Something strange is happening in Silicon Valley. Every three months, AI companies are doubling in value. Not because they released a breakthrough product. Not because they suddenly became profitable.
NVIDIA was worth less than a trillion dollars two years ago. Today? Five trillion. OpenAI added $200 billion to its valuation between March and October 2025, while projecting they’ll lose $14 billion by 2026. Perplexity jumped from $500 million to $20 billion in eighteen months, raising money every two months like clockwork.
Here’s what nobody’s saying out loud: this isn’t normal growth. This is what happens when everyone’s terrified of missing the next big thing. When being wrong about AI feels riskier than overpaying for it.
And the scary part? AI is real. It’s transformative. It will change everything. But that doesn’t mean these companies are worth what investors are paying for them. The technology can win while the valuations collapse.
History keeps trying to teach us lesson. We just keep refusing to learn it.
Press enter or click to view image in full size
The Infrastructure Layer: When the Moat Turned Out to Be a Puddle
Two years ago, if you asked anyone who’d win AI, the answer was obvious: NVIDIA. They had the chips. They had CUDA. They had a decade head start. Everyone needed them. The moat looked unbreakable.
Wall Street ate it up. NVIDIA’s stock jumped 169% in 2024. Five trillion dollars of pure belief that nobody could catch them.
Then reality walked in.
Google didn’t just build their own TPU’s they signed a deal with Anthropic for up to a million units. China started making chips despite every sanction threw at them. OpenAI announced they’re designing their own.
But the real gut punch came in January 2025. A Chinese startup called DeepSeek proved you could train advanced models for a fraction of what everyone thought was impossible to go below. NVIDIA lost $600 billion in market value. In one day.
Six hundred billion dollars evaporated because a startup nobody heard of showed that the moat was never that deep.
That five trillion dollar valuation? It’s built on the assumption that NVIDIA stays a decade ahead. But they’re not a decade ahead anymore. Maybe not even a year.
The Model Layer: Building Billion Dollar Wrappers
Then there’s the companies actually building the AI models. OpenAI started alone. Now the list is almost boring: Anthropic, Google, DeepSeek, Qwen, Alibaba, and others.
Anthropic’s worth over $350 billion. They raise every few months. Each round is bigger than the last, like clockwork.
But here’s what nobody wants to say out loud: when GPT-5 or Gemini 3 or whatever comes next launches, your parents won’t notice. They’ll still ask it the same questions. Get similar answers. Sure, developers will feel the difference. Coders will have slightly fewer bugs. But regular people? They’re not paying extra for 3% better responses.
Yet every time one of these companies raises money, the pitch is the same: “We’re going to change everything.” And investors keep believing it, keep writing checks, keep doubling valuations every quarter.
What changed between funding rounds? Did they crack AGI? Find the secret to consciousness? Or did everyone just decide that not being in the AI game is scarier than admitting these numbers don’t make sense?
The Product Layer: Competing with Free
Now we get to the companies building on top of these models. This is where the bubble gets really uncomfortable.
Take Perplexity. They hit $20 billion in September 2025. Their product? AI-powered search. Pretty useful, actually.
The problem is ChatGPT now does search. Google does AI search. The companies that build the models Perplexity uses are competing directly with them.
Perplexity raised $1.5 billion. Their valuation went from $500 million to $20 billion in eighteen months. They’re raising money every two months now. That’s not a company proving its model works that’s a company trying to raise enough money to survive until… what, exactly?
They tried to buy Google Chrome for $34.5 billion. A $20 billion company trying to spend $34.5 billion on distribution. That’s not confidence. That’s panic.
Then came the India deal, and this is where it gets almost sad. They partnered with Bharti Airtel to give 360 million customers free Perplexity Pro for a year. The subscription normally costs ₹17,000 annually. Now it’s free.
Sounds impressive, right? 360 million users!
Except India is brutally price sensitive. Netflix had to slash their prices by 60% just to compete. Amazon Prime raised their fee by ₹120 and people revolted. And Perplexity thinks after a year of free, these users will suddenly pay ₹17,000?
When that free year ends in 2026, those “users” will do exactly what makes sense: go back to ChatGPT, which is free forever. Or Google, also free forever. Perplexity will be left with vanity metrics that disappear the second you ask for a credit card.
At a recent conference in San Francisco, 300 AI founders and investors were asked which billion-dollar startup they’d bet against. Perplexity topped the list. The people building AI companies are already calling it dead.
That’s the product layer in a nutshell: you’re worth billions, but you’re competing with free, and the companies you depend on are becoming your competitors.
When Smart People Start Saying “Bubble”
The Bank of England warned about AI market corrections. The IMF agreed. Jamie Dimon compared it to the early auto industry, the technology succeeded, but most companies involved went bust.
The CEO of DeepL said AI valuations are “pretty exaggerated” with “signs of a bubble on the horizon.” Lyft’s CEO was even blunter: “Let’s be clear, we are absolutely in a financial bubble.”
When the people making money off the bubble start warning about the bubble, that should terrify you.
Michael Cembalest from JP Morgan pointed out that AI stocks account for 75% of S&P 500 returns since ChatGPT launched. When one sector is the entire market’s growth story, what happens when that story cracks?
My Take
AI is real. AI will change everything. I believe that completely.
But I also believe that believing in the technology doesn’t mean you have to believe in every valuation. The revolution can happen while the bubble pops. Both things can be true.
NVIDIA might lose half its value and still be a successful company. OpenAI might revolutionize work and still never justify a $500 billion valuation. Perplexity might have built something useful and still go to zero because they couldn’t compete with free.
The technology winning doesn’t mean the investors win. That’s the lesson from every bubble in history.
Right now, we’re in the gold rush phase. Everyone’s excited, everyone’s getting funded, and anyone asking hard questions gets told they “don’t understand” or they’re “thinking too small.”
But history doesn’t care about your pitch deck. Math doesn’t care about your TAM. And markets eventually always, eventually demand that the numbers make sense.
The question isn’t whether AI will change the world. It’s whether you’ll still believe these valuations when it does, Because the excitement is real. The fear of missing out is real. But so is gravity. And what goes up on hype alone always, always comes back down.
The only question is whether you’ll see it coming.