Nvidia (NVDA -0.72%) is considered a pioneer in the artificial intelligence (AI) hardware market, and rightly so, as the chip designer's graphics processing units (GPUs) have allowed cloud computing companies and others to train AI models and run inference applications.
The parallel computing power of Nvidia's GPUs makes them ideal for performing a large number of calculations simultaneously, which is precisely what's required for training AI models. Also, these chips are now gaining traction in AI inference as well, thanks to their ability to quickly make predictions and decisions using the trained model.
Not surprisingly, Nvidia has established a solid foothold in the AI chip market. It towers above its competitors with an estimated market share of 80% in AI data center accelerators. However, Nvidia's dominance wouldn't have been possible without its foundry partner Taiwan Semiconductor Manufacturing (TSM +0.90%), which is the real kingpin of the AI chip market.
Let's look at the reasons why TSMC is a bigger power player than Nvidia in AI chips.
Image source: Getty Images.
TSMC's dominant foundry position makes it the go-to manufacturer of AI chips
TSMC operates semiconductor fabrication plants across the globe, which are used to manufacture chips based on designs provided by its customers. It is worth noting that TSMC doesn't design its own chips. It simply makes chips for fabless semiconductor companies that don't have production facilities of their own.
NYSE: TSM
Key Data Points
Nvidia is one such company that utilizes TSMC's facilities for manufacturing its AI chips. Equity research and brokerage firm Bernstein estimates that Nvidia could account for over a fifth of TSMC's top line this year, up significantly from around 5% to 10% a couple of years ago. That's not surprising, as Nvidia has been aggressively looking to secure more of TSMC's chipmaking capacity.
Taiwan-based business newspaper Economic Daily News pointed out earlier this year that Nvidia has reportedly secured more than 70% of TSMC's advanced chip packaging capacity for 2025 in a bid to meet the robust demand for its AI GPUs. However, Nvidia is not the only company that's in line to utilize TSMC's fabs.
Apple (AAPL -2.66%) is another major customer, and its contribution toward TSMC's top line is expected to be identical to that of Nvidia's in 2025. The consumer electronics giant taps TSMC to manufacture the processors that go into popular devices such as iPhones and iPads, and it has reportedly pre-booked the foundry giant's 2-nanometer (nm) capacity to mass produce chips for its next-generation iPhones.
It is worth noting that Apple had reportedly booked all of TSMC's 3nm supply in 2023 to make processors for the iPhone and other devices. And now that Apple is looking to bolster the on-device AI capabilities of its devices, it is expected to move to the 2nm node so that it can pack more computing power and increase energy efficiency.
Apple, however, has company, as another smartphone chip designer -- Qualcomm -- is expected to produce chips based on TSMC's 2nm process node as well. On the other hand, Nvidia's peers in the AI accelerator market are also partnering with TSMC to manufacture advanced chips.
Marvell Technology, for instance, is reportedly going to adopt TSMC's sub-3nm process nodes to manufacture the next generation of its custom AI processors, which are in tremendous demand from cloud computing giants to reduce costs. Meanwhile, AMD is getting its central processing units (CPUs) and GPUs that power both servers and personal computers (PCs) manufactured by TSMC as well.
Clearly, TSMC is the power player in the AI chip market. Its plants manufacture chips that go into a wide variety of applications, ranging from smartphones to PCs to data centers, and all of these markets are on track to record secular growth because of AI. Importantly, TSMC is taking steps to ensure that it can meet the incredible demand from all of these markets.
An aggressive expansion plan should help it satisfy the booming AI chip demand
TSMC's 2025 capital expenditure forecast of $38 billion to $42 billion points toward a significant increase over its 2024 outlay of $30 billion. It is going to invest 70% of its 2025 capex on advanced process technologies that are used for making AI chips, which isn't surprising.
Moreover, the company has aggressive long-term expansion plans as well. It has outlined an investment of $165 billion in the U.S. to build more plants, while it is also building factories in Taiwan and Europe. These expansionary moves should enable TSMC to capitalize on the AI chip market's impressive long-term growth.
According to one estimate, the global AI chipset market could clock an annual growth rate of 31% through 2033, which means that TSMC has the ability to sustain its terrific growth for years to come. Not surprisingly, analysts are expecting a pick-up in TSMC's growth going forward.
TSM EPS Estimates for Current Fiscal Year data by YCharts
That's why it would be a good idea to buy this AI stock hand over fist right now, as it seems undervalued. TSMC's earnings are expected to jump by 34% this year, which is nearly five times the projected increase in the S&P 500 index's average earnings. With the stock trading at 28 times earnings, investors are getting a good deal on TSMC based on the potential upside it could deliver.
