In January, the Chinese start-up Deepseek introduced an artificial intelligence chatbot (AI) which quickly became the most downloaded free mobile application in the United States, the basic basic model behind the Deepseek application would have cost $ 6 million to train and exceed the United States. Models on certain benchmarks.
This rocked Wall Street because Openai spent more than $ 100 million in the training of its GPT-4 model and would have had access to more advanced fleas than Deepseek. Investors rushed to the conclusion that US companies have exceeded too much for AI infrastructure, and Nvidia (Nvda 0.90%)) The shares fell 17% on news, erase $ 600 billion in market value. It is the greatest loss of a day in American history.
NVIDIA shares have been down approximately 9% since Deepseek rocked the market, and the action is currently 13% reduction on its record level. But Nvidia shareholders have just received good news from Amazon (Amzn -4.05%)) And Alphabet (Googl -3.27%)). Read the rest to find out more.
Some Wall Street analysts are skeptical about the Deepseek story
Many analysts have praised Deepseek for his engineering breakthroughs. The company has reduced expenditure by reducing data processing with innovative training techniques. However, several analysts also wonder if Deepseek has been completely to come on its infrastructure and the associated costs.
For example, Dan Ives at Wedbush Securities wrote: “Saying that Deepseek was built for $ 6 million without the next generation equipment of Nvidia being probably a fictitious story.” Likewise, the semianalysis of the research company reports that Deepseek not only had new generation NVIDIA GPUs, but also hired around $ 1.6 billion in spending during the formation of its model.
Above all, savings could actually increase Nvidia’s demand Graphic processing units (GPU) by allowing faster diffusion of Artificial Intelligence (AI) through the economy. In other words, if new training methods reduce the development costs of the model, more software companies will create AI products, increasing the total demand from GPU Nvidia.
Image source: Getty Images.
Amazon and Google plan to invest even more aggressively in the IA infrastructure
Last week, Amazon and Google Parent Alphabet announced financial results in the fourth quarter. The two companies said Capital expenditure would increase considerably in 2025, as they are currently linked to supply with regard to AI infrastructure.
The CEO of Amazon, Andy Jassy, predicted that the costs associated with the development of AI models will continue to fall. But he also said: “I think it will make companies easier to infuse all their applications with inference and with AI Generative. “And financial director Brian Ossavsky told analysts that capital expenses could exceed $ 100 billion in 2025, compared to $ 83 billion in 2024, due to AI infrastructure demand.
The CEO of Alphabet Sundar Pichai provided similar information. He said that a higher percentage of capital expenses are moving towards inference training becomes cheaper, which implies the breakthroughs of Deepseek would not slow the expenses of the AI. Indeed, the financial director Ana Ashkenazi told analysts that capital expenses would total $ 75 billion in 2025, compared to $ 52 billion in 2024, mainly due to investments in data centers, servers and networks.
Capital expenditure directives of Amazon and Google Parent Alphabet corroborate the account that the demand for NVIDIA GPU could in fact increase as the training costs decrease because companies will transfer their resources to inference. Andy Jassy compared the situation to the way Cloud Computing sales have really increased over time despite cheaper cloud services.
“What is happening is that companies will spend much less per unit of infrastructure, and it is very, very useful for their businesses. But then, they are enthusiastic about what they could build that were always prohibitive Previously, and they generally end up spending much more in total technology once you make the unit.
Wall Street always expects Nvidia’s revenues to increase very quickly
Wall Street estimates that NVIDIA’s adjusted profits will increase to 52% per year during the year 2026, which ends in January 2026. This consensus makes the current valuation of 50 times the adjusted income is cheap. Admittedly, analysts have established a very high bar, so that the stock could drop sharply if Nvidia is missing these expectations.
Here is the essential: NVIDIA shareholders worried about the Deepseek drama can breathe deeply. Many analysts believe that more effective AI training techniques will lead to greater demand for NVIDIA GPU. And Proof of the largest hyperscal cloud companies supports this conclusion.
John Mackey, former CEO of Whole Foods Market, a subsidiary of Amazon, is a member of the board of directors of Motley Fool’s. Suzanne Frey, director of Alphabet, is a member of the board of directors of Motley Fool’s. Trevor Jennewine has positions in Amazon and Nvidia. The Motley Fool has positions and recommends Alphabet, Amazon and Nvidia. The Word’s madman has a Disclosure policy.