AI stocks have carried the market in a way that is hard to ignore. Since ChatGPT launched in November 2022, they have driven roughly 75 percent of S&P 500 returns, 80 percent of earnings growth, and 90 percent of capital spending growth. That is enormous. It means almost all of the excitement, profit, and investment in the market is coming from one theme.
"AI related stocks have accounted for 75% of S&P 500 returns, 80% of earnings growth and 90% of capital spending growth since ChatGPT launched in November 2022."
-JPM Cembalest pic.twitter.com/85s2a3HIHp
— Daily Chartbook (@dailychartbook) September 25, 2025
Nvidia’s deal with OpenAI is a perfect example of how this works. The company is putting cash into OpenAI, but most of that cash will come back to Nvidia in the form of leased chips. Profits are being engineered, and investors are celebrating, while the question of real, sustainable earnings gets blurry. This is old companies with real revenues riding the AI wave, combined with valuations that feel like the 1999-2000 tech bubble all over again.
This is a real headline published by the CNBC – Not even the CNBC can spin a different narrative to hide what’s really going on here: revenues fabrication https://t.co/aemzq5ILC3 pic.twitter.com/QG4WZ2Ljvt
— JustDario 🏊♂️ (@DarioCpx) September 25, 2025
The difference today is that these aren’t standalone startups. Back in 2000, investors were chasing pure hype with no revenue track record. Today, the AI beneficiaries are profitable, established companies. That makes it feel safer, but it also masks risk. Most of the hype is embedded in divisions of existing businesses. If AI growth slows or margins get pressured, the market could correct faster than people expect.
Alibaba’s recent announcement made that danger clear. Their GPUs match Nvidia’s performance but come with a gross margin around 40 percent, compared with Nvidia’s 72.5 percent. That built-in cost advantage could quickly win them business, and suddenly the pricing power and margins that underlie Nvidia’s valuations are not guaranteed.
MAG 7 = Slowgrowth massive revenue and profit oldco + 1999-2000 era newco valued at $PLTR level bubble
The difference between the 2000 bubble and today is the newco's in 2000 were standalone and today they are all parts of oldcos
— Andy Constan (@dampedspring) September 25, 2025
The lesson is simple. AI is driving huge growth and investors are making a lot of money. But profits are concentrated, hype is baked into valuations, and competitors are already emerging to challenge margins.