Some Deep-ish Thoughts About DeepSeek

Today’s market reaction should serve as a reminder that even disruptors are susceptible to disruption.  The sudden, adverse market reaction to DeepSeek, a Chinese open-source artificial intelligence platform, indicates that some of the key assumptions that have been driving the AI trade, and hence major indices, are getting reassessed today.

The advent of ChatGPT opened the doors to a wave of investment in artificial intelligence.  AI itself is not a new theme, per se.  It’s been around for decades, but ChatGPT moved it from the realm of data scientists to something accessible by almost anyone.  Their model, along with others, relies on a significant investment in computing power.  This notion was reinforced just last week with the announcement of Project Stargate, which provided a further lift to key providers and investors like NVDA, ARM, and Microsoft (MSFT).

As with any gold rush, savvy investors realized that perhaps the best way to profit was not from mining the “gold” itself – especially with ChatGPT’s parent, OpenAI, being a private company – but via the providers of the “picks and shovels” used by the miners.  Nvidia (NVDA) immediately pivoted from being known as the preferred provider of chips for crypto mining to that of an AI-focused company.  This is something we noted after NVDA’s earnings call in February 2023, when we noted that the company referred to AI 70 times – 37 of them even before the first question was posed.  Kudos to Jensen Huang for recognizing the potential, and to early investors who rode that wave.

But OpenSeek is forcing us to rethink that notion.  If an upstart developer can create an open-source platform that requires far less hardware to be productive, then it has the potential to puncture the expected demand for the type of chips provided by NVDA, Broadcom (AVGO), AMD, and the like.  The first two stocks are down about -15% as I write this, which in turn are helping pull the Philadelphia Semiconductor Index (SOX) down by about -8.5% – its worst day since 2020.

Just last week, when Apple was about -4% lower, we noted that a decline in the largest company had a depressing effect upon market capitalization weighted indices.  Well, it’s that much worse when the new #1, along with #6 are each down by 15%.  Today’s -1.8% decline in the S&P 500 (SPX) and -3.2% drop in the Nasdaq 100 (NDX) are direct consequences of their concentration.  The ‘Mag 7” plus AVGO make up more than 1/3 of SPX and nearly ½ of NDX.  Moves of today’s magnitude in the largest have inordinate effects on these key measures.  Interestingly, advances on the NYSE roughly match those of declining stocks, while on Nasdaq, decliners only lead by about a 3:2 margin.  Top-heavy indices benefit mightily when the biggest components lead the charge higher but fall disproportionately when the opposite occurs.

I believe that part of the today’s sudden adverse market reaction was a direct result of a wave of complacency that overtook the equity market last week.  On Tuesday we wrote:

To my mind, the most striking element of the market reaction continues to be decline in the Cboe Volatility Index (VIX).  The current level of 15.21 – which would represent the lowest close since the day after Christmas– shows that traders are anticipating relatively placid markets over the coming month.  Remember, VIX does not explicitly measure fear, but instead is constructed to portray the market’s best estimate of SPX volatility over the next 30 days.  Bear in mind that this period includes the peak of earnings season, an FOMC meeting, and the potential for tariffs against Canada and Mexico to begin on February 1st.  Virtually no one expects a rate change from the Fed next week, but the statement and press conference might offer important clues about the timing of future potential cuts.

Keeping in mind that 10 of the past 20 days have featured SPX closes of more than +/- 1%, perhaps we have gotten a bit too complacent.  The skew chart … shows that SPX volatilities are down pretty much across the board since last week, but especially in at-money and near-upside options.  Could the market be telling us that [it thinks] the easy money has been made?

I also reiterated a similar theme in media appearances on Wednesday and Friday.  We’ve had no shortage of volatile days – up and down – but somehow investors had decided that the likelihood of volatility diminished, major events notwithstanding.  That an unexpected event brought the volatility sooner than expected should remind us that the best time to buy an umbrella is often when the sun is shining. 

Going forward, we will need to decide if today’s events are a blip, an inflection point, or a turning point for the AI trade.  My gut says “inflection point,” though it is clearly too early to draw real conclusions in the midst of a major sector’s repricing.  At its best, OpenSeek will represent a boon for artificial intelligence, making it cheaper for all and thus improving the odds and the timeframe that its potential productivity boosts will be reflected in a wide variety of companies’ bottom lines.  Maybe not the one’s that we’d all expected, however.

My colleague Andrew Wilkinson put it succinctly in a message this morning with the phrase “where’s the moat?”  A good portion of the valuation premium that has been placed upon our largest tech stocks, our largest stocks overall, was based upon their entrenched competitive advantages.  That entrenchment is being reassessed right now, bringing some unwelcome moves for some stocks that seemed unassailable just last week.  I think back to the internet bubble, when many of us utilized computers provided by Sun Microsystems or Compaq, when Cisco (CSCO) was priced to rewire the whole world, and Yahoo was the search engine of choice.  The promise of the internet was realized, but remember that some of the companies that best utilize that promise – Alphabet (GOOG, GOOGL) and Meta Platforms (META) – only arose years later.

Go Source