Probably, there was no stock that was severely influenced by the news from DeepSeek as NVIDIA (NVDA -3.67 %)。 In a way, DeepSeek has verified its advantage by announcing that the H800 accelerator trained its model for less than $ 6 million. The cost is only a few of the hundreds of dollars that Openai spent on Chatgpt’s development costs.
However, the news questions the demand for NVIDIA accelerators. Low -cost AI is more likely to increase the demand for technology and may be useful for NVIDIA.
Nevertheless, if the entity can build an AI model with a cheap technology, it will probably have little need to pay a premium price to the state -of -the -art AI chip.
Knowing that, investors may be wondering how to invest in NVIDIA shares in the future. In this article, three Motley Fool contributors share their opinions on NVIDIA and what investors should do.
Long -term investors would be wise to adjust the noise of deep seek.
Jake Lelch: If it comes In my opinion about NVIDIA, I have not been moved to the recent Deepseek AI presentation.
beginning AllI am receiving news with a healthy spoon positive. Certainly, some AI experts have praise It is pointed out that the released open source code is an impressive and important breakthrough.
However, there are doubts about the use of NVIDIAGPU and several numbers to train this new model. But keep in mind that the US government has exported to the NVIDIA GPU. in particular To prevent Chinese profits from accessing them.
In other words, don’t stop your breath after waiting for the complete transparency of how DeepSeek was built. – And for what purpose.
In any case, the status of NVIDIA as a major horse in the race to design the best AI chips is the same. Furthermore, the next step of the AI Revolution is not about who will build the following: wonderful Large language model (LLM).
Instead, investors should do I will continue to concentrate In the overall picture, tools and applications with AI are ready to improve actual productivity in the economy as a whole. Logistics, healthcare, finance, and all other sectors create more sales, more Profit as an AI revolution Roll on。
In addition, this breakthrough does not reduce the need for a better and faster chips. After all, if the Technical Revolution for the past 40 years taught us what That one Nobody wants to stick to the old -fashioned technology. People, companies, and governments are always eager to pay above For a newer and faster calculation ability.
Remember: The world was not satisfied with the dial -up Internet, and we won’t do so satisfaction With NVIDIA A100sEither。 The company will continue to innovate with improved chips such as the latest The Blackwell. and That’s my belief These chips provide innovation that surpasses DeepSeek.
NVIDIA’s shares have just entered the idea of decreasing AI expenditures.
Justin Pope: It was a very stressful week for NVIDIA shareholders. The shares have dropped in two digits after DeepSeek, a Chinese AI company, recently sent Wall Street Scrambling. DeepSeek offers Openai’s performance to Openai, while the AI model is open source (the developers can use it freely), and argues that there are very few training and operation.
More specifically, DeepSeek argues that the V3 (chat) AI model was used using a 2,048 NVIDIA H800 GPU cluster. This is a surprising result, considering that tens of millions of AI companies, such as Openai and Google (owned by Alphabet) spent on NVIDIA’s excellent H100 chips. The company has a software engineering praising the results and is lying efficiently, not hardware to provide necessary performances.
The fear is that as other companies duplicate Deepseek’s technique, NVIDIA’s GPU sales may cool down. NVIDIA depends on a small number of AI hyper scallers for a considerable part of its revenue.
Multiple industry experts are very skeptical about $ 5.5 million training costs. NVIDIA claims that DeepSeek is working in the chip export restricted parameter, but there are questions about whether the numbers have excluded additional costs or whether the model has really been developed from zero. 。 However, Deepseek’s efficiency breakthrough seems to be plausible. Because the model is open source, anyone can see how the company has done and perform tests. In addition, DeepSeek has severely reduced the price setting of Openai. The R1 (reasoning) model API output is only $ 2.19 per million tokens, compared to $ 60 for Chatgpt O1.
The bottom line of NVIDIA is probably too early to respond. Technology companies like Microsoft are not yet able to support AI demand, so it is unlikely that these large -scale AI investment will drop overnight. The business remains powerful for NVIDIA. However, over time, if the efficiency is helpful to extend the life of existing AI chips through software engineering, the demand for GPUs may be stopped by seeking cost efficiency.
It definitely makes NVIDIA a more risky inventory than a few weeks ago. It is one of the best AI stocks that can still be purchased, but to minimize the risk, consider the average and diversification of portfolios.
Reduces enthusiasm for this AI accelerator stock
Will Health: The issue of investing in NVIDIA after Deepseek News is one attribute, a competitive advantage.
NVIDIA’s finances and shares have been burning since the spring of 2023 when Openai has driven NVIDIA AI Accelerators driving the GPT-4. Neither of the NVIDIA competitors to use the era of the era. The CUDA software corresponding to the NVIDIA accelerator is far ahead of a competitive product.
In fact, the demand for state -of -the -art accelerators is unlikely to disappear, and if analysts delve into DeepSeek’s breakthrough details, the NVIDIA’s competitive advantage remains the same. In addition, DeepSeek did not train models away from the NVIDIA universe.
However, due to high demand, the price of the state -of -the -art H200 accelerator exceeds $ 30,000. If there is demand in the top autumn, you can reduce the price of this technology to NVIDIA.
This is because the data center segment, which designed the accelerator, was 87 % of the first nine -month corporate revenue in the 2025 fiscal year (ended on October 27, 2024). In that period, $ 91 billion profits increased 135 % year -on -year.
Nevertheless, if this latest development reduces the price setting force, it can mean that the revenue per unit will decrease significantly.
Furthermore, according to the evaluation, NVIDIA’s stock may have a long way to fall. Many investors can overlook this problem because the ratio of 47 P/E is definitely low in stock. Unfortunately, the price of 46 prices and books far exceeded multiples of AMD’s book value 3. Investors will question multiple book value in environments that slow the growth of revenue.
In addition, NVIDIA has decreased by more than 50 % since 2018, reminiscent that NVIDIA’s momentum can change negatively and quickly.
NVDA data by YCharts
It is too early to assume that DeepSeek’s breakthrow will bring such a different decrease in stock prices, and even if it seems, NVIDIA is unlikely to lose its overall competitive advantage. Still, there are considerable drawbacks, so it may not be a good time to keep a large position in NVIDIA stocks.
