Nvidia’s boss dismissed the apprehensions that AI had hit a wall. Mint

When OpenAI’s boss Sam Altman posted a gnomic tweet this month, stating that “there is no wall”, his followers were very happy on the social-media site X. “Trump will make it,” one said. “No Pewall for ChatjPT?” The other took a pinch. Since then it has turned into a serious business case by a joke among the stupids.
The wall under consideration refers to the approach that the underlying forces of improvement in generative artificial intelligence (AI) have reached a limit over the last 15 years. Those forces are known as scaling laws. “There is a lot of debate on this: Have we created problems by cutting laws?” Microsoft boss Satya Nadella asked at his company’s annual conference on 19 November. A day later, Jensen Huang, the world’s most valuable company in the world, said that no.
Scaling laws are not physical laws. Like Moore’s rule, observation that processing performance for semi -activists doubles every two years, showing the assumption that AI performance in recent years has doubled every six months. The main reason for that progress is an increase in computing power that is used to train large language models (LLM). The fate of any company is not more associated with scaling laws compared to Nvidia, whose graphics processing units (GPUs) offer almost all computational omphs.
On 20 November, during the result presentation of Nvidia, Mr. Huaang defended the scaling laws. He also told The Economist that the first task of the latest class GPU, known as Blackwells, would be to train a new, more powerful generation of the model. He says, “It is very important for all these foundation-model-makers to run to the next level.”
The results of the NVidia quarter ended in October strengthened the spirit of speed upwards. Although the pace of growth has slowed down to some extent, its revenue has exceeded $ 35 billion, which is more than 94% year-on (see chart). And Nvidia estimated $ 37.5 billion in revenue for this quarter, more than Wall Street’s expectations. It states that the upward amendment was partially because he hoped that the demand for Blackwell GPU would be higher than before. Mr. Huang predicted that 100,000 Blackwells would be faster and would be employed to run the next generation of LLM.
Not everyone shares their optimism. Those who suspect scaling-law say Openai has not yet prepared a new general-purpose model to replace the GPT-4, which has outlined the Chatgpt since March 2023. He says that Gemini of Google is weak given the money spent on her.
But, as Mr. Huang says, scaling laws not only apply to early training of LLM, but also applied to the use of models or estimates, especially when complex arguments are involved. To explain the reason for this, he points to the latest model of Openai, O1, which has strong logic capabilities compared to GPT-4. This can perform advanced mathematics and other complex tasks by adopting a step-by-step approach, which its creator calls “thinking”. This advanced estimation process uses more computing power than a common chat GPT response, says Mr. Huang. “We know that whatever the approach, we need more calculations,” they say.
The more AI is adopted, the more important the conclusion will be. Mr. Huang says that the GPU of the previous generations of Nvidia can be used for estimates, but blackwells will make performance dozens of times. At least half of the infrastructure of Nvidia is already used to guess.
Sri Huang is clearly interested in painting the scaling in the best possible light. Some skeptical questions ask how meaningful progress in logic. Although some commercial models are being interrupted, many companies are struggling to adopt AI on a large scale, which can eventually affect the demand for technology. However, these are the early days. Tech giant continues to spend big on GPU, and Mr. Huang explains that it takes time to digest new technologies. Nvidia’s back is not yet adjacent to the wall.
© 2025, The Economist Newspaper Limited. All rights reserved. From The Economist, published under license. The original material can be found on www.economist.com
Hold everyone business News, Market News, Today’s latest news Events and Fresh news Update on live mint. Download Mint news app To get daily market updates.
MoreLess
(Tagstootransite) Nvidia (T) Nvidia (T) AI K Jensen Huang has rocked (T) SAM Altman (T) OpenAI (T) Generative Artificial Intelligence (T) AI Scaling Law (T) AI Scaling Law (T) Artificial Intelligence (T) NVDIA GPU (T) Blackwell GPU (T) Next generation of LLM (T) LLM training
#Nvidias #boss #dismissed #apprehensions #hit #wall #Mint