The landscape of artificial intelligence is starkly divided. OpenAI has withdrawn the release of a new model that became uncontrollable, even with a kill switch, lending credence to the notion of 'AI speed control.' In contrast, Nvidia has approved a massive $235 billion stock buyback plan, showcasing confidence that defies market concerns.
Nvidia's rationale is straightforward: its current stock price is undervalued. The 12-month forward price-to-earnings ratio stands at approximately 16.5, significantly lower than the 30 average of the past 15 years. Despite consistently exceeding market expectations each quarter, Nvidia's stock has not kept pace, largely due to skepticism surrounding 'AI peak-out' and investments in large data centers.
Apple has faced a similar situation. While its earnings have increased, growth expectations in the mature smartphone market remain stagnant, leading to a decline in stock prices. In May 2024, Apple initiated a $110 billion stock buyback, maintaining its dominance in the smartphone sector.
Nvidia's decision should not be viewed merely as a defensive move for its stock price. The demand for AI infrastructure from big tech companies represents a sustainable 'platform transition' in the medium to long term. As CEO Jensen Huang stated, we are in a once-in-a-generation acceleration of computing transformation. The substantial investments in accelerators and GPUs show no signs of slowing down, alleviating stock pressure in the AI value chain and continuing infrastructure investments.
Unlike OpenAI, which focuses on services, Nvidia is confident in its infrastructure role. The safety issues and malfunctions associated with AI agents stem from applications and algorithms. As models become more advanced, the complexity of understanding and realizing human intent increases exponentially, leading to technical challenges that existing algorithms cannot resolve.
As the competition intensifies, leading firms like OpenAI and Anthropic have raised the issue of speed control. However, even if the release of new models is delayed, competition will persist. The demand for infrastructure is heating up due to rival companies' pursuits, global corporations building their own large language models, and various countries' 'sovereign AI' projects.
A temporary setback in model research does not equate to a decrease in hardware demand. The immense computing power required for safety testing is still consumed on Nvidia GPUs.
The $235 billion stock buyback signals that AI is moving beyond expectations into practical commercialization and mass consumption. Issues of safety and kill switches are growing pains on the path to more advanced AI. The shift toward accelerated computing cannot be halted. Nvidia's boldness in maintaining its hardware dominance reflects a calm and calculated confidence rooted in the prevailing trends of the AI infrastructure market.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.