Analysts have identified a potential buying opportunity for NVIDIA following a recent dip in AI momentum.
Hwang Byeong Jun, a researcher at Yuanta Securities, noted that the current weakness in stock prices stems from underperformance among front-end customers and market disappointment regarding anticipated delays in the rollout of the upcoming Blackwell architecture chips, expected in the fourth quarter. He views this as a short-term concern. Investor anxiety surrounding consistent demand for high-performance semiconductors and inadequate AI monetization has negatively affected the company’s stock. Nevertheless, NVIDIA’s capital expenditures have consistently surpassed expectations since the second quarter of last year.
This increased spending reflects a robust demand for transitioning proprietary server data to the cloud, alongside the need for advanced semiconductor-based computing infrastructure to support AI model training and data analysis.
Demand for the structural AI infrastructure provided by cloud service providers (CSPs) is projected to endure. This infrastructure is pivotal for automating corporate functions and developing robotics and autonomous vehicles, including the synthetic data generation essential for training AI models.
Hwang predicts that, as NVIDIA CEO Jensen Huang has indicated, demand for AI chips will remain strong. This is driven by the escalating volume of data needed for model refinement, coupled with the growing demand for enhanced computing capabilities and cost efficiency.
The resurgence in front-end demand suggests that corporate growth momentum is on the cusp of recovery. Despite supply chain disruptions affecting the B series, momentum is expected to remain subdued until the first quarter of next year. Nonetheless, NVIDIA is positioned to thrive.
In July, the company increased production by 25% compared to the initial B series order volumes projected earlier in the year. Additionally, total shipments of the GB200 NVL72 and NVL36 models have reached 60,000 units, exceeding expectations of 40,000 units by 50%.
The unit price of the GB200 NVL36, which integrates 18 GB200s, is reported at $1.8 million. The projected revenue from NVL36 alone aligns with the upper end of last year’s guidance for H100 sales in 2024, which was set between 150,000 and 200,000 units.
Hwang attributes the recent slowdown in growth to anticipated demand for the B series from major players such as Amazon, Microsoft, and Meta. Additionally, production constraints at TSMC have contributed to these challenges. Advanced multi-chip packaging technology, CoWoS, is expected to be employed in the B series.
He explained that production capacity is forecasted to double this year compared to 2023, with a further increase of 70% expected next year. Thus, the ongoing supply chain issues may represent a strategic buying opportunity for investors.
Most Commented