Once in a while, narrative-changing information comes to the market.
That happened earlier this week with the DeepSeek news – if the claims are true.
Key Takeaways:
- Market Disruption: DeepSeek’s R1 model introduces a disruptive method, potentially eliminating the first-mover advantage in AI by significantly lowering compute costs and resource requirements.
- Cost Dynamics and Adoption: The drastic reduction in AI costs could accelerate adoption and shift value from infrastructure to application layers.
- Short- vs. Long-Term Impacts: In the short term, overbuilt AI infrastructure may face challenges, but long-term trends like the Jevon paradox suggest that increased efficiency will drive greater resource consumption and benefit companies like Nvidia.
- Strategic Repositioning: Megacap cloud companies remain well-positioned to capitalize on these shifts, while the AI investment narrative adjusts to new realities.
It was a historic moment for the stock market, as the NASDAQ slid 3.1% in one day, with Nvidia being worth $550B less than it was on Friday.
The AI investment cycle got thrown for a loop over the span of a weekend, all because of one mysterious company's claims. While this is definitely a near-term cause for concern, we need to ask ourselves several questions here.
First, are the claims substantiated? Second, what will this mean for the AI ecosystem? Third, what does this mean for AI stocks short term vs. long term?
Let’s break this down into three sections:
- Outlining what happened here,
- Walk through the “story” the market told itself
- The implications of it all, and
- How this will affect semiconductor companies in particular
By now, you've likely had time to digest some of the moves, so I wanted to take more time to give a sober view of where to go from here.
Background: What HAPPENED?!
DeepSeek is a Chinese AI lab founded in 2023 and headquartered in Hangzhou. Initially, DeepSeek was part of the quantitative hedge fund High-Flyer, which has an estimated $7B in AUM.
Last week, DeepSeek released R1, an open-source model which performs on par with leading models but allegedly uses less than 5% of the compute resources at 10% of the cost of leading models like ChatGPT-o1. DeepSeek-R1 matches OpenAI's GPT-o1 in math, coding, and reasoning benchmarks while costing 95% less. For example, o1 charges $15 per million input tokens compared to R1’s $0.55.
There were many skeptics of this company's previously released models because certain models can "game the system" by pre-training on certain data that lets them excel at benchmarks.
What we figured out over the weekend is that DeepSeek is very likely the real deal.
What we still need to figure out is whether DeepSeek had access to chips that they shouldn't have had and if their cost figures are actually accurate. Either way, the method that DeepSeek is now using will fundamentally disrupt the AI ecosystem.
DeepSeek uses a process called “distillation,” which essentially allows them to back into exactly how OpenAI built their model, copy it, and then open-source it. What you get after this is a massive first-mover disadvantage, which throws the whole incentive structure for building $100B+ datacentres in the trash.
This is where the market started running off with the narrative.
The Story the Market Told Itself
On Monday, the market was gripped with fear. And rightly so… this is a "WTF?" moment.
I will first outline how the market read this headline. But don't stop reading there because the truth of what we need to expect will lie somewhere in the middle of hope and despair.
The market narrative went like this:
- There is no longer a first-mover advantage to building frontier LLM solutions because someone can take the model, repurpose it, and provide the same functionality at a fraction of the cost.
- Therefore, spending hundreds of billions of dollars on CapEx is pointless because of this first-mover disadvantage.
- Therefore, we have drastically overbuilt CapEx and this will dramatically halt.
- Therefore, we don't need more semiconductors to build out more datacentres, nor do we need the racking systems and cooling systems to support them.
- Therefore, we don't need upgraded energy infrastructure because there will be no increase in energy demand because this new way of AI is much more energy efficient.
- However, since software companies no longer need to spend $$$ on these models as they get commoditized, their cost of goods sold (COGS) significantly falls, so software companies will benefit.
Following this narrative, the most sold-off companies of the day were energy companies like Vistra, Constellation, NRG, and Talen. These companies sold off 25%+. Cooling systems like Vertiv also sold off the same 25%+. Next, semis sold off 15-20%, datacentre REITs sold off 5-10%, while software companies were UP on the day. The remainder of the megacaps were mixed, with AAPL +4%, Google -4%, Microsoft -2%, Amazon flat, Tesla -2%, and Meta +2%.
Now that we can take a breath, let’s see what's next.
Where to Go from Here?
The reality is that the goalposts might have just significantly moved. We can no longer blindly spend hundreds of billions on CapEx just to stay ahead because "staying ahead" might not be as valuable anymore.
This will affect every AI company. But it will affect them differently.
One of the core impacts of this shift is that the value-accruing lifecycle will move from the infrastructure layer to the application layer much quicker. Previous compute cycle regime shifts have shown that we need the guts of the system built before we can build very useful applications on top of them.
Although we have begun to roll out AI into workloads, this was not yet at the point of ubiquity as most were in the experimental phase. The lowering of cost now reduces this friction, as AI can now be deployed much more rapidly with lower downside exposure.
Distribution will be the most important here. Megacap cloud companies are still positioned extremely well.
What Does This Mean for Nvidia?
So, does this mean that the semiconductor companies (particularly Nvidia) are in deep trouble? Not necessarily. Many who are close to the principles of lower-cost curve economics are familiar with the Jevons paradox, which states that increased efficiency in using a resource leads to increased consumption of that resource.
As AI proliferates due to lower costs, overall consumption will increase, which will still require higher compute loads.
Those who say that we are overbuilt in the very short term may well be correct, but over the longer term, I do believe the Jevons paradox will prevail.
What this creates is a short-term structural overbuild vs. a longer-term benefit.
A lot is still happening very quickly with a lot of moving pieces, but this is a market you can trade in over the short term that will provide compelling entry points if you have conviction in your long-term thesis.
Very interesting times ahead, indeed.
Strong convictions. Loosely held.
– Nick Mersch, CFA
Receive the latest tech insights in your inbox every month.
The content of this document is for informational purposes only and is not being provided in the context of an offering of any securities described herein, nor is it a recommendation or solicitation to buy, hold or sell any security. The information is not investment advice, nor is it tailored to the needs or circumstances of any investor. Information contained in this document is not, and under no circumstances is it to be construed as, an offering memorandum, prospectus, advertisement or public offering of securities. No securities commission or similar regulatory authority has reviewed this document, and any representation to the contrary is an offence. Information contained in this document is believed to be accurate and reliable; however, we cannot guarantee that it is complete or current at all times. The information provided is subject to change without notice.
Commissions, trailing commissions, management fees and expenses all may be associated with investment funds. Please read the prospectus before investing. If the securities are purchased or sold on a stock exchange, you may pay more or receive less than the current net asset value. Investment funds are not guaranteed, their values change frequently, and past performance may not be repeated. Certain statements in this document are forward-looking.
Forward-looking statements (“FLS”) are statements that are predictive in nature, depend on or refer to future events or conditions, or that include words such as “may,” “will,” “should,” “could,” “expect,” “anticipate,” intend,” “plan,” “believe,” “estimate” or other similar expressions. Statements that look forward in time or include anything other than historical information are subject to risks and uncertainties, and actual results, actions or events could differ materially from those set forth in the FLS. FLS are not guarantees of future performance and are, by their nature, based on numerous assumptions. Although the FLS contained in this document are based upon what Purpose Investments and the portfolio manager believe to be reasonable assumptions, Purpose Investments and the portfolio manager cannot assure that actual results will be consistent with these FLS. The reader is cautioned to consider the FLS carefully and not to place undue reliance on the FLS. Unless required by applicable law, it is not undertaken, and specifically disclaimed, that there is any intention or obligation to update or revise FLS, whether as a result of new information, future events or otherwise.