King of GPUs vs King of Chip Manufacturing: Who is More Worth Buying in the Era of AI Chips?

NVIDIA and TSMC, one designs and the other manufactures, both are top players in the AI chip ecosystem. However, the investment logic differs significantly. Today, let's discuss the differences between these two companies.

NVIDIA: The Absolute Monopoly in AI Infrastructure

NVIDIA's GPUs support the entire AI era—from cloud computing to autonomous driving, nothing can do without them. The latest financial report is impressive: third-quarter revenue reached $57 billion, a year-on-year increase of 62%; earnings per share are $1.30, with a growth rate of 60%.

The data center business is the real cash cow - with an income of $5.122 billion accounting for 89.8% of total revenue, a year-on-year increase of 66%. The shipment of Blackwell chips continues to explode, which is used for training and inference of large models. Coupled with the collaboration with OpenAI to build a super-large AI data center, NVIDIA's demand side is basically locked in.

Growth Expectations: The revenue growth rate for FY2026 is expected to be 57.9%, and 36.1% in 2027. This growth rate is indeed impressive among large companies.

TSMC: The Gatekeeper of Manufacturing Processes

TSMC controls the world's most advanced chip foundry capacity - it has already started mass production of 3nm, and 2nm is coming soon. Major companies like NVIDIA, Broadcom, and Marvell have to queue up for production.

Last year, AI-related revenue tripled, and this year it is expected to double again. In the third quarter, revenue was $33.1 billion, a year-on-year increase of 41%; earnings per share were $2.92, with a growth rate of 39%.

But TSMC also has its troubles. In order to keep up with the demand for AI chips, it plans to invest 40-42 billion USD in expansion by 2025 (having spent only 29.8 billion in 2024). Seventy percent of this will be allocated to advanced processes, which is necessary. The problem lies in the new factories in Arizona, Japan, and Germany, where high labor and energy costs and low initial utilization rates could compress gross margins by 2-3 percentage points each year, potentially lasting for 3-5 years.

A greater hidden danger is geopolitical issues—China is an important source of revenue for TSMC, and the risks of U.S. export controls and further sanctions should not be underestimated.

Growth Expectations: Revenue is expected to increase by 33.7% in 2025, and EPS is expected to increase by 44%. By 2026, the revenue growth rate is only 20.6%, and EPS is expected to grow by 20.2%. Growth is clearly slowing down.

Valuation and Price Increase Comparison

Both of these are not bad this year: Nvidia increased by 34.5%, and TSMC increased by 40.5%. But the valuation difference is significant—TSMC's price-to-earnings ratio is 23.47 times, while Nvidia's is 30.38 times.

NVIDIA's premium is actually reasonable because its growth rate is faster, its ecosystem is stronger, and the advantages of its hardware-software integration cannot be replicated by others. TSMC is cheap, but its growth rate is declining, and it is also being suffocated by geopolitical pressures and the costs of building factories overseas.

Investment Advice

NVIDIA is more worth investing in. Although both companies are good and can benefit from the AI boom, NVIDIA is the absolute winner of this game—faster growth, stronger moat, and clearer future. TSMC is also good, but it faces cost pressures and political risks in the short term, making it less attractive than NVIDIA.

Zacks has given Nvidia a rating of Buy (#2)，台积电是持有(#3), and the market sentiment has made it very clear.