Nvidia to report earnings in the midst of spending on infrastructure, deepseek concerns

Nvidia It is planned to report on financial results in the fourth quarter on Wednesday after the bell.

It is expected to put the final details in one of the most prominent years from the big company ever. Analysts who have surveyed FactSet expect $ 38 billion for a quarter of a quarter ended in January, which would be an increase for 72% on an annual basis.

The January quarter will take away the second fiscal year in which Nvidia sells more than doubled. It is a wonderful string guided by the fact that the Processing of the NVIDIA data processing units for the processing of graphics or GPUs are essential hardware for the construction and arrangement of artificial intelligence services such as Openai’s Chatgpt. In the last two years, the Nvidia section has increased 478%, making it the most respected US company over time with a market cap over $ 3 trillion.

But Nvidia’s supplies have slowed down in recent months, as investors have asked a question where Chip Company can get out of here.

Trafficking at the same price as last October, and investors are cautious about any signs that the most important NVIDIA customers could lift the belts after years of large capital expenditures. This is especially true of recent breakthroughs in AI from China.

Much of Nvidia sells goes to several companies that build a huge server farm, usually to rent to other companies. These cloud companies are usually called “hyperscalers”. Last February Nvidia said that one buyer made 19% of his total revenue in the fiscal 2024.

Morgan Stanley analysts estimated this month Microsoft It will make almost 35% of consumption in 2025. On Blackwell, the latest AI chip in Nvidia. Google is at 32.2%, Prophet at 7.4% and Amazon to 6.2%.

Because of this, it is any sign that Microsoft or his rivals could withdraw the consumption plans, he can shake off Nvidia’s supplies.

Last week, TD Cowen analysts said they learned that Microsoft had canceled tenants with private data operators, he slowed down his negotiation procedure for entering new tenants and adjusted plans for spending to international data centers in favor of US buildings.

The report raised the fear of growth of growth of the AI ​​infrastructure. This could mean less demand for Nvidia’s chips. Michael Elias TD Cowen said the finding of his team indicates the “potential position of excessive offer” for Microsoft. Nvidia shares on Friday fell 4%.

Microsoft pushed out on Monday, saying that he was still planning to spend $ 80 billion on 2025 infrastructure.

“Although in some areas we can strategically interpret or adjust our infrastructure, we will continue to grow strongly in all regions. This allows us to invest and distribute resources for growth areas for our future,” said CNBC spokesman.

During the last monthMost Nvidia key customers said big investments. The alphabet goal $ 75 billion In capital expenditures this year, Target will consume as much as $ 65 billion And Amazon is aimed at spending $ 100 billion.

Analysts say that half of the capital expenditures AI infrastructure ends with Nvidia. Many hyperscalers deal with AMD GPU and develop their own AI chips to reduce their addiction to NVIDIA, but the company holds most of the market for top-notch AI chips.

So far, these chips have been primarily used to train the New AI model, a process that can cost hundreds of millions of dollars. After AI develops companies such as Openi, Google and Anthropic, the NVIDIA GPU warehouses are needed to serve those models to customers. That is why Nvidia projects her income to continue to grow.

Another challenge for Nvidia is last year’s appearance of Chinese startup Deepseek, which has published effective and “distilled“Ai model. It had high enough performance suggesting that the billions of dollars of NVIDIA GPU did not need to be trained and used by top AI. This temporarily sank Nvidia shares, which made the company lost almost $ 600 billion in market boundaries.

NVIDIA Executive Director Jensen Huang will have the opportunity to explain why AI will still need more GPU capacity, even after last year’s mass construction.

Recently Huang talked about “Law on Scaling”, perception Since Openai 2020, for AI models to improve, more data and calculations are used when creating.

Huang said the Deepseek model R1 indicates a new pine in the Scalcling Act that Nvidia calls “Scaling time of testing. “Huang argued that the next large path to improvement of AI is the application of more GPUs to the process of implementing AI or conclusion. This allows Chatbot” Reason “or generate a lot of information in the process of thinking through the problem.

Ai models are only trained several times to create them precise. But AI models can be called millions of times a month, so the use of more calculating at the conclusion will require more Nvidia chips distributed to customers.

“The market responded to R1 as in:” Oh my God, and it’s done, “that Ai no longer needs to be counted anymore,” Huang said in the overwhelmed Interview last week. “That’s just the opposite.”