its custom Maia 100 AI chip for its Azure cloud platform last year. It will initially be used to power the company's AI services, including Copilot and the OpenAI services running on Azure.
While these mega-companies are still buying plenty of Nvidia's GPUs, options for training and running AI models are expanding. Nvidia's utter dominance of the AI chip market is almost certain to erode as time goes on.
Someone has to manufacture and package all those AI chips
Designing an AI chip is something that any large tech company with a big budget is capable of doing. Manufacturing those chips is another story. Today, there's only one viable option for manufacturing AI chips using the most advanced process. TSMC is the foundry market leader, and it has a significant technological lead over its competitors.
This situation is set to change in 2025 as Intel (INTC 5.41%) enters the foundry market in earnest. The chip giant has been executing its plan to roll out five new process nodes in four years, with the final node expected to overtake TSMC technologically. The Intel 18A process should be ready for volume production next year, and Intel has already snagged some important customers.
Intel has secured $15 billion of orders for its foundry business so far, spanning not only Intel 18A but also other nodes as well as advanced packaging services. Nvidia is reportedly planning to use Intel for some of its advanced packaging needs, although that rumor hasn't been confirmed. A deal with Microsoft to use Intel 18A for an undisclosed future chip has been confirmed, representing a massive win for Intel's fledgling foundry business.
Beyond Intel 18A, Intel has two additional processes on its roadmap meant to cement its manufacturing lead. Intel 14A is set to arrive in 2026, and Intel 10A will follow in 2027. Volume production of both processes will take time to ramp up following their introductions.
NASDAQ: INTC
Key Data Points
A $230+ billion opportunity
The foundry market is expected to more than double by 2032 to surpass $230 billion. Intel's market share rounds down to zero today, but if it succeeds in regaining its manufacturing edge and ramping up the capacity of its leading nodes, the company can rapidly grow that market share over the next decade.
Intel is valued at less than $200 billion today, with the stock still down significantly from its pandemic-era high. While Intel's core PC and server CPU businesses will remain key parts of the company, the foundry business has by far the best long-term growth potential.
Importantly, Intel's foundry can succeed regardless of how the semiconductor and AI industries evolve. Do GPUs remain the dominant method to train and run AI models? Intel could potentially manufacture them. Do custom AI chips proliferate? Intel could manufacture those as well. What if demand for AI chips falls short of forecasts? Intel can manufacture other types of chips destined for the data center, smartphones, PCs, and other devices.
While Nvidia is the AI darling today, it faces an onslaught of looming competition. That onslaught is good news for the foundry industry, and it's particularly good news for Intel as it aims to grow into the world's second-largest foundry by 2030.
