NVIDIA reported better-than-expected earnings in the fourth quarter of last year (November 2024 - January 2025), thanks to the performance of the industry's highest-performing graphics processing unit (GPU) 'Blackwell'. However, examining the detailed indicators reveals concerns that future revenue growth and gross profit margins are stagnating. Additionally, there are analyses indicating that the emergence of low-cost artificial intelligence (AI) models like DeepSeek and the U.S. government's semiconductor restrictions have increased uncertainty in this year's earnings outlook.
On the 26th (local time), NVIDIA announced that its fourth-quarter revenue was $39.33 billion, a 78% increase year-over-year. This figure exceeds Wall Street's expectations of $38.05 billion. Net income also rose by 80% year-over-year to $22.09 billion. Adjusted earnings per share (EPS) were $0.89, surpassing the market expectation of $0.84.
However, there are signs of instability in the earnings trend. NVIDIA's revenue growth rate for the fourth quarter of last year was 78%, the lowest among the last seven quarters. After the earnings announcement, NVIDIA's stock closed mixed. Major foreign media outlets analyzed that the slowdown in revenue growth, along with uncertainties stemming from the Trump administration's tariffs and restrictions, has put pressure on the stock price.
The gross profit margin, a key indicator of the company's profitability, has also decreased for three consecutive quarters. As the design and production processes for cutting-edge GPUs have become more complex, expenses related to research and development, design, and testing have surged. NVIDIA's gross profit margin for the fourth quarter of last year was 73%, down 3 percentage points from a year earlier. The forecast for adjusted gross profit margin in the first quarter of this year is also low at 71%, falling short of market expectations. Analysts suggest that concerns remain that NVIDIA's shift towards more complex and harder-to-produce products is leading to shrinking profits.
NVIDIA mentioned problems in the production process of Blackwell and announced plans to launch a new derivative model called 'Blackwell Ultra' starting in the second half of this year. CEO Jensen Huang said, "The technical issues that delayed the production of Blackwell have been completely resolved, and all new products are expected to be launched as planned." Huang explained that the basic usage method in systems remains the same, making the development of new products not particularly difficult.
NVIDIA expects its revenue for the first quarter (April-June) to reach around $43 billion. This is about 2% higher than the market research firm's estimate of $41.78 billion and represents a 65% year-over-year growth. However, the revenue growth rate for the quarter is showing signs of slowing down.
Concerns about the spread of low-cost AI models due to the emergence of DeepSeek remain. While it did not impact NVIDIA's performance in the fourth quarter of last year, there are observations that U.S. big tech companies may not actively pursue high-cost GPU purchases this year to reduce AI hardware investment expenses. An industry official noted, "U.S. big tech companies are incurring excessive costs in building AI infrastructure, which has been revealed through DeepSeek," adding that "this will become a significant threat to NVIDIA starting this year."
Colette Kress, NVIDIA's Chief Financial Officer (CFO), mentioned in response to a question about DeepSeek that "Long-term thinking and reasoning AI may require 100 times more computing power per task compared to a single inference." CEO Huang also stated that "most of our computing today is inference," predicting that next-generation AI algorithms will require millions of times the current computer capacity.
Meanwhile, the Trump administration's tariff pressures and strengthened restrictions on exports to China represent one of the risks to NVIDIA's performance this year. Bloomberg reported on the 25th (local time) that the Trump administration is considering further tightening semiconductor regulations and is looking to reduce the number of NVIDIA chips that can be exported to China without a permit.