China’s Moonshot Unveils AI Model, Fueling Tech Rout

**SPEAKER_1** (0:02)
Bloomberg Audio Studios, Podcasts, Radio, News.

**SPEAKER_2** (0:08)
This is Bloomberg Businessweek Daily, reporting from the magazine that helps global leaders stay ahead, with insight on the people, companies, and trends shaping today's complex economy, plus global business, finance, and tech news as it happens. The Bloomberg Businessweek Daily podcast with Carol Massar and Tim Stenovec. On Bloomberg Radio.

**SPEAKER_1** (0:31)
All right.

**Carol Massar** (0:32)
So let's get to it. There's a lot of stuff going on. As I said, I'm obsessed with some of the news out of China. A surprise breakthrough from Chinese AI startup, Moonshot, rippling through global markets. We're seeing that play out. It started overnight in Asia. Semiconductor stocks, Charlie just talked about it, sharply lower.
We're thinking, is this another deep seek moment? You had President Xi Jinping also overnight, his first appearance at China's premier AI summit, lending his support to the country's push to become a global leader in AI. You've also got alphabet tanking. There's another story kind of at play here.

**Emily Graffeo** (1:06)
Right. Shares of alphabet tanking as Google is months behind on schedule on delivering Gemini 3.5 Pro, its most powerful flagship AI model. It's trying to improve its capabilities, particularly encodings.
Then the credit markets are also reacting to this AI trade. We got to talk about credit, Carol.

**Carol Massar** (1:23)
The bonds sold by hyperscalers to fuel their AI ambitions have become a drag on investor portfolios falling prices and wider spreads to negative total returns. That is certainly your world. I feel like it's a lot, but it is always when it comes to what's going on with AI. We wanted to check in once again with Mandeep Singh. He's Bloomberg Intelligence Global Head of Technology Research.
Mandeep, I feel like we have the same conversation with you over and over again, but there's stuff that keeps coming at us. SK Hynex ADRs too are bouncing around, the COSPI route. Feels like nervousness continues to grow increasingly out of this AI trade. Is it fundamentally based? Is it valuation based? Is it just a little bit of fear of people being on the wrong side of the trade?
You watch the fundamentals. You always remind us what matters. How do you see it?

**Mandeep Singh** (2:11)
Yeah, I would agree with your assessment that it is fundamentals that are still holding up and it's the valuations and the nervousness around positioning is what is driving this volatility. Look, we know AI being expensive has been raised as a concern by a lot of companies lately. It doesn't surprise me that a model like Kimi 3, which is an open rate model, much cheaper to use, is gaining traction. There are companies out there that have explored using this model, notably Cursor, which is a company that XAI and SpaceX bought.
They use Kimi 3 under the covers. From that perspective, open source will have some adoption, but fundamentally nothing has changed. Everyone wants to use LLMs. It's just that they were expensive. That's why we had concerns around memory pricing.
Now we're talking about open source adoption, but I don't think fundamentally anything has changed. It's just a function of how folks were positioned when it comes to the valuations and certain sectors have gone parabolic within the tech space and you're seeing a drawdown over there.

**Emily Graffeo** (3:31)
Mandeep, tell us about this new Chinese AI model. Because as Carol and I were discussing just a few minutes ago, it wasn't too long ago in markets that we had that deep seek moment, that scare when a new model from a different country comes out and spooks the market. Is this just a deep seek 2.0? Is there something different about Moonshot? And when you look at the market right now, is it really reflecting the same type of fear that we saw when the deep seek model was released a little over a year ago?

**Mandeep Singh** (4:02)
Yeah, I mean, look at how far we have come since that deep seek moment. We had a company like Anthropic grow from 9 billion revenue run rate to 47 billion run rate in six months. So that's despite deep seek. And so look, open source gaining traction is something that has carried forward. Yeah, I should stand corrected, it's open weight, not open source. But all these companies are looking for ways to reduce their, you know, inference pricing, which is very expensive for frontier models. And we have been compute constrained. Those are the things that reflect the fundamentals. Everything else around, you know, nervousness around deep seek and new types of models just shows that there is a lot of investment being made. Companies are exploring how to reduce their costs, but they want to consume more capacity and deploy as much of LLM functionality as they can. So from that perspective, the fundamentals are still good. And that's what's reflected in the revenue run rates of these companies and the demand that we have seen, you know, in terms of the backlog. And we'll hear more from hyperscalers when they report, you know, over the next two weeks in terms of how they're guiding around their capex numbers.