*Investors are shaken by the latest update from the tech giant, Nvidia. The company is set to announce its fourth-quarter financial results after the bell on Wednesday, capping off a remarkable year. Analysts expect sales of $38 billion for the quarter ended in January, a 72% increase year-over-year. Nvidia’s sales have more than doubled in the past two fiscal years, driven by the demand for its data center GPUs for AI services like OpenAI’s ChatGPT. Despite a 478% rise in stock value in the past two years, Nvidia’s stock has slowed amid concerns about the company’s future prospects. Investors are cautious about signs that Nvidia’s key customers, such as hyperscalers like Microsoft and Google, may reduce spending on infrastructure, impacting demand for Nvidia’s chips. Recent reports about Microsoft scaling back infrastructure plans have further fueled these concerns. However, Microsoft clarified that it still intends to spend $80 billion on infrastructure by 2025. While some hyperscalers are exploring alternatives like AMD’s GPUs or developing their own AI chips, Nvidia remains a dominant player in the AI chip market. The company’s GPUs are crucial for training advanced AI models, with Nvidia holding a significant share of AI infrastructure capital expenditures. Despite some challenges, Nvidia continues to be a key player in the evolving AI landscape.*
Nvidia anticipates that its revenue will continue to grow steadily. However, a recent challenge for the company arose with the emergence of the Chinese startup DeepSeek, which unveiled an efficient and streamlined AI model last month. This model showed high performance levels that implied Nvidia’s GPUs worth billions of dollars may not be essential for training and using cutting-edge AI technology. Consequently, Nvidia’s stock experienced a temporary decline, leading to a substantial loss of almost $600 billion in market capitalization for the company.
Nvidia’s CEO, Jensen Huang, is set to address this issue on Wednesday, explaining why the demand for GPU capacity in AI is expected to rise despite a significant expansion in the previous year. Huang has recently discussed the concept of the “scaling law,” which was identified by OpenAI in 2020. This theory suggests that AI models improve as more data and computational power are employed during their creation.
Huang highlighted DeepSeek’s R1 model as a new aspect of the scaling law, termed “Test Time Scaling” by Nvidia. He argues that enhancing AI will now require deploying more GPUs for inference, enabling chatbots to engage in reasoning and generate substantial data while problem-solving. While AI models are typically trained a few times for refinement, they can be utilized millions of times monthly, necessitating increased computational resources, specifically Nvidia chips, for customers.
In response to the implications of DeepSeek’s R1 model, which initially suggested that AI no longer requires extensive computing, Huang emphasized in a recent interview that the exact opposite is true. He believes that the key to advancing AI further lies in applying more GPUs during the deployment and inference processes, contrary to the initial market reaction that signaled the end of AI’s computing era.
