Skip to content
Qualcomm Wants Your AI on Device - DTNS 5359 artwork

Qualcomm Wants Your AI on Device - DTNS 5359

Daily Tech News Show

September 23, 2026

Jason Howell reports from the Snapdragon Summit on the new Snapdragon Elite Gen 6 models, plus most of the world feels good about AI, so why is the West such a downer? Starring Tom Merritt, Huyen Tue Dao, and Jason Howell. Links to stories discussed in this episode can be found here.

Speakers Tom Merritt, Huyen Tue Dao, Jason Howell

TopicsTechnologyNews

Tom Merritt (0:08)

This is the Daily Tech News for Wednesday, September 22nd, 2026 We tell you what you need to know, give you important context and try our dear darndest to help each other understand.

Huyen Tue Dao (0:20)

Well today, Jason Howell reports in from Qualcomm's Big Chip event in Hawaii. Are billions of parameters on your phone attractive to you?

Tom Merritt (0:29)

Hmm, we'll find out. I'm Tom Merritt.

Huyen Tue Dao (0:32)

We do it now.

Tom Merritt (0:33)

Let's start with what you need to know with that big story.

From Maui, Jason Howell has been at the annual Snapdragon Summit. Qualcomm flew you out there. I'm a little jealous. Are you not enjoying yourself, I hope?

Jason Howell (0:50)

No, it's horrible. I would so much rather be in my studio in Petaluma.

No, a big thanks to Qualcomm for sponsoring my travel out here. Any chance to go to Maui? I mean, hey, I'm all for it.

Huyen Tue Dao (1:04)

Yeah, I don't blame you.

Tom Merritt (1:05)

But you get front seat. Well, I don't know if it's front row, but you get to be at the big announcement in person. Tell us all about what the new Snapdragon's are that they announced yesterday afternoon.

Jason Howell (1:16)

And it's not at all confusing that they have the eight elite Gen 6 and the eight elite extreme Gen 6

Tom Merritt (1:22)

Now, one of them is extreme.

Jason Howell (1:24)

Yes, one of them is extremer. We'll talk about what that actually means. But yes, the eight elite Gen 6 is the kind of next tier of flagship processor. They say the first mobile CPU to hit five gigahertz. So, you know, crossing that, you know, that nice round number target.

But what does that actually mean? Because I was like, OK, if you have five gigahertz, like who cares? It's about bursts of speed, not necessarily constant speed. So things like faster app launches on your smartphone, reduced game stutter. But it doesn't mean that suddenly everything is just screaming fast on long, long, for long times.

They're focused on putting a lot more onto the chip to kind of get around the memory shortage that we find ourselves in. So on-device AI has to fit in four of five gigs of RAM somewhere around there. And so they're basically trying to do more with the chip to kind of offset the reliance on RAM.

And so that's the 8 Elite Gen 6 They announced a cool feature, which we've talked on this show about the Apple Watch and this ongoing transcription process or whatever out in the real world. They have something similar with this chip called Personal Scribe, which is opt-in, but it listens, indexes your conversations, identifies who's speaking, things like that. Seems like we're getting more and more of those features, so that's gonna be a real challenge.

Tom Merritt (2:52)

And that's an on-device thing, right? Yeah. Yeah, that's interesting.

Jason Howell (2:55)

Yeah, very, very interesting. But Wi-Fi 8, an improved modem. I mean, it's interesting being at an event like this because if you watch the keynote or whatever, it's so information-dense, but really what it comes down to is bigger, better, faster, stronger. It's always that way with these things. And to that end, there is the extreme version of the 8 Elite Gen 6, which is a tier above the regular flagship. As you can imagine, this is a first for Qualcomm breaking it out like this.

A faster GPU, more efficient, 40 percent more efficient, whatever. But it runs 3 billion parameter AI models on device. And I think Apple's on device is like a 20 billion parameter. So, there's a little bit of a... Or sorry, not three, 30 It was 30

Tom Merritt (3:46)

Apologies. It's a pretty big number.

Jason Howell (3:48)

Yeah, it's a huge number.

That was my mistake. 30 billion on device. They did kind of clarify because it's like, okay, is the race now to completely remove dependence on cloud, on frontier models in the cloud? And they're like, no, that's always going to be there for certain tasks. That's always going to be necessary. But it is about doing more on the device for a number of reasons, privacy, security, all the things. And yeah, just the fact that you're capable of storing that much on a chip is pretty crazy.

Tom Merritt (4:21)

Yeah, I think the fact that we've got two different chips here is also interesting because it's the same architecture, but they're squeezing more out of one versus the other. Were you able to tell what the actual difference is between these two chips?

Jason Howell (4:38)

30 more minutes of transcript below

Thousands of transcripts fetched by people building searchable podcast archives

Fetch the whole transcript

The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000791339634