3. NVIDIA Corp (NASDAQ:NVDA)
Number of Hedge Fund Investors: 179
Barclays Tom O’Malley recently said in a note that spending on AI chips must rise to meet even modest expectations from large language model (LLM) developers. The analyst noted that the recent sell-off in AI chip stocks, especially after NVIDIA Corp (NASDAQ:NVDA) earnings, has raised concerns about whether the market is nearing a peak. However, O’Malley emphasized that these fears overlook the significant compute needs required to power future AI models.
Barclays identified three key takeaways for the semiconductor sector:
First, the need for compute resources is expected to surge. For instance, by 2027, just three frontier models with 50 trillion parameters each could require around 20 million chips for training alone. The demand for compute is anticipated to outpace improvements in chip performance.
Second, a dual approach to AI accelerators is emerging. NVIDIA Corp (NASDAQ:NVDA), AMD, and startups are expected to dominate in training and inference for frontier models, while hyperscale companies will rely on custom silicon for specialized tasks.
Third, inference, where AI models apply learned knowledge, will play a significant role in monetizing frontier models. NVIDIA Corp (NASDAQ:NVDA) claims that 40% of its chips are already used for inference, with more providers focusing on this market to drive down costs and improve return on investment.
O’Malley concluded that the AI chip industry is still in its early stages, and higher investments in compute resources are critical to meet future demands.
Alger Spectra Fund stated the following regarding NVIDIA Corporation (NASDAQ:NVDA) in its Q2 2024 investor letter:
“NVIDIA Corporation (NASDAQ:NVDA) is a leading supplier of graphics processing units (GPUs) for a variety of end markets, such as gaming, PCs, data centers, virtual reality, and high-performance computing. The company is leading in most secular growth categories in computing, and especially artificial intelligence and super-computing parallel processing techniques for solving complex computational problems. Simply put, Nvidia’s computational power is a critical enabler of AI and therefore essential to AI adoption, in our view. During the quarter, the company reported better-than-expected fiscal first quarter results driven by strong demand from data centers. Additionally, management noted that large cloud service providers, contributing approximately 45% of data center sales, recognize the high return on investment offered by Nvidia’s computing solutions, which are driving AI spending. The company also introduced its next-generation H200 chip, which nearly doubles the inference performance compared to the H100 chip, enhancing how trained AI models process new data. Lastly, management raised their fiscal second quarter guidance, noting that demand for their current H100 chips remains strong, and that demand for their next generation products is estimated to outstrip supply over the next year. We continue to believe the company is well positioned to potentially benefit from the growing AI data center workloads, which are driving demand for the increased interconnections and fully accelerated software stacks, thereby enabling leading application performance and fast result times.”
