For the semiconductor industry, Moore’s Law has long been a guiding principle. This observation—that chip performance improves as transistor density doubles every two years—has shaped innovation for decades. Now, however, it is quickly being replaced by the “scaling law” of artificial intelligence, according to an exclusive report from Financial Times.
Don't Miss our Black Friday Offers:
- Unlock your investing potential with TipRanks Premium - Now At 40% OFF!
- Make smarter investments with weekly expert stock picks from the Smart Investor Newsletter
While the scaling law has driven AI advancements, this shift could spell trouble for chipmaking giants like Nvidia (NVDA), FT reports.
What Is the “Scaling Law” of Artificial Intelligence?
At its core, the scaling law holds that feeding larger datasets into bigger AI models leads to smarter systems, but this requires increasingly powerful computing resources. The launch of ChatGPT last year amplified this approach, sparking rapid advancements in AI.
However, recent developments indicate the scaling law may be reaching its limits. OpenAI’s founder Ilya Sutskever remarked, “The 2010s were the age of scaling, now we’re back in the age of wonder and discovery once again.” This shift highlights a growing need for innovation beyond just bigger models.
Why Is the Scaling Law Losing Steam?
Initially, the scaling law primarily applied to “pre-training,” the foundation of large AI models. Yet, as venture capitalist Marc Andreessen noted on a recent podcast, the capabilities of these models are “topping out” during pre-training, requiring more work to maintain progress.
Nvidia’s CEO, Jensen Huang, acknowledged this shift on the company’searnings call explaining that while scaling remains important, “test time scaling” during inference—AI’s ability to “think” when responding—has become critical. This shift could still drive demand for Nvidia’s chips due to their need for higher computing power.
Increasingly, Big Tech companies are pivoting from building larger AI models to enhancing their practical applications. Microsoft’s (MSFT) Brad Smith predicts that AI infrastructure demand will evolve in line with more pragmatic uses of AI, reflecting a move toward efficiency over sheer scale.
Is NVDA a Good Stock to Buy?
Analysts remain bullish about NVDA stock, with a Strong Buy consensus rating based on 40 Buys and four Holds. Over the past year, NVDA has skyrocketed by more than 100%, and the average NVDA price target of $175 implies an upside potential of 25.9% from current levels.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
