Interconnected

Interconnected

Share this post

Interconnected
Interconnected
($) DeepSeek Diffusion
Copy link
Facebook
Email
Notes
More

($) DeepSeek Diffusion

It’s about open vs closed, not (just) about the US vs China

Kevin Xu's avatar
Kevin Xu
Feb 07, 2025
∙ Paid
8

Share this post

Interconnected
Interconnected
($) DeepSeek Diffusion
Copy link
Facebook
Email
Notes
More
4
Share

ICYMI: I did some press and podcasting on DeepSeek recently on CNBC and the inestimable ChinaTalk podcast. Please give them a read and listen if you are interested!

This post is for our premium members only. If you are reading this in full, thank you for being an Interconnected Premium member! If you aren’t, I hope you become one by scrolling down and tear down that paywall! 😎


All the buzz around DeepSeek has so far had a familiar, if not tiresome, overtone: it’s all about the U.S. vs China. Had DeepSeek come out of, say, a Canadian AI lab, I doubt there would be such a confrontational setup to the discussion (though President Trump might beg to differ as he tries to conquer Canada as the 51st state).

These days, no technology topic exists in a vacuum without geopolitical implications, so the U.S.-vs.-China lens is not without merit. But as a student of not just geopolitical history but also technological history, I believe this perspective, if taken too far, misses a far more important and reliable pattern: the age-old battle between open source and closed source technology.

The diffusion power of open source technology tends to be underestimated. Despite all the cybersecurity and data privacy fears drummed up about DeepSeek’s models, their open source nature has led to rapid and global adoption.

Almost every major Chinese cloud platform, from Alibaba and Tencent, to Huawei and Baidu, now offers easy deployment of DeepSeek models, even though all these companies are developing competing models of their own. But it’s Chinese adopting Chinese, so nothing remarkable, right? Well, a long list of prominent Silicon Valley startups are adding DeepSeek to their underlying model layer too, from the high-flying Google killer wannabe, Perplexity, to fast growing AI coding products, like Cursor, Sourcegraph, and GitHub Copilot. AI-focused neocloud platforms, like Nebius, and newer startups in this space, like Together.ai, have quickly offered hosted solutions for DeepSeek. It’s everywhere in Europe too.

What’s most remarkable is that even U.S. hyperscalers—AWS and Azure—along with smaller cloud providers like Cloudflare, and even Nvidia itself (whose single-day market cap drop triggered by DeepSeek was one for the history books) are all offering DeepSeek models to their customers.

I honestly can’t keep track of them all. Depending on when you read this post, there are likely even more technology vendors offering DeepSeek models as part of their products. The “DeepSeek Diffusion” is real. But it is not new. In fact, in the annals of technology history, we’ve seen this movie play out over and over again.

A Familiar Script

This post is for paid subscribers

Already a paid subscriber? Sign in
© 2025 Kevin Xu
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture

Share

Copy link
Facebook
Email
Notes
More