# SpaceX Eyes Borrowing for Nvidia Chips, Apple Plans Smart Home Push

An episode of [Bloomberg Tech](https://invisiblenetwork.ai/podcast/1611767434), 7 Oct 2026, 44 min.

Audio: https://podtrac.com/pts/redirect.mp3/traffic.omny.fm/d/clips/e73c998e-6e60-432f-8610-ae210140c5b1/41764a4f-fc64-4e11-89ba-ae7c0030ab5e/acac6294-654d-4eec-b439-b4dc010ff44d/audio.mp3?utm_source=Podcast&in_playlist=9caafc41-289c-4115-995d-ae7c0030ab75

Bloomberg’s Ed Ludlow takes a look at SpaceX in talks with banks and investors to borrow $40 billion to buy chips from Nvidia. Plus, Apple plans its entry into the smart home space with a doorbell, thermostat, and other accessories developed with LG; and JD Moriarty, Vice Chair and head of Global Tech Equity Capital Markets at Bank of America Securities, discusses the upcoming mega AI IPO pipeline. See omnystudio.com/listener for privacy information.

## Transcript

[0:02] Speaker 1: Bloomberg Audio Studios. Podcasts.
[0:05] Speaker 2: Radio. News. Bloomberg Tech is live from the heart of
[0:16] Speaker 2: Silicon Valley with Ed Ludlow in San Francisco.
[0:22] Speaker 1: This is Bloomberg Tech. Coming up, the AI build-out keeps
[0:24] Speaker 1: driving record borrowing, with SpaceX in talks to borrow $ 40
[0:28] Speaker 1: billion to buy chips from NVIDIA. Plus, Apple plans its
[0:32] Speaker 1: entry into the smart home space with a doorbell, thermostat,
[0:35] Speaker 1: and other accessories developed with LG. And we'll discuss the
[0:39] Speaker 1: mega AI IPO pipeline with J.D. Moriarty, vice chair and
[0:43] Speaker 1: head of global tech equity capital markets at Bank of America.
[0:45] Speaker 1: of America Securities. Hello from the world of tech. So
[0:50] Speaker 1: stocks closed at records Tuesday, including the Nasdaq 100, very
[0:55] Speaker 1: tech heavy index. We're now hitting pause on that, largely
[0:57] Speaker 1: because of concerns about inflation. Oil is pushing higher. Inflation
[1:02] Speaker 1: is still the concern of the day. But there's a
[1:04] Speaker 1: lot in the news cycle. So let's get to our
[1:06] Speaker 1: top story and take a look at today's big number.
[1:09] Speaker 1: $ 40 billion. Bloomberg's reporting that SpaceX is in talks with
[1:13] Speaker 1: banks and investors to borrow $ 40 billion to buy chips
[1:17] Speaker 1: from NVIDIA, all according to sources. I want to get
[1:19] Speaker 1: the details with Bloomberg's chief Wall Street correspondent, Sridhar Nadarajan.
[1:23] Speaker 1: Let's start with the structure, right? There is a split
[1:25] Speaker 1: between loan and some debt and the players involved based
[1:30] Speaker 1: on Bloomberg's reporting.
[1:31] Speaker 3: Look, every other morning we seem to be waking up
[1:34] Speaker 3: to these big, unfathomable numbers almost. It's $ 60 billion one week,
[1:38] Speaker 3: $ 40 billion next week. It's in tens of billions of dollars.
[1:41] Speaker 3: This latest one involves SpaceX trying to finance these Nvidia
[1:45] Speaker 3: chips with a $ 40 billion package, $ 10 billion of it
[1:49] Speaker 3: to be syndicated through banks. $ 30 billion of it will
[1:52] Speaker 3: likely go through private market players led by Apollo. PIMCO
[1:55] Speaker 3: is possibly involved. They will syndicate it to the rest.
[1:58] Speaker 3: But it is very clear that the tech story today
[2:01] Speaker 3: is the debt story. And there is going to be
[2:03] Speaker 3: this increasing confluence between some of the biggest private market players,
[2:07] Speaker 3: some of the biggest firms with the deepest balance sheets,
[2:10] Speaker 3: and the tech companies that so desperately need to invest
[2:14] Speaker 3: and borrow so many hundreds of billions of dollars to
[2:18] Speaker 3: fund and finance this AI infrastructure build-out.
[2:22] Speaker 1: Sri, this morning, Bloomberg's Danny Berger had the opportunity to
[2:25] Speaker 1: speak to John Cortese, who's the co-head of corporate credit
[2:28] Speaker 1: at Apollo, one of the parties named in our reporting.
[2:32] Speaker 1: And she asked about this scoop.
[2:34] Speaker 4: Can't talk about SpaceX, but I think if you look at.
[2:39] Speaker 4: All of these companies that are the best companies in
[2:41] Speaker 4: the world, the largest companies in the world, the largest
[2:43] Speaker 4: parts of equity portfolios. It's not surprising that the largest
[2:46] Speaker 4: companies are looking at this marketplace and saying, we'd like
[2:49] Speaker 4: to find structured solutions for complicated build-outs or project build-outs,
[2:53] Speaker 4: AI being one of them, chip financing being one of them.
[2:55] Speaker 4: So it's a huge amount of that happening. I would
[2:57] Speaker 4: expect more of it. It's been very clearly stated to
[3:02] Speaker 4: the marketplace what the CapEx needs are to build that AI.
[3:05] Speaker 4: So I think you're going to see more and more.
[3:08] Speaker 1: We also got a warning signal through the CDS, right?
[3:11] Speaker 1: Five-year credit default swaps jumping to their highest level since
[3:14] Speaker 1: they started trading in June. But listening to that Apollo
[3:17] Speaker 1: non-response and the parties on the Wall Street side we're
[3:21] Speaker 1: talking about, there's an appetite to back these kinds of deals, right?
[3:25] Speaker 1: There's no real concern that the capital can't be found.
[3:29] Speaker 3: And the numbers show that. If you look at the
[3:31] Speaker 3: amount raised for data centers, for stacking these data centers
[3:35] Speaker 3: with chips, for the power required to power these data centers,
[3:38] Speaker 3: we've raised about $ 360 billion so far this year. That's
[3:42] Speaker 3: about three times the pace last year And if you
[3:45] Speaker 3: talk to analysts across the street, that number is only
[3:48] Speaker 3: going to keep going up in the next year or two.
[3:50] Speaker 3: We are talking about trillions of dollars in requirements, and
[3:53] Speaker 3: these companies will keep coming back to the market. Yes,
[3:56] Speaker 3: like you point out, with what is happening in the
[3:58] Speaker 3: broader rate environment, with some of the concerns starting to
[4:02] Speaker 3: seep in, if there is too much ebullience and too
[4:04] Speaker 3: much enthusiasm out there, you will see some of these
[4:07] Speaker 3: risk gauges starting to go up. SpaceX, you can see
[4:10] Speaker 3: that in the CDS a couple of weeks ago. We
[4:12] Speaker 3: talked about Oracle and we saw the moves in Oracle CDS.
[4:15] Speaker 3: But for the large part, these are not crazy blowouts.
[4:18] Speaker 3: These are contained moves. And for the biggest, most credit
[4:22] Speaker 3: worthy players, there will be demand. Because when you listen
[4:25] Speaker 3: to what else people like the senior executives at Apollo
[4:28] Speaker 3: have to say, they're not really trying to make a
[4:30] Speaker 3: bet on whether these chips will depreciate in value or
[4:33] Speaker 3: hold their value. They are basically betting on the underlying
[4:36] Speaker 3: credit worthiness of the borrower. And as long as that
[4:39] Speaker 3: is solid, they are willing to extend their balance sheet
[4:41] Speaker 3: to them.
[4:43] Speaker 1: Bloomberg Shridhar Natarajan, thank you very much. Now, staying with SpaceX,
[4:46] Speaker 1: its Crew 12 is heading home. The four-person crew undocked
[4:50] Speaker 1: from the International Space Station this morning after spending nearly
[4:54] Speaker 1: eight months in orbit. Their Dragon capsule is now on
[4:58] Speaker 1: a 27-hour journey back to Earth. Splashdown is set for
[5:02] Speaker 1: Thursday evening. in the Pacific, and we will track the
[5:06] Speaker 1: progress of that mission. Another Musk story. Elon Musk setting
[5:09] Speaker 1: the record straight on his chip-making ambition, saying there should be, quote,
[5:14] Speaker 1: zero doubt about who's in charge of that initiative. Earlier
[5:18] Speaker 1: this week, Musk confirmed discussions with TSMC over his planned
[5:21] Speaker 1: terror fab in Texas, which raised questions about Intel's role
[5:25] Speaker 1: and if they'd be displaced. Now, Intel CEO Lit Bhutan,
[5:29] Speaker 1: speaking overnight in Tokyo, said, says his company is still
[5:32] Speaker 1: there as a technology partner, while Musk is saying TSMC
[5:36] Speaker 1: won't be running that fab. Bloomberg's global tech editor, Peter Elstrom,
[5:40] Speaker 1: is with us. The twists and turns. Try and summarize
[5:44] Speaker 1: where we've landed.
[5:46] Speaker 5: Yeah, let's pause for a second. Musk talked about TerraFab
[5:49] Speaker 5: back in March. And at the time, it was unclear
[5:52] Speaker 5: exactly where he was going to go with this venture.
[5:54] Speaker 5: He talked about it as an epic adventure in chip building.
[5:58] Speaker 5: He said that there was such a shortage in the
[6:00] Speaker 5: chip industry that he was going to get into the
[6:02] Speaker 5: business of making the chips himself. He said it was
[6:04] Speaker 5: important enough that they wanted to get into the kind
[6:06] Speaker 5: of logic chips that TSMC makes, and they're going to
[6:09] Speaker 5: make memory chips at the same time. And the vision
[6:11] Speaker 5: he laid out was really quite exceptional at that point.
[6:15] Speaker 5: It was called a fever dream by some people. And
[6:18] Speaker 5: one analyst estimated that it would cost $ 5 trillion to
[6:20] Speaker 5: $ 13 trillion to be able to build out this whole project.
[6:23] Speaker 5: So early on, they brought Intel on as the partner
[6:26] Speaker 5: to help them. If you don't have this chip expertise,
[6:29] Speaker 5: you need to be able to use the IP from someplace. Intel,
[6:32] Speaker 5: of course, has that kind of expertise. They began working together.
[6:36] Speaker 5: But then, as you mentioned, there was this report that
[6:38] Speaker 5: TSMC was somehow involved. In fact, Musk confirmed that there
[6:41] Speaker 5: were talks at that point without going into any kind
[6:44] Speaker 5: of detail. So, as you mentioned, we talked with Luke Buten,
[6:46] Speaker 5: the CEO of Intel Overnight in Tokyo. He confirmed they're
[6:49] Speaker 5: still working with Musk and Terafab at this point. They
[6:52] Speaker 5: plan on continuing that work. That's a positive for Intel.
[6:56] Speaker 5: And then also Musk responded to something that suggested TerraFab
[7:01] Speaker 5: would never actually own the fabs. In fact, it would
[7:03] Speaker 5: be TSMC that would build the fabs and own the fabs.
[7:05] Speaker 5: And he said, that's not at all the case. TerraFab,
[7:08] Speaker 5: Musk and his team are going to build these fabs themselves.
[7:11] Speaker 5: TSMC may be involved somehow, but they're not going to
[7:14] Speaker 5: own the property.
[7:15] Speaker 1: I think it's really useful to go back to the
[7:18] Speaker 1: basics of what TerraFab is, right? It's a joint venture
[7:21] Speaker 1: between SpaceX and and Tesla to currently separate companies. But
[7:26] Speaker 1: I think Musk has always talked about the idea that
[7:29] Speaker 1: there isn't enough chip supply for the things that both
[7:31] Speaker 1: those companies want to do. So, Musk's solution, build it themselves.
[7:36] Speaker 1: Is that kind of a fair sum?
[7:38] Speaker 5: Yeah, that's exactly it.
[7:40] Speaker 6: Yeah.
[7:40] Speaker 5: And he said at the time back in March, he
[7:42] Speaker 5: said in order to accomplish what he wants to accomplish
[7:44] Speaker 5: with rockets and robots and autonomous cars, They need chips.
[7:50] Speaker 5: And he said if the chip industry is not gonna
[7:52] Speaker 5: be able to supply those chips, and he thinks it
[7:54] Speaker 5: doesn't have that capacity right now, then they have to
[7:57] Speaker 5: build it themselves. So it sounded like they were pushing
[8:00] Speaker 5: into this business partly to motivate the chip industry, to
[8:04] Speaker 5: move more quickly, to expand more quickly. Now, the other
[8:06] Speaker 5: side of this, of course, is the chip industry is
[8:08] Speaker 5: terribly cyclical. These companies invest billions and billions of dollars,
[8:12] Speaker 5: and then they find themselves with too much capacity. They've
[8:14] Speaker 5: done this again and again over the years. So they've
[8:16] Speaker 5: been a little bit cautious about how quickly they've expanded,
[8:19] Speaker 5: especially on the memory chip front. But now Musk is
[8:21] Speaker 5: trying to push that a little bit faster. Talk about
[8:24] Speaker 5: capital needs. I mean, they may need to raise billions
[8:26] Speaker 5: of dollars, tens of billions of dollars, if not trillions
[8:28] Speaker 5: of dollars, to be able to accomplish some of the
[8:30] Speaker 5: goals he has for TerraFab.
[8:32] Speaker 1: Bloomberg's Peter Ahlstrom. Thank you. SpaceX is pushing the Trump
[8:37] Speaker 1: administration to make it easier for rocket companies to get
[8:40] Speaker 1: access to bands of spectrum, a move that could help
[8:43] Speaker 1: increase launches. But those efforts are facing pushback from rivals
[8:47] Speaker 1: in the aerospace industry. So Boeing, Lockheed Martin and GE
[8:50] Speaker 1: Healthcare are among those warning the changes could disrupt everything
[8:54] Speaker 1: from flight tests of commercial jets and missiles to hospital
[8:59] Speaker 1: equipment and and even TV broadcasts. The Federal Communications Commission,
[9:03] Speaker 1: which is being asked to scrap a third-party entity that
[9:06] Speaker 1: controls how spectrum is shared, did not respond to requests
[9:10] Speaker 1: for comment. Coming up here on Bloomberg Tech, Apple's push
[9:14] Speaker 1: into smart home devices includes an unusual partnership. We've got
[9:19] Speaker 1: the details next. This is Bloomberg Tech. Apple's upcoming push
[9:31] Speaker 1: into smart home devices will include a doorbell, thermostat, and
[9:35] Speaker 1: other accessories developed through an unusual partnership with LG Electronics.
[9:39] Speaker 1: The products will be part of an ecosystem that works
[9:42] Speaker 1: with Apple's new smart home hub, according to Bloomberg sources.
[9:46] Speaker 1: Let's get the details with Bloomberg's Mark Gurman, who leads
[9:48] Speaker 1: our coverage of consumer technology. So interesting to read this one.
[9:52] Speaker 1: And we'll bring the diagrams back up that were in
[9:54] Speaker 1: your story. But Apple is partnering on this range of devices.
[10:01] Speaker 7: So I've been reporting for years now that Apple has several...
[10:04] Speaker 7: first party in-house smart home devices coming. This is part
[10:08] Speaker 7: of a gigantic reboot to its home strategy. HomeKit launched
[10:12] Speaker 7: in 2014 has been an unmitigated disaster compared to what
[10:16] Speaker 7: you're seeing from Ring, from what you're seeing from Google,
[10:19] Speaker 7: from what you're seeing from others. So Apple is starting over.
[10:21] Speaker 7: And so they'll announce a new HomePod mini, a smart
[10:24] Speaker 7: home hub with a display that's like an Amazon Echo Show,
[10:26] Speaker 7: and a new Apple TV. But yesterday, I revealed there's more.
[10:30] Speaker 7: In addition to the flagship devices, Apple's basically coming out
[10:33] Speaker 7: with what I'm calling a smart home starter kit. If
[10:36] Speaker 7: you're building a smart home, what would you want? You
[10:38] Speaker 7: would want a smart lock. You'd want a smart thermostat.
[10:41] Speaker 7: You'd want a smart doorbell. And you would want smart
[10:43] Speaker 7: security cameras inside, outside. It may be a floodlight camera.
[10:47] Speaker 7: So they're coming out with all of that. And they're
[10:49] Speaker 7: partnering with LG on it. according to my sources. And
[10:53] Speaker 7: I searched to the FCC database and found that these
[10:57] Speaker 7: LG branded devices are in midst of getting regulatory approval
[11:00] Speaker 7: in the United States.
[11:03] Speaker 1: Okay. So how unusual is the step for Apple to say,
[11:06] Speaker 1: go to another company and be like, you guys are
[11:07] Speaker 1: good at this and we're good at this and we
[11:10] Speaker 1: can work together. And then the big picture of why
[11:13] Speaker 1: a lock and a doorbell and a thermostat works with
[11:18] Speaker 1: what Apple's trying to do in the home.
[11:21] Speaker 7: It's extremely unusual, but it's not unprecedented. They've done this
[11:25] Speaker 7: a couple times. They've built devices with Belkin and put
[11:28] Speaker 7: them under the Belkin brand before. They've built some iPad
[11:32] Speaker 7: accessories with Logitech and put it under the Logitech brand before.
[11:35] Speaker 7: But this LG one is kind of crazy because LG
[11:37] Speaker 7: doesn't make smart home devices, right? Like this is new
[11:40] Speaker 7: territory completely for LG. Belkin has made iPhone accessories for years.
[11:44] Speaker 7: Logitech has made iPad accessories for years. I mean, LG
[11:47] Speaker 7: is basically completely out of left field and completely random.
[11:50] Speaker 7: This is why this news was so shocking. They don't
[11:52] Speaker 7: make a smart home devices. They don't make a smart
[11:55] Speaker 7: cameras for the home. They don't make door bells. They
[11:57] Speaker 7: don't make locks. They don't make any of this stuff. Uh,
[12:00] Speaker 7: so it's fascinating. And Apple and LG have a big relationship, right? LG,
[12:04] Speaker 7: the sister companies, LG electronics, LG display, make screens for
[12:07] Speaker 7: Apple devices. Uh, one of the pieces of the LG
[12:10] Speaker 7: conglomerate make camera modules. So there is some relationship there. And, uh,
[12:15] Speaker 7: To me, these look pretty nice. And like I said earlier,
[12:17] Speaker 7: if you're wanting to build a smart home as a consumer,
[12:21] Speaker 7: you're going to start with a doorbell. You're going to
[12:23] Speaker 7: start with a lock. You're going to start with a thermostat.
[12:25] Speaker 7: And so that's why Apple's going after these products.
[12:28] Speaker 1: Bloomberg's Mark Gurman. And don't forget to read Power On,
[12:30] Speaker 1: his newsletter, and listen to the new podcast, Power On.
[12:34] Speaker 1: Thank you very much. Meanwhile, the leader in the smart
[12:37] Speaker 1: doorbell market, Ring, is out with its own new products.
[12:40] Speaker 1: The Amazon-owned home security company releasing its first smart lock
[12:44] Speaker 1: and new default security encryption for users' data. Delighted to
[12:48] Speaker 1: say that Ring founder, Jamie Siminoff, joins us now. Let's
[12:51] Speaker 1: start with what's new here. Lots of the audience familiar
[12:54] Speaker 1: with Ring, the doorbell. Goes beyond that now with a
[12:57] Speaker 1: range of things.
[12:58] Speaker 8: Yeah, so we have a large range of products from
[13:01] Speaker 8: outdoor cameras, indoor cameras, alarm system. We started with the doorbell,
[13:05] Speaker 8: and now I call it the peanut butter and jelly.
[13:07] Speaker 8: I mean, the lock and the doorbell have always, we thought,
[13:09] Speaker 8: come together, but we didn't have an angle on the
[13:12] Speaker 8: lock or an innovation on the lock that we thought
[13:14] Speaker 8: warranted us putting our brand on it until now.
[13:18] Speaker 1: Jamie, let's talk about that innovation. There are, as Mark's
[13:22] Speaker 1: kind of outlined, an increasing number of options of smart
[13:27] Speaker 1: home devices. What is it that you're good at? What
[13:29] Speaker 1: is it that you think you're better at on the
[13:31] Speaker 1: hardware side, the software side relative to the competition?
[13:35] Speaker 8: First of all, we start everything around a mission to
[13:38] Speaker 8: make neighborhoods safer. I think with that North Star that
[13:41] Speaker 8: really drives our business into a direction that so far
[13:44] Speaker 8: no one has been able to follow. Secondly, I'm an inventor.
[13:47] Speaker 8: I've invented the smart doorbell, the floodlight camera, and many
[13:51] Speaker 8: other innovations around the home. including today with the smart
[13:54] Speaker 8: lock that now has a rechargeable power.
[13:59] Speaker 1: Dial on it.
[13:59] Speaker 8: So that if the smart, I mean, you know, when
[14:01] Speaker 8: a smart lock loses its batteries, it becomes a not
[14:03] Speaker 8: smart lock or a dumb lock.
[14:04] Speaker 1: Right.
[14:05] Speaker 8: And so, you know, now instead of having to have
[14:07] Speaker 8: a backup key that's hidden in the area where the
[14:09] Speaker 8: spiders are around your house, you literally just twist this
[14:12] Speaker 8: power dial. It comes back on and now you get in.
[14:15] Speaker 8: And so that's, you know, Ring really mission to make
[14:17] Speaker 8: neighborhoods safer. We're very innovative. I consider myself an inventor.
[14:21] Speaker 8: And that's what we've been able to do so far,
[14:23] Speaker 8: which is, I think, how we've become what I believe
[14:25] Speaker 8: is the largest home security company in the world.
[14:27] Speaker 9: Right.
[14:28] Speaker 1: Jamie, I have a smart doorbell from from another brand.
[14:31] Speaker 1: But the main point I'm getting at is that it's
[14:34] Speaker 1: part of an integrated system of multiple devices of different types.
[14:38] Speaker 3: Right.
[14:39] Speaker 1: What kind of data do you see about a consumer
[14:41] Speaker 1: that buys one Ring product in aggregate with lots of others, right?
[14:46] Speaker 1: They're kind of bulk buyers of systems rather than just
[14:48] Speaker 1: one piece of it.
[14:49] Speaker 8: Yeah, certainly it's the rings of security and everything you
[14:52] Speaker 8: can buy from Ring works on its own or works
[14:54] Speaker 8: together sort of on the app and in the ecosystem.
[14:57] Speaker 1: It also works with Alexa Plus.
[14:59] Speaker 8: So we do see that people, you know, they certainly
[15:01] Speaker 8: starting at the doorbell as your first ring of security,
[15:04] Speaker 8: working your way around alarm system, outdoor cameras, some indoor cameras.
[15:09] Speaker 8: smart lock now, you know, we really help people to
[15:11] Speaker 8: build the security they need and also, you know, choose
[15:14] Speaker 8: your power. I mean, one of the big things is,
[15:16] Speaker 8: you know, we offer batteries, solar, and powered options in
[15:19] Speaker 8: basically every camera that we have.
[15:23] Speaker 1: Jamie, you heard Mark Gurman's reporting there on Apple's entry
[15:26] Speaker 1: into this category, which includes a smart lock and thermostat,
[15:31] Speaker 1: et cetera, based on his reporting. How do you react
[15:34] Speaker 1: to that? What's your feeling about Apple coming into the space?
[15:37] Speaker 8: I've had competitors since day one. I've stuck to my
[15:40] Speaker 8: mission of making neighborhoods safer, and so far that's paid
[15:42] Speaker 8: dividends for us. So we're just going to keep focusing
[15:44] Speaker 8: on our neighbors.
[15:45] Speaker 1: Which is what we call our customers.
[15:46] Speaker 8: I'm going to keep innovating, bringing new-to-world products, and I
[15:50] Speaker 8: feel like if I do that, I'll remain successful and
[15:52] Speaker 8: keep being the leader in the market.
[15:55] Speaker 1: Here. I'm genuinely interested in the economics of consumer electronics, right?
[16:01] Speaker 1: Especially when you push into getting electronics getting smaller and
[16:04] Speaker 1: smaller and smaller. And Ring within Amazon, we showed... In fact, Tim,
[16:09] Speaker 1: put the table back up with the price points for
[16:11] Speaker 1: this generation of Ring gear.
[16:13] Speaker 9: Like.
[16:15] Speaker 1: How do you get that right, the pricing? Is it
[16:17] Speaker 1: a benefit to have Amazon scale that you can price
[16:20] Speaker 1: such as you do? Is this a profitable enterprise?
[16:24] Speaker 8: Certainly Ring is very profitable at this point. And we also,
[16:27] Speaker 8: we are big enough now in terms of scale that yes,
[16:30] Speaker 8: certainly being at Amazon is not a detriment, but we're
[16:33] Speaker 8: big enough now that we can get scale in the
[16:35] Speaker 8: pieces and the things that we buy. And we're very
[16:38] Speaker 8: innovative in how we are able to choose the chips,
[16:41] Speaker 8: build the products, build them from scratch, try to use
[16:43] Speaker 8: first principles effectively. and really build the best products at
[16:46] Speaker 8: the best cost to democratize home security for everyone.
[16:50] Speaker 1: Jamie, what happens next in this field? How does the
[16:53] Speaker 1: smart home change in the future, to your mind?
[16:56] Speaker 8: I mean, with AI and superintelligence, certainly the smart home
[16:59] Speaker 8: is going to continue to change. I believe we're the
[17:03] Speaker 8: appendages for AI and superintelligence. It's going to need to
[17:07] Speaker 8: have the physical, it's going to need to have a lock,
[17:09] Speaker 8: it's going to need to have a doorbell, it's going
[17:10] Speaker 8: to need to have cameras, it's going to need to
[17:11] Speaker 8: be able to see and sense things And when you
[17:13] Speaker 8: put all of that into a superintelligence, I do believe
[17:15] Speaker 8: that that is going to superpower our ability in the home,
[17:20] Speaker 8: our peace of mind in the home, and doing things
[17:22] Speaker 8: for us. I mean, in the end of the day,
[17:24] Speaker 8: it doesn't matter the technology. It's how is it better
[17:27] Speaker 8: for us to live in our homes when we have
[17:28] Speaker 8: all this stuff? And I do believe the age of
[17:31] Speaker 8: abundance is ahead of us, and we're going to have,
[17:32] Speaker 8: you know, around our area, we're going to be living
[17:35] Speaker 8: in our homes safer, more secure, and it's going to
[17:37] Speaker 8: be more peace of mind.
[17:39] Speaker 1: Ring founder, Jamie Siminoff. Thank you very much for your time.
[17:42] Speaker 5: On the show.
[17:44] Speaker 1: Okay, coming up, the hidden risks of open source AI.
[17:48] Speaker 1: Why powerful models running offline are becoming harder to police.
[17:53] Speaker 1: We have an important piece of Bloomberg reporting next. This
[17:55] Speaker 1: is Bloomberg Tech. The rise of open source AI is
[18:08] Speaker 1: creating a new challenge for law enforcement. Once powerful models
[18:12] Speaker 1: are downloaded and run privately, their safeguards can be stripped away.
[18:16] Speaker 1: And investigators say they're increasingly being used to create illegal
[18:20] Speaker 1: child sexual abuse material that can be nearly impossible to track.
[18:25] Speaker 1: It's the topic of today's Big Take. And Bloomberg's Big
[18:27] Speaker 1: Tech editor, Sarah Fryer, joins us now. Sarah, you edited
[18:31] Speaker 1: this story with the team. We talk a lot about
[18:34] Speaker 1: the risk of open source AI. And the benefits of
[18:37] Speaker 1: open source also on this program, but very simply, this
[18:40] Speaker 1: story explains how open source tech makes illegal activity easier
[18:46] Speaker 1: for human users. Can you explain why?
[18:49] Speaker 10: Well, once it's on somebody's personal device, there's no commercial
[18:53] Speaker 10: entity that it's interacting with. You can train it for
[18:56] Speaker 10: whatever purpose you have, and then it can just run locally.
[19:01] Speaker 10: And so that makes it very difficult for investigators to
[19:04] Speaker 10: find this illegal activity and stop it. And it also
[19:09] Speaker 10: means that these communities of people who engage in child
[19:13] Speaker 10: exploitative imagery, they're passing notes with each other on forums
[19:18] Speaker 10: talking about how to use open source tech to make
[19:21] Speaker 10: this content and sharing their models for people to download
[19:25] Speaker 10: on their own devices and create unlimited amounts, which is
[19:31] Speaker 10: quite concerning.
[19:33] Speaker 1: The story details case studies of child exploitation crimes, right?
[19:37] Speaker 1: And in the reporting, in talking to investigators, what have
[19:42] Speaker 1: those investigators, and I guess the industry, noticed about how
[19:46] Speaker 1: AI is changing and accelerating, frankly, what is harm to miners?
[19:52] Speaker 10: The limit to what can be created And the amount
[19:57] Speaker 10: of content is just the perpetrator's imagination. And they can
[20:04] Speaker 10: ask AI to take an image of a child that
[20:07] Speaker 10: is just a normal photo and decide to create whatever
[20:12] Speaker 10: kind of results they want, possibly an illegal result in
[20:17] Speaker 10: sexual positions. So what we're seeing here is that for investigators,
[20:22] Speaker 10: this flood of images, the dramatic increase in the amount
[20:26] Speaker 10: of illegal content that is being created is making it
[20:28] Speaker 10: very difficult for them to figure out of these children
[20:32] Speaker 10: we see. First of all, who are the real children?
[20:34] Speaker 10: Who are the generated images of real children? And then
[20:39] Speaker 10: who might be in danger today? Who might be in
[20:42] Speaker 10: physical danger in their communities that we need to help save?
[20:46] Speaker 10: And it's just becoming more difficult than ever.
[20:50] Speaker 1: Sarah, can anything be done to remedy it?
[20:53] Speaker 10: Well, it's very difficult to fix the problem in this space,
[20:57] Speaker 10: in part because as models are doing their red teaming
[21:01] Speaker 10: and trying to figure out how these tools could be
[21:04] Speaker 10: used for illegal activity, they cannot themselves create this kind
[21:09] Speaker 10: of content. It's just too illegal to do.
[21:13] Speaker 1: If you're a company. However.
[21:15] Speaker 10: There are some investigators that are trying to figure out
[21:17] Speaker 10: if they can determine which models have been fine-tuned for
[21:21] Speaker 10: this purpose and trained for this purpose to prevent them
[21:25] Speaker 10: from coming on to the model of sharing groups like
[21:30] Speaker 10: Hugging Face so they never get uploaded so that other
[21:33] Speaker 10: people can't use them.
[21:34] Speaker 1: And what Sarah was just talking about, red teaming, so interesting.
[21:37] Speaker 1: That's the sort of adversarial testing process where humans or
[21:40] Speaker 1: sometimes AI agents try to expose those vulnerabilities in the
[21:45] Speaker 1: underlying model. All of that detailed in the big take.
[21:48] Speaker 1: Bloomberg's Sarah Farrar, who edited the big take and led
[21:51] Speaker 1: the team that reported it out. Thank you very much indeed. Okay,
[21:55] Speaker 1: coming up in the show, we'll get the credit reaction
[21:57] Speaker 1: from Bloomberg Intelligence on our top story, which was SpaceX's
[22:01] Speaker 1: potential $ 40 billion Nvidia chip financing package split across loan
[22:05] Speaker 1: and debt. It's a big one in the market. See
[22:07] Speaker 1: the equities involved in that story. SpaceX, but also Nvidia
[22:11] Speaker 1: kind of under pressure. Remember Tuesday, technology stocks, but stocks
[22:16] Speaker 1: in general closed at record highs on the S &
[22:18] Speaker 1: P 500 and NASDAQ 100. We're hitting the pause button
[22:22] Speaker 1: on that a little bit. Particular underperformance in semiconductors, the
[22:26] Speaker 1: SOX or Philadelphia Semiconductor Index down 2%. And again, we're
[22:29] Speaker 1: going to go back into that story. about SpaceX. Also
[22:34] Speaker 1: in the background, the world of IPOs. There's a lot
[22:36] Speaker 1: more to come. It's halftime here in the city of
[22:39] Speaker 1: San Francisco. Thanks for being with us. You're watching Bloomberg Tech.
[23:01] Speaker 1: Welcome back to Bloomberg Tech. Let's get back to today's
[23:03] Speaker 1: top story, which is SpaceX in talks to borrow $ 40
[23:06] Speaker 1: billion to buy NVIDIA chips. From the equity part of
[23:10] Speaker 1: the story, the stock under pressure down about 2%. Bloomberg
[23:13] Speaker 1: Intelligence already out with a credit react saying the deal
[23:16] Speaker 1: underscores the substantial pools of capital available to fund a
[23:20] Speaker 1: broader AI build out. Bloomberg Intelligence analyst Robert Schiffman joins
[23:24] Speaker 1: us with that react. I saw the CDS chart, right?
[23:28] Speaker 1: But at the end of the day, you and I
[23:30] Speaker 1: have been talking regularly and for some time now that
[23:33] Speaker 1: the capital is there and there's a willingness from capital
[23:36] Speaker 1: markets to support what's going on. Just take me through
[23:39] Speaker 1: the react.
[23:41] Speaker 6: Yeah, listen, quite frankly, CDS has barely budged and bonds
[23:44] Speaker 6: have barely budged. We have told people to buckle up.
[23:48] Speaker 6: There is a lot of this to come. And quite frankly,
[23:51] Speaker 6: today's announcement, or at least last night's announcement, I think
[23:54] Speaker 6: is wildly positive. Coming up with creative ways to finance
[23:58] Speaker 6: balance sheets that are just not straight, unsecured debt. Make
[24:02] Speaker 6: a ton of sense. And I think having partners like Apollo, NVIDIA,
[24:06] Speaker 6: PIMCO helping potentially to securitize chips to help NVIDIA raise
[24:11] Speaker 6: $ 40 billion is fantastic. Quite frankly, they may have a
[24:15] Speaker 6: couple hundred billion dollars to raise over the next few years.
[24:18] Speaker 6: So I think this sort of creativity is good for
[24:20] Speaker 6: the market. And I think demand for bonds like this
[24:23] Speaker 6: are going to be through the roof.
[24:26] Speaker 1: Robert Schiffman of Bloomberg Intelligence. Check out his react on
[24:28] Speaker 1: the Bloomberg terminal. Thank you very much. Okay, two of
[24:31] Speaker 1: the biggest names in artificial intelligence are moving toward the
[24:33] Speaker 1: public markets, but in different timelines. And the potential scale
[24:37] Speaker 1: of open AI and anthropic could test investor appetite for
[24:41] Speaker 1: AI like never before.
[24:42] Speaker 9: J.D.
[24:43] Speaker 1: Moriarty is Vice Chair and Head of Global Technology, Equity
[24:45] Speaker 1: Capital Markets at Bank of America Securities. He's in San
[24:48] Speaker 1: Francisco for Bank of America's 15th Annual Technology Innovation Summit, where...
[24:53] Speaker 1: AI is the anchor of these conversations, one would suspect. JD,
[24:57] Speaker 1: what do you think everyone's going to be talking about
[25:00] Speaker 1: in town this week?
[25:01] Speaker 2: I think you just told us, and I think you're
[25:03] Speaker 2: probably right. The majority of the discussion will be certainly
[25:06] Speaker 2: around AI. The conference has been going on for, this
[25:09] Speaker 2: is the 15th year that we've done it, and the
[25:12] Speaker 2: Tech Innovation Summit is intended to give companies an opportunity
[25:16] Speaker 2: to talk to our technology department, to better understand our needs,
[25:20] Speaker 2: and for us to discover interesting companies along the way, right?
[25:23] Speaker 2: So it's really, this is a conference not put on
[25:26] Speaker 2: by investment banking, but put on by our technology department.
[25:30] Speaker 2: And there's been an unbelievable track record of many companies
[25:33] Speaker 2: that you're aware of today that, you know, 10 years
[25:35] Speaker 2: ago were at this conference and became partners of ours.
[25:38] Speaker 1: Just real quick on the... corporate side, the use of
[25:42] Speaker 1: AI in your industry, is there a good economic proposition?
[25:46] Speaker 1: When you're offered different tools, do you think the banks
[25:48] Speaker 1: look at it and say, yes, it's worth spending money on?
[25:52] Speaker 1: We see a benefit for this.
[25:53] Speaker 2: Yeah.
[25:53] Speaker 1: Listen, I think that's evolving.
[25:56] Speaker 2: I think the definition of that economic proposition will continue
[26:00] Speaker 2: to evolve. Certainly in terms of You can't afford to
[26:04] Speaker 2: wait for that economic proposition to be totally clear given
[26:08] Speaker 2: the rate of change and given what all of our
[26:10] Speaker 2: competitors are doing. But even just this year, in terms
[26:13] Speaker 2: of our internal utilization, we've seen that play out. The
[26:20] Speaker 2: narrative is going to evolve in terms of the use
[26:24] Speaker 2: case for all the major guys that you talk about
[26:26] Speaker 2: in the context of the prospective IPOs and then how
[26:29] Speaker 2: vertically oriented it will be as well. But it's certainly
[26:32] Speaker 2: helping us in terms of internal efficiency.
[26:34] Speaker 1: Okay, so those prospective IPOs, Bloomberg's reporting is that Anthropic's
[26:39] Speaker 1: timeline is to try and go before Thanksgiving. But as
[26:42] Speaker 1: a case study, right, Anthropic or OpenAI, how is it different?
[26:47] Speaker 1: to IPOs in the past in terms of what those
[26:50] Speaker 1: companies need to say to investors to help them understand
[26:55] Speaker 1: their futures. You know, I'm not necessarily even talking about financials.
[26:58] Speaker 1: I'm talking about, yes, their capital needs in the future,
[27:02] Speaker 1: but also like this idea of the risk disclosures being
[27:06] Speaker 1: completely novel, you know, a threat to human life existentially.
[27:09] Speaker 6: Yeah.
[27:10] Speaker 1: That's a hard one to put into a document.
[27:13] Speaker 2: It is rare that we're having a conversation in the
[27:15] Speaker 2: media about risk factors, for sure.
[27:20] Speaker 1: I wouldn't look at those two.
[27:23] Speaker 2: Prospective IPOs as indicative of the broader IPO market. Obviously,
[27:27] Speaker 2: we're dealing with something that's, whether it's SpaceX or Anthropic
[27:29] Speaker 2: or OpenAI, very different with respect to just a sheer scale, right?
[27:33] Speaker 9: Yes.
[27:34] Speaker 2: And recognize that the IPO market has evolved a great
[27:37] Speaker 2: deal because of the abundance of private capital. And crossover
[27:41] Speaker 2: investors that are already in each of SpaceX before its IPO,
[27:46] Speaker 2: in OpenAI already, Anthropic already. So recognize that institutionally, those
[27:51] Speaker 2: companies are very well known by the buy side long
[27:54] Speaker 2: before the IPO process.
[27:56] Speaker 1: May I give you a case study?
[27:58] Speaker 9: Sure.
[27:58] Speaker 1: Emma Norchette from T. Rowe Price was on the show yesterday.
[28:02] Speaker 1: She does private investments. She explained they have multibillion-dollar exposure
[28:07] Speaker 1: to both open AI and anthropic, but that part of
[28:10] Speaker 1: their underlying thesis is that there is room for both
[28:12] Speaker 1: of them at the frontier in the world in the end. Now,
[28:15] Speaker 1: from the public market investor's side, from an IPO story's perspective,
[28:20] Speaker 1: is there room for two of these things to happen
[28:23] Speaker 1: in proximity? Three, if you go back to June and SpaceX.
[28:27] Speaker 2: Well, we certainly started the year, as we talked about SpaceX,
[28:30] Speaker 2: and everybody in our world was questioning exactly that. What
[28:35] Speaker 2: would be the impact on the market with the SpaceX IPO?
[28:39] Speaker 2: We were looking at how much cash was on the sidelines.
[28:41] Speaker 2: What would need to be sold to buy that IPO?
[28:44] Speaker 2: And as a reminder, just as that IPO went on
[28:48] Speaker 2: the road, Alphabet came and raised a huge amount of
[28:51] Speaker 2: capital in the public market.
[28:53] Speaker 5: Right.
[28:53] Speaker 1: And it was a blip, right?
[28:56] Speaker 2: The market absorbed it quickly, just as the market has
[28:58] Speaker 2: absorbed and rewarded SpaceX, right? So there's less conversation about
[29:03] Speaker 2: capacity today.
[29:04] Speaker 1: That's so interesting because you heard what Robert was discussing,
[29:09] Speaker 1: this Bloomberg report around SpaceX in talks to borrow $ 40 billion.
[29:14] Speaker 1: But what was amazing about the period post-SpaceX IPO was
[29:18] Speaker 1: the speed at which they returned to markets. Again, Anthropic
[29:24] Speaker 1: and OpenAI, they go next. But do investors accept that
[29:28] Speaker 1: that's the operating environment?
[29:30] Speaker 2: I think investors certainly accept that the traditional construct of
[29:35] Speaker 2: you go public and you don't do anything for six
[29:37] Speaker 2: months for these companies is irrelevant. Right.
[29:40] Speaker 1: Serve them well.
[29:49] Speaker 2: In fact, just this morning, we were having a discussion
[29:51] Speaker 2: with our colleagues in our investment grade area about the
[29:55] Speaker 2: amount of CapEx that the hyperscalers project. And if you
[29:58] Speaker 2: look at their behavior in 25 and 26, where essentially
[30:03] Speaker 2: their CapEx projections went up by 20% in 25 and
[30:06] Speaker 2: remains to be seen on 26 final numbers, but if
[30:09] Speaker 2: you project that out to 2030, it projects a staggering
[30:13] Speaker 2: amount of CapEx that will be supported by the investment
[30:16] Speaker 2: grade market And historically, about half of that capex has
[30:21] Speaker 2: been there. That will have huge implications for the high-grade index.
[30:27] Speaker 2: So additional pools of capital are actually very good for
[30:31] Speaker 2: all of those companies and for the functioning of the market.
[30:34] Speaker 2: So if you just look at it, it's timely that
[30:37] Speaker 2: you're seeing that announcement because what is going on in
[30:42] Speaker 2: that market to fund the hyperscaler capex has implications for
[30:47] Speaker 2: all of our markets. Our markets are more connected than
[30:49] Speaker 2: they've ever been.
[30:49] Speaker 1: I think the challenge of understanding what's going on is,
[30:53] Speaker 1: you know, Daniela Amodei said herself at a Bloomberg event
[30:56] Speaker 1: in June, the IPO, the rationale is that's the mechanism
[30:59] Speaker 1: to raise capital at the levels you need. But then Alphabet,
[31:03] Speaker 1: as you said, was able to do it in the
[31:06] Speaker 1: equity markets quite willingly. I guess let's end our conversation
[31:09] Speaker 1: with like, what was the lesson from SpaceX's IPO, the
[31:12] Speaker 1: scale of it? And what it foreshadows, foreshadows is a
[31:16] Speaker 1: harsh term, what it sets us up for in markets.
[31:20] Speaker 2: The lesson is that the capacity is absolutely there in
[31:24] Speaker 2: the equity market. The lesson is that they are able
[31:28] Speaker 2: to tap other markets as well.
[31:31] Speaker 5: Right.
[31:32] Speaker 2: The lesson is that retail investors very much want to participate.
[31:34] Speaker 1: So you, therefore, in response, prioritize AI deals?
[31:37] Speaker 6: Yes.
[31:38] Speaker 1: Because you see the demand there.
[31:39] Speaker 9: Oh, of course.
[31:40] Speaker 2: The demand is there both institutionally and from retail.
[31:42] Speaker 1: Yeah, so for your job, I mean.
[31:44] Speaker 2: Oh, for sure.
[31:44] Speaker 1: Can I just ask you what the first half of
[31:47] Speaker 1: next year looks like to you if you had to
[31:51] Speaker 1: project something today and maybe how your thinking's changed?
[31:55] Speaker 2: Well, listen, it's been a wonderful market, but not for everyone, right?
[31:59] Speaker 2: So it is a market that is, we've exceeded 2021
[32:03] Speaker 2: levels of productivity in the IPO market on a proceeds basis.
[32:07] Speaker 2: with many fewer deals, much larger deals, right? Software has
[32:12] Speaker 2: been absent because of the potential disruption of AI. And
[32:16] Speaker 2: so if we fast forward to next year, that will
[32:19] Speaker 2: be interesting to see whether there's a viable market for
[32:22] Speaker 2: more traditional technology.
[32:25] Speaker 1: Are you hiring people for those specialisms?
[32:27] Speaker 7: Sure.
[32:28] Speaker 1: Yeah, to try and accommodate that changing opportunity.
[32:32] Speaker 2: But I would say hiring on a go-forward basis is
[32:34] Speaker 2: probably less specialized, right? You have to be more nimble
[32:38] Speaker 2: across business model. And so to just say software, AI
[32:42] Speaker 2: is changing the definition of a number of these companies
[32:45] Speaker 2: and subsectors.
[32:46] Speaker 1: JD Moriarty of Bank of America. It's really great to
[32:49] Speaker 1: have you here. Great to meet you. Thank you very
[32:51] Speaker 1: much indeed. Coming up, Mecca, a startup working to build
[32:54] Speaker 1: the foundation layer of deploying physical AI, just raised $ 60 million.
[32:58] Speaker 1: We're going to speak with CEO Josh Gao. That's next.
[33:01] Speaker 1: This is Bloomberg Tech. Mecha AI, a startup that designs
[33:13] Speaker 1: systems to help robots understand and operate in the physical world.
[33:17] Speaker 1: There's already $ 60 million in new funding. The Series B round,
[33:20] Speaker 1: led by Sequoia Capital, we're backing from NVIDIA, Microsoft's M12,
[33:24] Speaker 1: and Qualcomm, among others. Mecha co-founder and CEO Josh Gow
[33:28] Speaker 1: joins us here in San Francisco. I want to start
[33:30] Speaker 1: with what Mecha AI is and does. You know, we
[33:33] Speaker 1: do a lot on this show about world models. that
[33:36] Speaker 1: underpin physical AI deployments, but MechEd's doing something slightly different.
[33:41] Speaker 1: Thanks so much for having me, Ed.
[33:43] Speaker 9: Yeah, I think a big part of this is that
[33:45] Speaker 9: robotics is entering this deep learning era. A big challenge
[33:49] Speaker 9: for a lot of traditional robotics has been they weren't
[33:51] Speaker 9: able to generalize.
[33:53] Speaker 1: The world is incredibly.
[33:54] Speaker 9: Chaotic in the amount of noise, light changes, even the
[33:58] Speaker 9: dexterity of a lot of motions that we have to
[34:00] Speaker 9: go through. You can imagine just understanding physics is a
[34:04] Speaker 9: very challenging problem.
[34:05] Speaker 8: Yes.
[34:05] Speaker 9: And so for us, when we think about models, ultimately
[34:08] Speaker 9: it comes down to three things in this new deep
[34:10] Speaker 9: learning era for us. And that is, one, we have
[34:13] Speaker 9: to have compute algorithms and data.
[34:16] Speaker 1: And so some of the work that.
[34:17] Speaker 9: We do is how can we digitize a lot of
[34:20] Speaker 9: the physical world? There is incredible detail in the world
[34:24] Speaker 9: about how much pressure.
[34:26] Speaker 1: It takes to interact with objects.
[34:29] Speaker 9: What happens if I drop a cup and it shatters
[34:31] Speaker 9: on the floor? How do I know where the pieces
[34:33] Speaker 9: are going to be? This imagination is really the data
[34:35] Speaker 9: that we're trying to digitize.
[34:37] Speaker 1: And also how that data originates, right? So I think
[34:41] Speaker 1: a case study I always think about is me watching
[34:44] Speaker 1: engineers control a robotic arm repetitively to generate data of
[34:49] Speaker 1: moving an object from point A to point B. But
[34:52] Speaker 1: there is now a huge role for synthetic data. Getting
[34:55] Speaker 1: lots of real-world data is hard. How are you taking
[34:59] Speaker 1: that on?
[35:00] Speaker 9: I think a great example is if you look at AlphaGo,
[35:02] Speaker 9: for example. AlphaGo had behavior cloning with millions of different
[35:07] Speaker 9: Go games played over time. And then ultimately, through the
[35:10] Speaker 9: rules of the game, we were able to have self-play.
[35:14] Speaker 9: And the big difference here is that just the rules
[35:17] Speaker 9: of the game of Go is the entire world. In
[35:20] Speaker 9: our situation, the entire physical world can't be digitized right
[35:24] Speaker 9: into rules right away. And so the gap for us
[35:28] Speaker 9: to be able to create what is a perfect world
[35:30] Speaker 9: model or a perfect simulator for us to have synthetic
[35:33] Speaker 9: data is a long ways away. And hopefully we can
[35:35] Speaker 9: fill that gap across many different modalities and domains.
[35:39] Speaker 1: How does Mecca's business run? What is it that you
[35:41] Speaker 1: sell or what is your product? There's ultimately three stacks
[35:45] Speaker 1: in robotics. Can we handle the hardware, the software, and
[35:49] Speaker 1: the operations really well? So the first piece of the
[35:51] Speaker 1: business is ultimately, does the hardware exist to sensorize.
[35:55] Speaker 9: This information in the world? And so a great example
[35:58] Speaker 9: actually is force data. Typically when I'm grabbing a cup,
[36:02] Speaker 9: when I'm opening a door handle, when I'm twisting a bottle,
[36:05] Speaker 9: how many newtons does it take to apply the pressure
[36:09] Speaker 9: to And so our work is first, can we create
[36:12] Speaker 9: those sensors if they don't exist? We manufacture, we design,
[36:15] Speaker 9: we produce them at scale. And the second is ultimately operations.
[36:19] Speaker 9: We distribute and run logistics internationally so that we can
[36:23] Speaker 9: go into different scenarios, all the edge cases, different types
[36:26] Speaker 9: of businesses. An excellent example in this case is we've
[36:30] Speaker 9: also had Tony Hsu backing us. The founder of Doge
[36:34] Speaker 9: is probably the best operator or top operator in the
[36:38] Speaker 9: world at this point. And so we've had to go
[36:40] Speaker 9: through this muscle building of what does it take to
[36:42] Speaker 9: actually get the physical.
[36:45] Speaker 1: Intelligence into the world ultimately. Before we started the segment
[36:50] Speaker 1: with you, you said to me during the commercial break,
[36:52] Speaker 1: we are just getting started. But you've raised significant amount
[36:56] Speaker 1: of capital. You know, what are the priorities for you
[36:59] Speaker 1: where you need to hire people and what it is
[37:02] Speaker 1: you're going to do to scale?
[37:04] Speaker 9: I think a big piece of even thinking about the
[37:07] Speaker 9: scale that we need to be at, you can look
[37:09] Speaker 9: at Tesla in terms of spending a decade to solve
[37:12] Speaker 9: autonomous driving. They solve for two degrees of freedom of
[37:15] Speaker 9: acceleration and then direction. In robotics, we have many dozens
[37:19] Speaker 9: of degrees of freedom, even in our hands, every joint movement,
[37:23] Speaker 9: every body movement. And that scale of data is quite exponential.
[37:27] Speaker 9: So a lot of this funding and what we're trying
[37:28] Speaker 9: to do is we're trying to lay the infrastructure to
[37:31] Speaker 9: make this possible.
[37:32] Speaker 1: In many regions, even to support this data.
[37:34] Speaker 9: We have to set up hundreds of gigabytes of upload
[37:37] Speaker 9: bandwidth that doesn't exist before. And so set up the infrastructure,
[37:40] Speaker 9: scale in terms of hiring, attracting top-tier robotics talent, infrastructure, operations, hardware,
[37:47] Speaker 9: and ultimately be able to scale up.
[37:49] Speaker 1: Josh Gow, CEO and co-founder of Mecco. It's great to
[37:51] Speaker 1: have you on the show. Thank you very much indeed. Okay,
[37:54] Speaker 1: coming up, pictures, not prompts. Marissa Mayer is back with
[37:58] Speaker 1: a new AI assistant that turns your camera roll into
[38:01] Speaker 1: a tool for getting things done. Really looking forward to
[38:04] Speaker 1: this one. That's next. This is Bloomberg Tech. Most AI
[38:10] Speaker 1: assistants start with a prompt. Marissa Meyer is betting yours
[38:14] Speaker 1: should start with your camera roll. The former Yahoo CEO
[38:17] Speaker 1: and longtime Google exec is launching Dazzle, a personal AI
[38:20] Speaker 1: system that uses your photos and screenshots to learn what
[38:24] Speaker 1: matters to you. Marissa Merge joins us now here in SF.
[38:27] Speaker 1: This is so interesting. Again, the basics, you know, we're
[38:31] Speaker 1: used to saying, do this for me or through the voice,
[38:34] Speaker 1: do this for me or and the agent acts. This
[38:37] Speaker 1: is live real time monitoring of camera roll. Yes.
[38:40] Speaker 11: So what Dazzle does is a personal AI assistant that
[38:44] Speaker 11: learns from your camera roll. And it does two things. One,
[38:46] Speaker 11: it looks over your history of your photos and it
[38:49] Speaker 11: tries to understand you. Who's important in your life?
[38:52] Speaker 1: What are your hobbies? Where do you like to go?
[38:54] Speaker 11: How do you like to spend your time? And then
[38:56] Speaker 11: we also have an AI agent that monitors your camera roll.
[38:59] Speaker 11: So when you take a new photo or a new screenshot,
[39:03] Speaker 11: it monitors and says, wait, why did Ed take this
[39:05] Speaker 11: picture today?
[39:06] Speaker 1: What can I do to help him? Give me a
[39:09] Speaker 1: case study. I have one for you, but you give
[39:11] Speaker 1: me one first. A case study that helps the audience
[39:13] Speaker 1: understand why this is a good idea, like why it works.
[39:16] Speaker 7: Sure.
[39:16] Speaker 11: Well, a couple of different case studies. So one is,
[39:19] Speaker 11: the idea is like if I see something I like,
[39:21] Speaker 11: I'm shopping online, I'm busy, and so I can take
[39:24] Speaker 11: a screenshot, or I could, if I saw something, I
[39:26] Speaker 11: could take a picture of it. And Dazzle actually comes
[39:28] Speaker 11: back to me and says, did you want to buy this?
[39:31] Speaker 11: And it can overall make it a seamless purchase.
[39:35] Speaker 1: We're showing that kind of case study on the screen
[39:37] Speaker 1: right now. Exactly.
[39:39] Speaker 11: And the other thing that it can do is it
[39:42] Speaker 11: really understands at a very deep level Things that you're into,
[39:46] Speaker 11: things that your kids are into. My kids are off
[39:48] Speaker 11: of school on Friday. I asked Dazzle what should they
[39:51] Speaker 11: do on Friday. And I said, you know, Marielle's been
[39:53] Speaker 11: really into magic lately. Why doesn't she and her friend
[39:59] Speaker 11: go to a magic shop? And they picked a specific
[40:03] Speaker 11: one here in San Francisco, Misperceptions Magic Shop, where they
[40:07] Speaker 11: can go buy tricks, learn tricks. and spend the afternoon.
[40:10] Speaker 11: And so the fact that I could actually understand that
[40:12] Speaker 11: my daughter is going through a phase where she's really
[40:15] Speaker 11: entranced with magic is pretty fun and pretty interesting. Same thing,
[40:20] Speaker 11: I asked to tell my mother to buy skis. And
[40:23] Speaker 11: I said, well, because she's 76 years old, and because
[40:25] Speaker 11: this is where she skis, and roughly her stature, and
[40:27] Speaker 11: she loves the Rossignol brand, why.
[40:30] Speaker 1: Doesn't she buy this pair of skis? Based on photos.
[40:32] Speaker 11: Based on photos, because it's got pictures of her skiing.
[40:34] Speaker 1: Of those experiences. Let's talk about technology. How did you
[40:39] Speaker 1: build this? How did you get to what we're showing
[40:42] Speaker 1: is Dazzle AI today? And tell me a little bit
[40:46] Speaker 1: about the underlying technologies that support the agent. Sure.
[40:49] Speaker 11: Well, we obviously are using some frontier models to handle users'
[40:53] Speaker 11: requests and chats. But what we're doing is we're also building...
[40:56] Speaker 11: a deep context layer. So we try to understand who
[40:59] Speaker 11: you are and add that context as relevant to the
[41:03] Speaker 11: queries that we see. So when you're chatting with Dazzle,
[41:07] Speaker 11: we're using that context to augment and basically make it
[41:10] Speaker 11: a more personal form of AI. But in addition, we're
[41:14] Speaker 11: also using an agent to screen your photos, new photos
[41:18] Speaker 11: that you're taking, to understand how can we help, how
[41:21] Speaker 11: can we turn these into tasks that aren't necessarily a
[41:24] Speaker 11: to-do list for you, they're a to-do list for our agent,
[41:26] Speaker 11: because then our agents can kick in. and say, how
[41:28] Speaker 11: can we buy it? How can we book this repair?
[41:30] Speaker 11: How can we add this to your calendar?
[41:31] Speaker 1: So that goes to the definition of an agent. Is
[41:33] Speaker 1: that agent genuinely autonomous? It's able to make a transaction
[41:37] Speaker 1: or an action that's real. In Dazzle AI's case, it is.
[41:40] Speaker 1: That's right.
[41:41] Speaker 11: So our agents do go ahead and they are able
[41:43] Speaker 11: to do most of those tasks. We also think it's
[41:46] Speaker 11: important because some tasks are either more efficient or are
[41:50] Speaker 11: high value enough that it makes sense to have humans
[41:53] Speaker 11: in the loop. And that's another thing that we do
[41:54] Speaker 11: at Dazzle. So most of the work is done by agents.
[41:58] Speaker 11: But for the more complicated and intricate tasks, we actually
[42:01] Speaker 11: have humans in the loop that make sure that we've
[42:04] Speaker 11: got high accuracy and high fidelity. Last week, someone actually
[42:08] Speaker 11: went about replacing their roof using Dazzle. Took pictures of it,
[42:12] Speaker 11: got three quote people to come out and quote it,
[42:15] Speaker 11: picked it, and booked it through Dazzle. And for that,
[42:18] Speaker 11: I would argue that I think at this particular moment
[42:21] Speaker 11: in AI, maybe not forever, don't know what happens next year,
[42:24] Speaker 11: but for this moment in AI, it's actually reassuring that
[42:26] Speaker 11: to have a human in the loop guiding that process,
[42:29] Speaker 11: even if the agents are doing the majority of the work.
[42:32] Speaker 1: So you're taking us to this idea that in Dazzle's case,
[42:34] Speaker 1: this is for the individual, the consumer. Last week when
[42:36] Speaker 1: I was with Sam Altman at OpenAI Dev Day, you know,
[42:39] Speaker 1: dots is targeted at work for the enterprise. You're very
[42:43] Speaker 1: experienced in the world of technology. Did you arrive pretty
[42:47] Speaker 1: quickly at the idea this is consumer first?
[42:50] Speaker 2: Yes.
[42:50] Speaker 11: Well, I think that, you know, photos, I would make
[42:52] Speaker 11: the argument, If I had to be your executive assistant,
[42:55] Speaker 11: work assistant for a week, email is what you would
[42:59] Speaker 11: want though. I can figure out what to do from
[43:01] Speaker 11: your email. For your personal assistant, Right. Where are you
[43:04] Speaker 11: going to dinner tonight? What are you doing this weekend?
[43:07] Speaker 11: Buying you clothes, buying gifts for people that you care
[43:09] Speaker 11: about your the visuals of your life. Your camera roll
[43:13] Speaker 11: is a much richer source of context for me to
[43:17] Speaker 11: do that work in a really excellent way. That's why
[43:19] Speaker 11: we said, you know, look, we really want to affect
[43:21] Speaker 11: someone's personal life and be a good personal assistant. We
[43:25] Speaker 11: think photos, you know, maybe we add email over time
[43:27] Speaker 11: as well. But photos are a great.
[43:29] Speaker 1: Place to start.
[43:29] Speaker 6: Yeah.
[43:30] Speaker 1: Marissa Mayer-Abdallah, it's great to have you back with us
[43:32] Speaker 1: on the show. Thank you very much indeed. That does
[43:35] Speaker 1: it for this edition of Bloomberg Tech. There are so
[43:38] Speaker 1: many big stories out there, lots of them related to
[43:40] Speaker 1: the world of AI. Recap all of the conversations that
[43:43] Speaker 1: we've had and the reporting from Bloomberg News on the podcast.
[43:46] Speaker 1: You know where to find it, on the Bloomberg Terminal,
[43:48] Speaker 1: as well as online on Apple, Spotify, and iHeart, just
[43:52] Speaker 1: halfway through the working week here in the city of
[43:54] Speaker 1: San Francisco. Thanks for being with us. This is Bloomberg Tech.
[44:03] Speaker 3: Thank you.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.