Earlier this week, AI stocks had their worst day in recent memory after a new Chinese AI model, dubbed DeepSeek—a Chinese version of ChatGPT that is nearly as good but supposedly cost 95% less to develop—sparked fears that companies may not need to spend as much on creating new AI models as previously thought.
AI powerhouse Nvidia (NVDA) dropped more than 15%. A variety of other AI stocks—such as Nebius, Credo, Vistra (VST), Constellation Energy (CEG), Astera Labs (ALAB), Lumentum (LITE), Fabrinet (FN), and Ciena (CIEN)—plunged between 20% and 40% in a single day.
It felt like the AI bubble was bursting.
But it wasn’t. Instead, the DeepSeek drama earlier this week simply created a generational buying opportunity in top-tier AI stocks.
To understand why, we first have to break down what happened.
The Impact of DeepSeek’s Cost-Effective Model on AI Stocks
Tech stocks—and in particular, AI stocks—crashed after DeepSeek, a Chinese AI startup, released a new AI model that could represent a disruptive paradigm shift in foundational AI models.
The DeepSeek model functions similarly to ChatGPT. It is a chatbot with complex reasoning capabilities—and it is very good. The app has quickly become the most downloaded in the United States, and early users broadly believe it is as good as (if not slightly better than) ChatGPT.
But here’s the real kicker: DeepSeek is significantly cheaper than ChatGPT.
Training ChatGPT-4 reportedly cost about $80 million. Google’s (GOOGL) Gemini Ultra reportedly cost nearly $200 million. Broadly, incumbent foundational AI models in the U.S. have required $100 million or more to train.
DeepSeek, however, claims it trained its AI model for less than $6 million—despite being nearly as good as (or potentially better than) those other $100-million-plus AI models. That’s 95% cheaper.
The market panicked over this reported cost breakthrough, fearing that companies may not need to spend as much on AI model development in the coming years. That could mean less money flowing into AI infrastructure, lower spending by companies supporting that buildout, and, ultimately, lower stock prices for those firms.
Those fears are misplaced. In reality, the opposite will likely happen. Lower AI training costs will mean more AI spending.
That’s because AI is not your average commodity.
If you reduce the price of your average commodity — like eggs — you’ll reduce the total amount of money people spend on that commodity. If egg prices drop from $10 to two dollars, we’ll all spend less on eggs. Sure, some of us will buy more eggs. But the demand for eggs is pretty consistent. It won’t change dramatically with price. If egg prices go down, the total amount of money we spend on eggs every week will go down, too.
AI is not eggs.
AI is a high-demand and high-value commodity. When you have a high-demand and high-value commodity like AI, cost reductions tend to be met with quantity increases, and overall consumption goes up.
Jevons Paradox and the AI Spending Surge
This is not a novel idea. It’s an economic principle called “Jevons Paradox.”
Penned by 19th-century British economist William Stanley Jevons, Jevons Paradox states that improvements in the efficiency of resource use tend to increase — rather than decrease — the overall consumption of that resource because greater efficiency lowers the cost of using the resource, which can lead to increased demand for it.
This happened with coal usage in steam engines. Improvements in steam engine efficiency reduced coal consumption per unit of output. However, these improvements also made coal-powered technology more economically attractive, leading to broader adoption and ultimately increased overall coal consumption.
It happened with the internet. As computers got smaller and cheaper — and access to the internet became essentially free — pretty much everyone bought a computer and plugged into the internet 24/7.
And we believe it’s happening with AI right now. As AI model training and inference costs get slashed, more companies and people will build AI models, paving the path for AI to become a global ubiquity.
That means more — not less — AI spending.
More AI models mean more AI chips. More AI chips mean more AI compute. More AI compute means more AI energy. All that means more AI spending.
We aren’t alone in thinking this way about the DeepSeek developments.
Giant semiconductor equipment supplier ASML (ASML) reported earnings on Tuesday. In its quarterly conference call with analysts, ASML management said the cost efficiencies unlocked by DeepSeek would just open the door for more AI models to be created, which will lead to more AI chips, and more demand for the semiconductor equipment needed to make those chips.
They see the DeepSeek breakthrough only boosting the AI spending boom.
Microsoft (MSFT) and Meta (META) seem to agree.
They both reported earnings on Wednesday. In their quarterly conference calls, both Microsoft and Meta execs said the DeepSeek breakthrough wouldn’t change their AI spending plans. Both companies doubled down on their plans to spend tens of billions of dollars in 2025 on creating new AI infrastructure.
Semiconductor firm KLA Corp. (KLAC) said much the same in its quarterly earnings call this week. It believes the increased compute efficiency unlocked by DeepSeek will enable more adoption of AI due to “clearly elastic” demand for AI. Another important player in the AI infrastructure world — Celestica (CLS) — echoed similar thoughts. More efficient compute means more apps, which means more demand for the stuff that builds those apps.
The Final Word
The tech world has spoken. DeepSeek’s compute efficiency breakthroughs will be a net positive for the AI industry.
That, of course, means the huge drop we saw in AI stocks this week is nothing more than a golden buying opportunity.
We should be taking advantage of this opportunity right away.
To help us find maybe some of the best AI stocks to buy on this dip, we’re looking towards the world’s richest man – Elon Musk – and his big AI venture, xAI.
Click here to learn more about xAI now.
On the date of publication, Luke Lango did not have (either directly or indirectly) any positions in the securities mentioned in this article.
P.S. You can stay up to speed with Luke’s latest market analysis by reading our Daily Notes! Check out the latest issue on your Innovation Investor or Early Stage Investor subscriber site.