# DeepSeek showed frontier work can come from outside major labs

Markdown copy of https://realalpha.app/s/deepseek-showed-frontier-work-can-come-from-outside-major/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

- Speaker: [Marc Andreessen](https://realalpha.app/people/marc-andreessen/) · Marc Andreessen is cofounder of a16z. · Co-founded Netscape and co-created Mosaic, then co-founded a16z, a venture firm backing AI, crypto and frontier tech.
- Host: [Erik Torenberg](https://realalpha.app/people/erik-torenberg/) · General partner, Andreessen Horowitz · Helped build Product Hunt; founded On Deck, Village Global and Turpentine.
- Show: a16z Podcast
- Episode: [Marc Andreessen's 2026 Outlook: AI Timelines, US vs. China, and The Price of AI](https://realalpha.app/v/xRh2sVcNXQ8/), Jan 7, 2026
- Moment: 29:50 to 32:47 of the episode (3 min), playing on https://realalpha.app/s/deepseek-showed-frontier-work-can-come-from-outside-major/; the episode from there: https://realalpha.app/v/xRh2sVcNXQ8/?t=1790

## Quote

> “this was **not a national champion tech company** at the time that DeepSeek was released.”
> — [Marc Andreessen](https://realalpha.app/people/marc-andreessen/) at [31:00](https://realalpha.app/v/xRh2sVcNXQ8/?t=1860) · a16z Podcast

## Takeaways

1. DeepSeek came from a Chinese quant hedge fund.
2. Marc says its release appeared to surprise even the Chinese government.
3. The release helped shift U.S. policy attention toward competing with China.

## About this segment

Marc Andreessen says DeepSeek came from a quant hedge fund rather than a national champion or research lab, and its open release surprised observers. He sees this as evidence that capable teams need not be giant incumbents.

From “Marc Andreessen's 2026 Outlook: AI Timelines, US vs. China, and The Price of AI”, Jan 7, 2026. [Watch the full episode](https://realalpha.app/v/xRh2sVcNXQ8/)

## Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[29:54](https://realalpha.app/v/xRh2sVcNXQ8/?t=1794) The big kind of supernova moment this year was the DeepSeq release. The DeepSeq release was surprising on a number of fronts. One was just how good it was. And again, along this line of it took the capability set that were running in large models in the cloud and kind of shrunk it onto a, you know, into a sort of a reduced size, you know, a smaller version with sort of equivalent capabilities that you could run on small amounts of local hardware.

[30:19](https://realalpha.app/v/xRh2sVcNXQ8/?t=1819) And so there was that. And then it was also a surprise that it was released as open source and particularly open source from China because China does not have a long history of open source. And then it was also a surprise that it actually came from a hedge fund. So it didn't come from a big R&D, you know, sort of university research lab. It didn't come from a, you know, from a big tech company. It came from a hedge fund.

[30:39](https://realalpha.app/v/xRh2sVcNXQ8/?t=1839) And it, like, as far as we can tell, it basically is this somewhat idiosyncratic situation where you just have this incredibly successful quant hedge fund with all these, you know, super geniuses. And the founder of that hedge fund, you know, basically decided to build AI. And, you know, at least externally indications are this was a surprise to even the Chinese government. It's impossible to prove, you know, what the Chinese government was surprised by or not.

[31:01](https://realalpha.app/v/xRh2sVcNXQ8/?t=1861) But, you know, there's at least the atmospherics are that this was not exactly planned. And this was not a national champion tech company at the time that DeepSeek was released. It was a, it sort of came out of left field, which by the way, is very encouraging for the field that it was possible for somebody to do that kind of who was unknown, right? Because it kind of means that maybe you don't need all these, you know, super genius, superstar researchers.

[31:19](https://realalpha.app/v/xRh2sVcNXQ8/?t=1879) Maybe actually smart kids can just build this stuff, which I think is the direction things are headed. And so that kicked off, I would say like this kind of, I don't know, copycat's the wrong word, but that was sort of, it feels like the success of DeepSeek and the success of DeepSeek from China as open source kind of kicked off a sort of trend in China of releasing these open-source models. You know, look, the cynics, you know, in D.C. would say, you know, yeah, like they're dumping, right?

[31:42](https://realalpha.app/v/xRh2sVcNXQ8/?t=1902) They're obviously dumping. They're trying to, you know, they see that the West has this opportunity to build this China industry. You know, they're trying to commoditize it right out of the gate. You know, there's probably something to that. You know, the Chinese industrial economy does have a history of, you know, sort of, let's say, subsidized production that leads to selling, you know, selling things below cost in some cases.

[32:00](https://realalpha.app/v/xRh2sVcNXQ8/?t=1920) But I think also it's like, I think that's almost too cynical of a view also, because it's just like, all right, wow, like, they're really in the race. Like, open source, closed source, whatever. Like, you know, they're actually really in the race. You know, we've talked in the past, I think, on LP calls about, you know, these policy fights. that we've been having in D.C. for the last two years. And, you know, there was a pretty big push within the U.S. government, you know, two years ago to basically, you know, restrict, you know, or outright ban, you know, a lot of AI.

[32:26](https://realalpha.app/v/xRh2sVcNXQ8/?t=1946) And, you know, it's very easy for a country that is the only game in town to have those conversations. It's quite another thing if you're actually in a foot race with China. And so I think actually the policy landscape in D.C. has, I would say, has improved dramatically as a consequence of sort of an awareness now that this is actually a two-horse race, not a one-horse race.