Is DeepSeek a Game-Changer or Just Panic?

DeepSeek sent shockwaves through the market, triggering panic among investors. But is the fear justified, or an opportunity in disguise?

Deep Panic.

That’s what many investors felt today.

If you weren’t sure what to do today … don’t feel like you’re alone.

Today I’m going to lay out the facts so you can decide if you should be panicked or not.

To be clear, panic is an investors best friend.

Especially the informed investor.

So let’s get into it ..

Semiconductors and AI software aren’t iPhones or Nike shoes.

Many investors knowledge is only surface level.

They know what Nvidia does … but they don’t carry it around in their pocket.

They might use ChatGPT, but to them - it’s just Alexa or Siri on steroids.

That makes the industry ripe for confusion and panic like we saw today.

This has happened before.

In 2023 you might recall LK-99.

People lost their minds for a few weeks thinking a small research firm in Korea developed a material with superconducting properties.

There was even a “white paper”

But if you did 30 mins of internet research, you realized pretty quick this was likely a hoax.

I even wrote a blog: Don’t Fall for the LK-99 Hype.

Keep in mind when I wrote that - I had my boss and 3 copywriters telling me what a great stock promotion LK-99 would be.

I could have made money promoting LK-99 to people that wouldn’t know any better.

I told them to fuck off.

Fast forward to today.

People are flipping out about DeepSeek like it’s LK-99.

Here’s what we know:

• It was pre-trained using ChatGPT

• Most benchmarks show it’s marginally better than what’s on the market today.

• The company said V3 cost about $5.5M to develop

• But that figure doesn’t include initial research, scraping the data data from ChatGPT, Capex for their various pre-training clusters, etc

• Numerous sources, including Elon Musk believe the Chinese based company is not telling the truth about how many GPUs it has because they violate US sanctions.

• The site was shut down to new users today.

• All the data, including keystrokes and location data is sent back to mainland China.

Based on those facts:

This is why semiconductor stocks like Nvidia NVDA 0.00%↑ sold off hard today:

• If it only cost $5.5M to train a LLM that competes with the best on the market - CapEx spending is way over exaggerated.

• Companies like Microsoft, Amazon and Google won’t need to spend billions on Nvidia GPUs if the cost to train/launch is so cheap.

This ignores the fact that the vast majority of compute requirements for AI is actually in the “inference” or running of the AI models.

This is why companies need to either throttle your access to the model - or charge you a fee like $20/month.

It’s not “cheap” to run applications like ChatGPT to a wide audience.

DeepSeek didn’t solve that.

In fact, the site shut itself down likely because it didn’t have enough compute power to meet the surge in demand.

The fact investors ignore these details is not surprising.

It also ignores the fact that DeepSeek omitted all costs except a final training run.

And if you believe a Chinese based company for its word, the Nigerian Prince who sent you an e-mail last week is still waiting for your Western Union wire transfer.

That’s not racist to think that way … it’s just the reality.

Also, ask yourself this ….

If DeepSeek was going to dramatically reduce the cost to launch AI - why did China just announce an $8.2B investment fund a week ago?

Seems odd.

Even odder … the rumors around DeepSeek having access to an Nvidia GPU cluster larger than most US based firms is not new.

Dylan is worth following on twitter. I have access to his paid stuff.

This isn’t a guy making shit up for clicks.

He’s got big time clients paying big money for his info. He doesn’t need followers or fans.

Note that this tweet is from November 20,2024.

The hype and misinformation around DeepSeek goes back months.

DeepSeek is actually old news.

It’s just the social media algorithms picked up the hype this weekend and sent uniformed normies into a panic.

The news media sees it - and communication majors who were experts in foreign policy last week start writing about semiconductors.

This is how the “fake news” cycle operates.

Meanwhile, anyone with any knowledge of the semiconductor industry knew DeepSeek’s $5.5M number was bogus.

So now what?

DeepSeek changes nothing.

China was always going to develop their own AI.

For obvious reasons.

Just like Temu or Alibaba didn’t crush Amazon, DeepSeek isn’t going to eliminate American AI companies.

LLMs are new. Most do a few things well, but are far from perfect.

By next month, OpenAI, Meta, Anthropic, Grok, and Google will release models that are better than DeepSeek.

This pattern will continue for a while.

The average person isn’t using AI.

As the use cases expand - the need for more compute increases.

But as adoption increases cost of hardware always goes down.

Remember what your first computer or HD TV cost?

Nvidia is unique since it has very little competition. The faster AI is adopted, the more they benefit due to this.

Based on this information - I bought stocks that were whacked on DeepSeek panic.

Keep in mind that I can both afford to buy more if the panic continues - and if the panic turns into reality … and Nvidia’s days are over, my kids eat.

I’ve also spent the last several month buying (and recommending) beaten down stocks like Boeing, Nike, Starbucks, Chipotle, McDonalds, Louis Vuitton, etc - so if you’ve ignored building a balanced portfolio - today is a good wakeup call.

DeepSeek’s hype underscores the market’s vulnerability to fear. While it changes little for the AI industry’s trajectory, it serves as a reminder: informed, balanced investing beats panic every time.