The US stock market experienced a bloodbath, with the Nasdaq Composite plunging 3%, fueled by the panic triggered by DeepSeek. This disruptive force threatens to upend the economics of the entire AI industry in the US. With chaos unfolding and many moving parts at play, it’s time to conduct a battle damage assessment. More importantly, many of you are wondering whether Big Tech are a buy, hold, or sell at this point?
Damage Limited to the ‘AI Efficiency Revolution’ Theme
While the Nasdaq dropped sharply, the Dow Jones managed to rise 0.65%. There were 2,810 gainers and 3,297 losers in the U.S. stock market, indicating that this wasn’t a broad market crash but rather a concentrated sell-off in a specific group of stocks.
The main casualty was unsurprisingly Nvidia, the previous most valuable company by market cap, which had the most to lose. Nvidia broke its own record with the largest single-day loss of $589 billion in market value.
It also lost its crown as the world’s most valuable company, falling two places to third. Nvidia’s market cap is now below $3 trillion.
DeepSeek doesn’t manufacture chips or GPUs, but it has upended the economics of AI computing by demonstrating that AI can operate 10 times cheaper, using significantly fewer GPUs and far less energy. This breakthrough permanently shifts the AI industry’s economics, and Nvidia, as the dominant player in GPUs, is expected to generate far less revenue in this new paradigm. (We’ve covered this on Finbite Insights and YouTube, so we won’t repeat it here.)
The fallout extended beyond Nvidia, impacting its entire supply chain and ecosystem:
- TSMC, which fabricates Nvidia’s chips, is expected to see fewer orders. Its stock dropped 13%.
- A slowdown in TSMC’s foundry expansion means less demand for chipmaking equipment. As a result, Applied Materials, ASML, KLA, and Lam Research saw declines of 7%, 6%, 6%, and 5%, respectively.
- ARM, which relies on chip royalties, faced a 10% decline due to reduced demand for GPUs.
- Dell, which produces server computers for data centers, dropped 9% as demand expectations fell.
- Even data center real estate wasn’t spared. Equinix, the world’s largest data center REIT, saw its stock fall 4%.
- Lower AI computing needs also hit electricity providers hard. Vistra, Talen Energy, and Constellation Energy plummeted 28%, 22%, and 21%, respectively.
The broader picture shows that the stocks impacted all belong to the “AI Efficiency Revolution” theme—those hurt by the push toward greater efficiency in AI computing.
DeepSeek Has Already Achieved Its Bannister Effect
The Bannister Effect, named after Roger Bannister breaking the four-minute mile, describes how achieving an impossible feat inspires others to follow suit. DeepSeek has done exactly that in the realm of AI computing efficiency.
Before DeepSeek, the prevailing belief was that advancing AI required packing more power into the infrastructure—more GPUs, more energy, more cost. DeepSeek shattered this notion by demonstrating a much more sustainable path. The result? A structural shift in the AI industry. Whether or not DeepSeek can beat ChatGPT is irrelevant; the efficiency revolution it sparked is here to stay. Other AI programs will inevitably pursue this path, fundamentally altering the demand and growth trajectory for AI chips and energy.
This doesn’t mean Nvidia is obsolete—not by a long shot. Nvidia remains the leading GPU provider for AI and will continue to dominate for the foreseeable future. However, the growth story has changed. With efficiency now the industry’s focus, Nvidia’s growth rate has fallen. As a result, its valuation and share price must adjust accordingly. How much? That’s anyone’s guess. What’s clear to us is that Nvidia no longer offers an attractive risk-reward profile at this stage.
We firmly believe this is a structural trend, not a passing phase. It’s not about DeepSeek being “better” but about the entire AI industry embracing efficiency. The “Genie of Efficiency” is out of the bottle, and there’s no putting it back.
There will be beneficiaries
Not all AI-related stocks are set to suffer in this shift. We believe the cloud giants are a key group poised to benefit. We’ve stated this before, and we stand by this position, even as Microsoft and Alphabet saw their share prices fall 2% and 4%, respectively. Meanwhile, Amazon managed to eke out a 0.2% gain.
The reason is simple: higher efficiency reduces costs. Cloud giants won’t need to purchase as many of Nvidia’s expensive chips, which have been a major driver of their rising capital expenditure. As efficiency improves, we can reasonably expect their profitability to increase as well.
There’s another side to this story: cheaper AI computing costs could expand the market. Enterprises that previously avoided AI adoption due to high costs will now find it more accessible and less risky to experiment with AI solutions. This shift could significantly broaden the customer base for cloud services, creating new growth opportunities for the giants.
We are living in exciting times. The AI industry has just taken a major turn, and it’s clear this is only the beginning. Expect more surprises to unfold.
I want to add:
1. Many pointed out that increased efficiency will actually drive higher demand. This aligns with the classic demand curve, where lower prices lead to higher volumes. Microsoft CEO Satya Nadella also referenced the Jevons Effect in this context.
2. Previously, I commented that increased efficiency would reduce usage, thereby lowering demand for Nvidia chips. However, I realize I didn’t fully elaborate on the context behind that view.
3. I maintain my view that Nvidia’s stock faces a higher risk of downside, and here’s my refined reasoning.
4. The key caveat is that demand must grow at a pace that exceeds the improvements in efficiency to sustain higher demand for Nvidia chips.
5. Another mental model that might apply here is the hype cycle. Every new technology experiences a peak of inflated expectations before crashing into the trough of disillusionment.
6. Over the past two years, we’ve seen the generative AI craze, accompanied by cloud giants rapidly building out their infrastructure. However, this has likely resulted in overcapacity. Anecdotally, when I surveyed people around me, very few have meaningfully incorporated AI into their workflows.
7. With improved efficiency, this overcapacity issue could worsen. As a result, cloud providers may not feel incentivized to continue stocking up on Nvidia chips.
8. A similar situation unfolded with Amazon during COVID. The company aggressively built warehouses, only to realize it had overcapacity. This led to warehouse closures and a $10 billion write-down, causing the stock to slump temporarily. Eventually, the business recovered, and so did the share price.
9. Something similar could be happening with Nvidia GPUs. The race to acquire these limited chips may have outpaced end-user adoption. This is extremely hard to estimate, and by the time the market acknowledges the overcapacity, it could already be too late.
10. Nvidia’s results for the next two quarters are likely to remain strong, as these will reflect pre-DeepSeek conditions and continued purchases by hyperscalers.
11. However, the following quarters will be the true test to determine whether demand wanes. If it does, it would indicate excess capacity, which could negatively impact Nvidia’s valuation.
12. Correcting this overcapacity could take 1–2 years. By then, AI developments are likely to stabilize along the slope of enlightenment, leading to more sustainable growth.
13. To be clear, I am not against AI—I use LLMs daily and firmly believe in its long-term potential. However, we may face a significant mid-term correction due to overcapacity, and improved efficiency could accelerate this issue.