You can listen to the audio version here:
Alright, listen up, because this is wilder than your ex showing up at your wedding with her new man. The AI market just got wrecked, and it’s all because of some Chinese AI lab called DeepSeek.
We’re talking about a full-blown market panic—Nvidia ($NVDA), Microsoft ($MSFT), Alphabet ($GOOGL), and other big tech stocks took a massive nosedive, wiping out over a trillion dollars in market value. And what was the reason? Some AI nerds in China claimed they could train a ChatGPT-level model for pennies on the dollar compared to OpenAI.
How Bad Was It, you ask?
Nvidia ($NVDA): Down 17% 🔻 evaporating about $600 billion in value overnight.
Microsoft ($MSFT): Slipped 2.5% 🔻 not as bad, but still a hit.
Alphabet ($GOOGL): Dipped 4% 🔻
ASML (Dutch chip giant): Down 7% 🔻
And just like that, the AI hype bubble got a serious reality check.
(Sources: Reuters and Yahoo Finance)
Why the market freaked out?
DeepSeek basically walked into the AI arms race and said, "Hey, OpenAI, you spent $100 million on GPT-4? Cute. We did it for $5.6 million. Suck it." And boom—Wall Street lost its f*cking mind.
The second investors saw this, they had two major panic attacks:
If AI models can be built this cheap, who needs Nvidia's overpriced GPUs?
If China can do this now, what's stopping them from running the AI game in a couple of years?
People started selling Nvidia stock like it was a goddamn rug-pull sh*tcoin, thinking the AI gold rush just got a whole lot cheaper—and that’s bad news for anyone betting big on expensive computing power.
But here’s the thing, and this is why we are covering this:
Is this even f*cking real⁉️
Like, did DeepSeek actually figure out some AI wizardry, or are they just running the best marketing scam of the decade? That's what we're about to find out. Because, this whole thing smells fishier than a Tinder date that "just got out of a long-term relationship, and is more like a home body"
2. What Is DeepSeek, and Why did it cause panic?
Alright, so let’s talk about DeepSeek, the Chinese AI upstart that just crushed Nvidia’s stock price and sent the entire AI industry into a tailspin.
This ain't some random-ass AI side project as many people like to put it. DeepSeek is backed by big Chinese investors and is making some bold claims—basically, they’re walking into the AI world with their d*ck swinging, telling OpenAI, Microsoft, and Google to get out of the way.
So who the hell are these guys?
DeepSeek is a Chinese AI company that popped up in 2023.
They dropped a series of models under the “DeepSeek” name, with their latest one, DeepSeek-V3, being their claim to fame.
Unlike OpenAI, they’ve been flying under the radar until now—but that all changed when they claimed to train their AI for next to nothing compared to the big dogs.
Their AI app skyrocketed to #1 on the Apple App Store in the U.S., beating out even ChatGPT.
DeepSeek came out swinging with a single claim that turned the AI industry upside down:
"We trained our AI model for just $5.6 million."
Now, to put that in perspective, OpenAI reportedly burned over $100 million training GPT-4.
So imagine you’re Nvidia, OpenAI, Microsoft, or any AI-heavy tech company. You just spent Fort Knox money building AI models, and now some Chinese upstart is saying they did it for pocket change?
So tha's why investors had a full-blown panic attack.
If training AI models just got that cheap, Nvidia’s overpriced GPUs might be more useless than crypto bro’s “financial advice".
If DeepSeek actually pulled this off, OpenAI and Google and the US might not be the AI kings for long.
If AI can be developed this efficiently, every single AI stock is overvalued as hell.
But back to our original question:
Is this even f*cking real⁉️
Well we did some digging and by "we" I mean "I", and by "some" I mean "a lot more than a healthy amount". I went through both ChatGPTs and Deepeek's whitepapers: (I included the links to them, if you don'T have a family or loved ones and want to burn your time on them. )
And I found some pretty exciting stuff, so let me break it down for you with a simple fast food analogy / story:
You walk into OpenAI and ask for a burger:
OpenAI (ChatGPT): Our burger costs $100 million
You: Thanks sound cool, I take it!
OpenAI (ChatGPT): Here you go enjoy.
End of story.
Now let's look at the same scenario with Deepseek
You walk into Deepssek and ask for a burger:
Deepseek: Our burger only costs $5.6 million
You: Wow awsome, I take it!
Deepseek: Here you go your total will be $1.6 billion
You: What?! Why?!
Deepseek: well our burger patty costs $5.6 million but you have to add in:
The bun (data collection & processing)
The ketchup (infrastructure & hardware costs)
The fries (energy & cooling expenses)
The drink (engineers & R&D salaries)
And of course, the restaurant itself (servers, deployment, and maintenance)
So yeah, that "cheap" burger? Turns out it costs just as much—if not more—than OpenAI’s. But DeepSeek just left that part out of the ad. They “forgot” to mention a sh*tload of fine print.
DeepSeek’s Whitepaper: What They Actually Said
DeepSeek’s official whitepaper lays out exactly where the $5.6 million number comes from.
And guess what? It only covers one thing:
🚨 GPU hours used during training. That’s it. 🚨
Here’s the full breakdown of their own numbers:
Pre-training: 2.66 million GPU hours → $5.3M
Context extension: 119K GPU hours → $0.24M
Fine-tuning: 5K GPU hours → $0.01M
Total: 2.78 million GPU hours = $5.6 million. (Source: DeepSeek Whitepaper)
That’s like saying you built a supercar for $10,000—and then revealing you only counted the cost of the engine.
Meanwhile, OpenAI Counted the Whole Burger
In comparison, OpenAI’s $100M+ cost estimate for GPT-4 includes everything:
✔ GPUs (obviously)
✔ Data collection & processing (a.k.a. the fuel for the AI)
✔ Hardware infrastructure (massive servers & networking costs)
✔ Energy & cooling (AI models don’t run on good vibes)
✔ R&D & engineering salaries (because someone has to build the damn thing)
✔ Security & compliance (especially to stop lawsuits)
✔ Deployment & inference costs (keeping the model running post-training)
That’s why DeepSeek’s "cheap burger" analogy is an illusion. They only counted the patty. The bun, fries, and drink? That's all extra.
So, What’s the Real Cost of DeepSeek’s Model?
Some AI analysts dug into what DeepSeek probably actually spent when you count the full burger meal. Their estimate is:
Between $1.3 billion and $1.6 billion.
Now, just to be clear, that’s NOT confirmed by DeepSeek, but in my and in many analysts opinion something definatlly smells fishy. If they actually built this for just $5.6M, they’d be selling a trillion-dollar secret, not just an app.
So, is DeepSeek full of sh*t?
I say chances are yes at least partially, either that or they found some AI magic that no one else in the world has. But realistically? They just hid the real costs behind a sexy headline. And Wall Street ate it up like it was the second coming of ChatGPT.
Why this matters, and why Nvidia got f*cked)
Now you see why Nvidia investors lost their minds over this. If AI models could actually be trained this cheap, then:
❌ No need to buy as many GPUs from Nvidia
❌ Big tech companies overpaid for AI development
❌ Smaller AI startups could suddenly compete
But since DeepSeek almost certainly left out key costs, the panic sell-off might have been completely overblown—a fake crisis caused by bad math. So yeah, investors got played because they didn’t check the receipt.
Just to reiterate once again this is not a 100% fact, but very likely be the case. But analysts have some other theories of what could have happaned some have ground, while others sound totall bullsh*t to me, so let's put on our tinfoil hats and let's look at some:
Theory 1: Did DeepSeek get Nvidia chips despite sanctions?
The U.S. has been cockblocking China from getting its hands on Nvidia’s best AI chips. The have imposed sanctions to stop Chinese companies from building GPT-4-level models. But we all now that "life finds a way" The U.S. government is now investigating whether DeepSeek got their hands on Nvidia’s chips anyway—through backdoor channels in Singapore. Translation: If true, DeepSeek had access to banned AI hardware that wasn’t accounted for in their “cheap” cost estimates.
What this all means⁉️
If DeepSeek illegally acquired high-end GPUs, their entire operation might be running on forbidden tech.
This would mean they didn’t actually train their AI model for cheaper—they just cheated the system.
Theory 2: Did DeepSeek steal data from OpenAI?
Another wild theory is that Deepseek didn’t train their AI from scratch at all—they just stole OpenAI’s homework. OpenAI is already investigating whether DeepSeek used data scraping (a.k.a. copying tons of ChatGPT responses) to train their model instead of actually building their own dataset.
What this all means⁉️
If DeepSeek scraped ChatGPT’s answers, they essentially reverse-engineered OpenAI’s model without doing the real work.
It would explain how they built a competitive AI model so fast and on the cheap.
If OpenAI proves this, DeepSeek could face legal hell and get their whole model blacklisted internationally.
Theory 3: Was this just a big-ass market manipulation play?
Some analysts are whispering that this entire DeepSeek situation wasn’t about AI at all—it was about f*cking with the U.S. stock market.
What this all means⁉️
Chinese state-backed accounts hyped up DeepSeek’s AI model on social media right before Nvidia’s stock tanked.
The timing was too perfect—DeepSeek’s “cheap AI” news dropped right before the biggest Nvidia stock crash in history.
Some believe this was a coordinated effort to trigger a sell-off, shaking investor confidence in U.S. AI companies.
If this theory is true, then DeepSeek didn’t just disrupt AI—it was used as a weapon in the financial war between the U.S. and China.
Theory 4: Is deepSeek’s AI a security & censorship risk?
Now, let’s talk about the biggest concern—DeepSeek’s AI could be a major security risk.
What this all means⁉️
The app is already huge in the U.S., topping the App Store charts.
China controls DeepSeek’s AI, meaning they could manipulate search results, filter information, and spread propaganda however they want.
There’s also a concern that DeepSeek might be collecting U.S. user data, because, let’s be honest, China isn’t exactly known for respecting privacy.
If DeepSeek’s AI becomes a major platform in the U.S., it could be used to control narratives, push misinformation, or even censor certain topics, all without users realizing it.
So, what’s the takeaway here? DeepSeek’s “cheap AI” story is a lot messier than it looks.
The $5.6M cost claim is misleading. But HOW they actually pulled this off is Still unclear.
If even one of these theories is true, DeepSeek isn’t a breakthrough—it’s just another AI scandal.
🚨 If the Nvidia GPU smuggling theory is true → DeepSeek could face bans & sanctions.
🚨 If the data scraping theory is true → OpenAI might sue their asses off.
🚨 If the market manipulation is true → More government crackdowns on AI stocks.
🚨 If the security risk theory is true → DeepSeek could get banned in the U.S., like TikTok 2.0.
So, What Happens Next?
This whole DeepSeek debacle has completely changed the AI game. Here’s what to watch for:
🔥 AI cost efficiency war incoming – Now that DeepSeek set the bar low (even if it’s fake), OpenAI, Google, and Meta will have to prove their AI is worth the cost. Expect pricing pressure and more AI models dropping fast.
🔥 Regulatory crackdowns – If DeepSeek scraped data or smuggled Nvidia GPUs, they could be hit with major legal trouble. The U.S. might crack down harder on Chinese AI as a response.
🔥 Nvidia’s comeback arc? – The sell-off might’ve been overblown, but if AI development actually gets cheaper, Nvidia will need to prove they’re still essential. Expect them to drop new AI chips with better efficiency to stay on top.
🔥 Is DeepSeek getting banned? – If U.S. lawmakers take the security risks seriously, we could see DeepSeek banned in the West, just like TikTok is facing U.S. bans right now.
Comments