Original video in Chinese.
Key Takeaway
- Massive infrastructure spending: AI research is optimistic, and compute demand is exploding. OpenAI is teaming up with NVIDIA and AMD to place huge bets across the entire chip/data center/power chain. If you don’t secure capacity first, you’ll fall behind and be left cold.
- Strategic positioning: Build a personal AI assistant like Windows. ChatGPT as the core entry point, plus API and hardware, across C-end and B-end scenarios, while maintaining continuity of experience and historical data.
- Ecosystem and Sora inspiration: Connecting third-party apps enables collaboration at the front end instead of pipe-like extraction from the back end. Sora lowers the barrier to creation and unleashes ordinary people’s demand; the explosion of niche subcultures signals a coming surge in AI demand.
You all definitely need to watch Sam Altman’s latest interview. He talked with Ben Thompson for more than half an hour and personally explained OpenAI’s strategy.
Recently, a lot has happened at OpenAI, including infrastructure investment, the launch of Sora 2, and new things at DevDay. All of it points to one judgment: AI is about to come in a big way.
Based on my own understanding, I’ve distilled this interview and want to share the key points with everyone. Sam mainly talked about these key issues.
First, why are they suddenly burning money like crazy on infrastructure?
I believe this is the issue the market cares about most right now.
OpenAI first partnered with NVIDIA to deploy at least 10 gigawatts of systems. NVIDIA plans to invest as much as $100 billion into OpenAI.
Right after that, OpenAI also partnered with AMD to deploy 6 gigawatts of AMD GPUs. And as part of the deal, if OpenAI reaches the chip deployment milestones, AMD will grant OpenAI an option to purchase a 10% stake in AMD.
A lot of people think this is just companies banding together to prop up a bubble. But Sam said in the interview that progress in AI research is extremely optimistic. They’re seeing explosive growth in AI compute demand. So at this point, they have to make company-level, massive bets.
This is a huge undertaking. Chips, data centers, electricity, software, business models, and everything else all have to move forward at the same time.
And you have to act decisively now. Otherwise, someone else will move first and reserve the capacity, and you won’t get a piece of this wave. In AI, if you miss one wave, you’re basically done.
Second, what exactly does OpenAI want to do?
The host said OpenAI wants to build the Windows of the AI era. Sam didn’t exactly agree out loud, but the concept he put forward was a “personal AI assistant,” and the core meaning is actually the same.
Sam believes you not only need a C-end super entry point like ChatGPT, but also an API business. These are all meant to let users use the same AI anywhere, while keeping the experience continuous—because the AI knows you and has your historical data.
It’s true that everyone has their own preferences when using AI. For example, my wife really likes ChatGPT, while I use Gemini and Grok in my daily life. We all want this AI experience to extend across all kinds of scenarios.
So OpenAI needs ChatGPT as its core entry point, plus integration with various apps, along with API services, and eventually the AI hardware that will appear in the future. All of this is an incarnation of the “personal AI assistant” in different scenarios, all meant to run through a user’s life.
Third, why let ChatGPT connect to third-party apps?
One huge concern we have about AI is that it will pipeline all apps and websites. These apps and websites are only providing services and content in the back end. On the front end, users basically don’t know where it comes from. It’s like AI search today. This is a kind of “AI vampire mode.”
What OpenAI wants to achieve is an “AI empowerment mode.” After ChatGPT connects to third-party apps, if users want to use a certain app’s service, they have to clearly tell ChatGPT the name of that app. Then ChatGPT calls that app and lets it take over.
The benefit of this model is that these partner apps get summoned, move from the back end to the front end, and work with ChatGPT to fully satisfy users’ needs. They’re not just sitting quietly in the background.
From the perspective of ecosystem building, this really is a good approach.
Fourth, what lessons can we draw from Sora’s success?
In the past, all creation tools were hard to use. As a result, good works became the privilege of a small number of people. In other words, the so-called 1% provided good content, and the remaining 99% could only be consumers.
But now, Sora has made the transformation from idea to work extraordinarily easy. It has greatly lowered the barrier to high-quality creative expression and unleashed ordinary people’s suppressed demand to create.
That’s the logic at the model level. At the product level, the core audience for the Sora app is actually users in small circles. OpenAI found that users sharing fun meme images within niche communities is a far bigger demand than we imagined.
That’s why so many so-called analysts and big influencers don’t get it. They don’t understand what’s fun about those simple, blunt AI videos. They can only make sense of mainstream entertainment, but they know absolutely nothing about the subcultures that emerge inside niche circles.
If you’re an investor or founder in AI, what Sam Altman shared will definitely give you some inspiration. My own judgment is that within the next 12 to 18 months, we may see a real explosion in AI demand. So over the next few months at the end of the year, I’ll put more money into the stock market.
OK, that’s it for this episode. If you want to understand AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you next time!