Sell-Off Keeps Accelerating

While NVIDIA (Nasdaq: NVDA) is down 17% today, many stocks across technology are rallying. Let’s take a look at a few notable stocks:

• Apple (Nasdaq: AAPL):Is up nearly 4% today. Apple has pursued a strategy of keeping AI computing on device. After the announcements from Deep Seek that make models much more efficient, Apple’s strategy looks more sound.
• Salesforce (Nasdaq: CRM): Is up nearly 5% today. The company is one of the leading names around the growth of ‘agents,’ which could benefit from significantly lower computing costs as Deep Seek’s advancements are replicated by other AI models.

We’re continuing to compile reactions to Deep Seek from across the investing and technology world.

Let’s look at what Andrej Karpathy, the former director of AI had Tesla had to say.

As some background, on December 26th Andrej commented on how efficient DeepSeek V3 model was. While the world is reacting in an outsized way to R1 today, Andrej was ahead of the curve recognizing the breakthroughs being pioneered by DeepSeek.

Here’s some highlights from his commentary today:

• “I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed in AI. You may not always be utilizing it fully but I would never bet against compute as the upper bound for achievable intelligence in the long run.”
• “Data has historically been seen as a separate category from compute, but even data is downstream of compute to a large extent – you can spend compute to create data. Tons of it.”

One reading of Andrej’s thoughts would be that while the near-term could be rocky for many of the computing infrastructure stocks declining rapidly today, their long-term outlook looks robust.

We’ve been looking at what Wall Street is saying about the release of DeepSeek’s newest models, let’s check what some tech leaders are saying.

Gavin Baker, who is the CIO of Atreides Management, posted a long and thoughtful note about DeepSeek.

Some key takeaways:

• He highlights some of the real breakthroughs in DeepSeek’s new models
• But also notes that access to leading U.S. models and a technique for shrinking them (distillation) is another key factor in DeepSeek’s breakthrough.
• He notes that the biggest consequences of this announcement could be computing moving to the ‘edge’ – which would create a massive upgrade cycle for devices like smartphones.

The entire post is worth a read – we’re now seeing some stocks like Meta Platforms (Nasdaq: META) that were down to begin the day move to the green.

It is very likely DeepSeek’s announcement benefits companies like Meta and other hyperscalers with unique data and distribution advantages.

Let’s take a look at where some notable names levered to the AI buildout are in early trading today:

• NVIDIA (Nasdaq: NVDA): Down 13%
• Broadcom (Nasdaq: AVGO): Down 12.5%
• Vistra (NYSE: VST): Down 24%
• Vertiv (NYSE: VRT): Down 25%

We’ll see where markets trend at the end of the day. Yet, for now, investors appear to be selling first and asking questions later.

Bernstein released a research note on what Deep Seek means for NVIDIA and other AI stocks this morning, let’s look at some of the key takeaways:

• Did DeepSeek Build OpenAI For $5 Million? 

Bernstein first addresses DeepSeek’s cost to refute the $5 million number cited in coverage of the model.

Bernstein says:

“There are actually two model families in discussion. The first family is DeepSeek -V3, a Mixture of Experts (MoE) large language model which, through a number of optimizations and clever techniques can provide similar or better performance vs other large foundational models but requires a small reaction of the compute resources to train…

The oft-quoted “$5M” number is calculated by assuming a $2/GPU hour rental price for this infrastructure which is fine, but not really what they did, and does not include all he other costs associated with prior research and experiments on architectures, algorithms, or data…

Overall, our own (semi-informed layperson’s?) reaction is not to buy into the doomsday scenarios currently playing out in the Twitterverse. We continue to like the AI stories in our coverage, particularly NVDA and AVGO (both rated OP).

While AI hardware companies are seeing large losses in premarket trading, the efficiency gains from Deep Seek could accelerate adoption from software companies.

So, how are they performing premarket?

Let’s check in on a software stocks with the most to gain if the cost of AI models plummets:

• Salesforce (NYSE: CRM): Down .6% premarket
• ServiceNow (NYSE: NOW): Down .87% premarket
• Snowflake (NYSE: SNOW): Down 2.27% premarket

Clearly the panic hasn’t spread to these stocks and investors may increasingly rotate out of hardware AI stocks and into them in the weeks to come.