The release of DeepSeek R1 last week shocked the AI industry. At a fraction of the cost, it could match the performance of OpenAI o1. DeepSeek claimed that its R1 model took only two months and less than $6 million to create. This cost contrasts with that of OpenAI and other US-based LLM creators, which have already spent billions.
Did US AI companies overspend?
The release of this low-cost DeepSeek R1 model rattled the US stock market. The Nasdaq100, which includes many big US tech companies, crashed by around 3%. Chipmaker stocks were devastated, with a double-digit percentage drop. NVIDIA dropped by ~17% overnight.

What Happened?
DeepSeek has released an open-source artificial intelligence model that it claims took only two months and less than $6 million to develop. This is a massive breakthrough in the AI industry! So far, the high cost of developing an AI model has made it prohibitive for most players without significant funding to enter the market. This is why large tech companies dominate the AI landscape.
DeepSeek just disrupted this landscape, which is a massive win for the AI ecosystem. Competition is usually good for innovation. There will be winners and losers, but overall, this seems to be a positive development for the AI industry.
If this was positive news, why did the US market experience a significant drop afterward?
The answer could be the future expectation. US big tech players may have been overspending, and future adjustments may be needed to stay competitive.
Let’s look into some implications this can have on the key AI players today.
Possible Impact to Major AI Players
Other AI Model Players
The other AI model players, such as OpenAI and Anthropic, probably feel the most significant impact because their business model relies solely on the AI model itself. They have spent billions of dollars developing their models, but DeepSeek matches them with less than $6 million. Can they justify their spending?
Furthermore, they are not profitable yet and still rely on investors’ money to continue their development. Now, questions will be raised on whether those billions are needed.
Because their costs are high, they also charge users more to sustain their operations. On the other hand, DeepSeek is low-cost to operate and open source. Competition is heating up here, and the competitive moat that OpenAI was perceived to have is deteriorating.
In our opinion, AI models have slowly become commodities. This development is good for the AI industry as more and more utilizations can be created on top of these AI models.
Apart from OpenAI and Anthropic, Google and Meta are somewhat affected because they also develop their own AI models. However, because their primary business model is advertising, they can benefit from DeepSeek’s low-cost model. The AI-related capital expenditures they need may become smaller while still achieving comparable results.
Additionally, they can also adopt DeepSeek’s breakthrough into their own models.
Chipmakers
Generally, with such a low-cost model, the demand for expensive and high-end chips may be lower in the short term. Most AI players can cut their spending while still achieving their goals.
However, demand is likely to increase in the long run. This is widely known as the Jevons Paradox.
The Jevons paradox is an economic theory that states that when a resource becomes more efficient, its demand increases, which can lead to a rise in total consumption.
As AI models become more accessible and affordable, various applications will be developed, and many businesses will embrace AI technology, increasing the demand for these chips again.
AI Hyperscalers
AI Hyperscalers provide infrastructure and computational power to AI businesses. Some of the largest ones are Amazon, Google, and Microsoft. How is DeepSeek’s breakthrough possibly impacting them?
In the short term, the demand for AI computational power may be lower due to the lower cost offered by the DeepSeek model. In the long run, this may be counterbalanced by the ability to serve more customers with the same computational resources.
The main benefit for these hyperscalers is probably the lower CapEx required to serve the low-cost models. They do not need to invest too much in the more expensive and high-end hardware, thus lowering their capital expenditure.
AI Applications
DeepSeek’s affordable model will likely benefit all applications that utilize AI models. If the cost drops by 90-95%, AI will be more accessible to a broader audience, leading to more adoption and innovation.
Many big tech companies, such as Meta, Google, Microsoft, Salesforce, Adobe, etc., have been using AI models to supercharge their products and offerings. If costs can be significantly reduced, they can deliver much more at a much lower price. What a wonderful world that would be!
What Do We Do?
DeepSeek’s low-cost model is a breakthrough that should positively impact the AI ecosystem in the long run. We are still early in the AI revolution. More and more innovations will be introduced along the way, even when we least expect it.
In the long run, we still believe that AI will be an integral part of our lives, just like how the internet has become our way of life today.
As long-term investors, we should focus first on the long-term fundamentals. Because we still believe that the AI narrative is intact, we will stay invested in the fundamentally strong companies that contributed to this AI revolution. We will view short-term price fluctuations as an opportunity to add to our positions.
However, as previously mentioned, we may implement a higher discount rate when evaluating stock prices for certain market segments, such as chipmakers, to account for the uncertainty surrounding their demand after the introduction of DeepSeek’s low-cost model.
How about you? Did you make any moves in the market?
If you want to invest in US equities, you need to use an investment broker with access to the US market. You can follow our summary of some of the best online brokerage accounts in Singapore.