Markets Are Getting DeepSeek And Nvidia So Wrong

I bought 5 shares of Nvidia today, ramping up my stake in the company from 3 years ago (I originally believed in the stock for its potential in Web3 gaming), as a cash-strapped person bootstrapping 2 startups with my own money.

\The markets are supposedly always right, but they are also so funny.

So, everyone in the world seems to be talking about this new model called DeepSeek. It’s kinda wild, because until recently, it was barely a blip in AI circles, let alone the mainstream (the company has been around a while and shipping top models consistently)

But the conversation around it has spiraled into something else entirely — US v/s China. 🥊

DeepSeek is an open-source AI model, in the same vein as Meta’s Llama series.

The fact that an open-source model, developed on a shoe-string budget (I mean as far as AI training goes — you remember when Sam Altman was asking for 7 billion effin dollars?), without access to top-of-the-line chips, could rival the most-funded closed-source models is astonishing, is it not.

Nasdaq plunged. Hard. Nvidia plunged. Hard. The cause? Everyone thinks China is winning over the AI war over the U.S.

Umm. Let’s hold on a second.

Nvidia Is Still the Core AI Hardware Beneficiary

DeepSeek’s team trained the model using Nvidia chips— just maybe not the top-shelf H100 or A100 GPUs. But it was still Nvidia hardware!  If Trump (and Biden before) let them, they’d probably be buying every top-tier GPU Jensen Huang’s factories can churn out.

Besides, the demand for AI training hardware right now is off the charts.

According to recent market analyses, the global AI chip market could grow to over $250 billion by 2030. (And that’s a conservative estimate by some accounts.)

Not to mention that Nvidia has a near-monopoly in the current generative AI boom. So, do we really think Nvidia is about to go out of style because of one open-source model? Hardly.

The Ever-Growing GPU Shortage

There is a huge GPU shortage. A bunch of big tech players—Meta, xAI, OpenAI, Google, and yes, Chinese tech titans like Baidu and Tencent — are all in a mad dash to grab as many advanced chips as possible. As of mid-2023, there have been multiple reports of 6–9 month wait times for Nvidia’s H100 GPUs, because they’re so popular among AI research labs and commercial outfits.⁵

Meanwhile, alternatives like Groq or Graphcore have interesting tech but haven’t matched Nvidia’s scale or ecosystem yet. Even the bigger chip companies like AMD are trying to catch up with their Instinct series, but the developer ecosystem heavily favors Nvidia’s CUDA software stack.

AI Is Moving From Novelty to Utility

Even beyond the big research labs, AI usage is becoming mainstream in everyday life. OpenAI just released Operator (kind of a “do-stuff-for-me AI” that can browse the web, order groceries, and carry out tasks).

AI agents are coming.

Whether they are labeled AutoGPT, Meta Agents, or something else, the idea is the same — AIs that can operate independently, making decisions and executing tasks to completion.

That’s a lot of computing horsepower needed, and guess whose hardware they use for inference and training? Yep, Nvidia again. Even if partial inference moves to CPU or specialized chips over time, the big heavy-lifting for training new models is definitely going to soak up GPUs.

Don’t Fall for the Panic

Yes, DeepSeek is amazing. Yes, China is aggressively funding AI research. But no, that doesn’t automatically mean we throw our hands up in despair or watch Nvidia crash forever.

If anything, the more players (be it in the U.S., China, or anywhere else) push the boundaries of AI, the more GPUs, networking, and data infrastructure they’ll need. Nvidia (and a few of its top competitors) will remain massive beneficiaries.

So here’s my two cents:

• Appreciate DeepSeek’s ingenuity.

• Acknowledge China’s ambitions in AI.

• Don’t succumb to the market’s doom-and-gloom knee-jerk around Nvidia.

Agents are coming. AI usage is blowing up. And guess which hardware is indispensable to train and run all of it? I’m not selling my shares anytime soon.

Stay calm, stay curious, and as always — invest responsibly.