Oracle Chairman Larry Ellison warns that AI models like ChatGPT and Gemini are rapidly becoming commodities because they are trained on the same public internet data, arguing that the only sustainable competitive advantage lies in companies that own and train on proprietary, exclusive datasets.
Larry Ellison $ORCL highlighted something critical: models like ChatGPT, Gemini, Grok, and Llama are all trained on largely the same public internet data.
When everyone trains on the same information, models inevitably converge. That’s why AI is moving toward commoditization.… pic.twitter.com/rhG7H0Eheb
— Daniel (@danielisdizzy) March 1, 2026
Larry Ellison $ORCL highlighted something critical: models like ChatGPT, Gemini, Grok, and Llama are all trained on largely the same public internet data.
When everyone trains on the same information, models inevitably converge. That’s why AI is moving toward commoditization.
The real moat isn’t the model itself. It’s the proprietary data behind it.
Companies that can train on exclusive datasets gain an advantage competitors can’t replicate.
Having data that no one else has will allow you to dominate your market.
Every AI lab thinks they’re in a compute race.
They’re not.
Ellison: “For these models to reach their peak value, you need to make private, privately owned data available to those models.”
Every major lab trained on the same foundation.
ChatGPT. Anthropic. Grok. Llama.
Ellison: “They’re all trained on all of the data on the internet.”
The public web is fully commoditized.
Every lab scraped it.
The advantage is zero.
But here’s what nobody’s saying out loud.
The internet isn’t ground truth.
It’s humanity’s most distorted mirror.
Algorithms engineered to reward rage.
Bots amplifying the most extreme five percent.
Industrial-scale manipulation running continuously for two decades.
That’s the substrate every frontier model was built on.
You’re not training on reality.
You’re training on reality’s psychotic reflection.
And you’re compressing that distortion into a system that may eventually rewrite the world.
The actual moat isn’t compute.
Isn’t architecture.
It’s private data that never touched the web.
Offline archives. Proprietary records. Ground truth that can’t be scraped, reconstructed, or reverse-engineered.
Train on that and you don’t get marginal gains.
You get a model that sees the world as it actually is instead of as the internet performed it.
The race to AGI won’t be won by whoever scraped the most websites.
It’ll be won by whoever controls the data no one else can access.
That door is open right now.
And the most consequential data grab in human history is happening in silence.