I think the next big wealth creation opportunity in AI trade is **access to unique data**.
Memory trade is probably near the later stage now. There can still be upside, but mostly I see it as swing trade rather than another huge long-term opportunity.
GPU hype trade already peaked in 2025. Everyone knows GPU and ASIC story now.
Then there is fabric/infrastructure trade like Marvell. Personally I don't really see the unique value proposition there. A lot of hype but not enough real revenue growth. I would rather own memory or compute companies than these.
So what is actually next?
I think it is **data**.
More specifically, real-time human data that AI companies cannot simply reproduce themselves.
Look at what **Elon Musk** has been doing with **X**.
When xAI acquired X, he basically put the AI model and one of the largest real-time human discussion platforms under the same roof.
And since then look at how aggressive X has become about protecting its data.
**On legal side,** X is suing data scraping companies:
[https://news.bloomberglaw.com/privacy-and-data-security/elon-musks-x-corp-sues-israeli-company-over-data-scraping](https://news.bloomberglaw.com/privacy-and-data-security/elon-musks-x-corp-sues-israeli-company-over-data-scraping)
**On technical side,** X has made automated access much harder with login restrictions, rate limits and other barriers:
[https://www.wsj.com/tech/twitter-limits-number-of-posts-users-can-read-prompting-disruptions-for-some-15bc01cc](https://www.wsj.com/tech/twitter-limits-number-of-posts-users-can-read-prompting-disruptions-for-some-15bc01cc)
This makes complete sense to me.
Why let everyone scrape your most valuable asset for free?
***Musk is basically walling up X data.***
First, Grok gets access to something other AI companies cannot easily get.
Second, X can charge other companies who want access to that data.
[https://www.cbc.ca/news/business/x-third-parties-user-data-1.7356152](https://www.cbc.ca/news/business/x-third-parties-user-data-1.7356152)
And I think this type of data is becoming more valuable, not less.
AI already consumed huge amount of static internet data. What becomes more interesting now is **live discussion, opinions, arguments, recommendations, reactions and actual human interaction.**
Apart from X, there is really one public company sitting on a massive amount of this data:
**Reddit.**
And Reddit is slowly doing the same thing.
Google and OpenAI already signed agreements with Reddit for data access. At the same time Reddit is restricting unauthorized crawlers and making it harder for AI companies to simply take the data for free.
This is where I think people are still underestimating **Reddit**.
If AI companies increasingly need fresh human discussion, but Reddit controls access to it, then the value of Reddit's data goes up significantly.
The AI trade started with compute.
Then memory.
I think the next one is **who owns the data AI (LLM) needs but cannot generate itself.**
And **Reddit** may be sitting on one of the most valuable piles of that data on the internet.
