Amazon Debuts New AI Chips and Models at AWS

2 Dec 2025 · 44 min

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Bloomberg Tech Podcast Episode Summary

Episode Title

Amazon Debuts New AI Chips and Models at AWS

Hosts

  • Caroline Hyde
  • Ed Ludlow

Episode Highlights

The episode centers on significant announcements from Amazon's AWS Re:Invent conference, including new AI products, a notable philanthropic initiative by Michael and Susan Dell, and Netflix's proposal for Warner Brothers Discovery.

---

Key Discussions

  1. Amazon's AI Innovations at AWS
  2. New Chips and Models:
  3. Amazon introduced its latest generation AI chip, Tranium 3, designed for training and inference.
  4. The focus is on expanding the use of Tranium servers beyond the current primary client, Anthropic.
  5. Amazon is looking to demonstrate the cost-effectiveness and performance metrics of its new offerings, particularly against competitors like NVIDIA.
  6. Large Language Models:
  7. Amazon is also advancing its large language models, including Nova 2, which can process multimodal inputs (text, video, image).
  8. The company is striving to not only host other models but develop its own that provide tangible benefits for enterprise customers.
  1. Shift in Capital Expenditure Among Tech Giants
  2. Discussion about the shift in spending habits among major tech companies (Amazon, Microsoft, Alphabet) with projected capital expenditures exceeding $380 billion.
  3. Investors are responding variably to this shift; Microsoft saw a stock increase post-earnings report, whereas Meta faced scrutiny for not detailing ROI on its spending.
  1. Cryptocurrency Market Update
  2. After a significant drop in Bitcoin prices, a recovery was noted, with Bitcoin rising over 5%.
  3. Discussion on how retail investors have been adversely affected by previous market volatility.
  1. Philanthropy: The Dell Donation
  2. Michael and Susan Dell announced a $6.25 billion donation aimed at jump-starting investment accounts for 25 million children in America.
  3. The initiative is part of the broader Invest in America initiative, aiming to foster long-term financial stability for children.
  1. Warner Brothers Discovery Bids
  2. Netflix made a significant cash offer for Warner Brothers Discovery, surprising many as it was expected to be stock-based.
  3. Other bids were noted from Paramount and Comcast, each with strategic interests in different parts of Warner Brothers.

---

Key Takeaways

  • Amazon's Positioning in AI: AWS is intensifying its focus on AI, aiming to become a leader not just in cloud computing but also in AI solutions.
  • Market Dynamics in Tech Spending: The substantial shift in capital expenditure signals a new era for big tech, moving away from capital-light strategies.
  • Philanthropic Impact: The Dells' donation highlights a significant movement towards ensuring future financial security for children.
  • Media Landscape: Competition and strategic maneuvers in the media sector are heating up, particularly with high-stakes bids for Warner Brothers Discovery.

---

Upcoming Segments

  • Interviews with key figures from Amazon, including AWS CEO Matt Garman, highlighting further insights into AWS's strategic direction and innovations.

---

Conclusion The episode encapsulates pivotal movements in technology, philanthropy, and media, offering insights into current trends and future implications for the industry.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:01Bloomberg Audio Studios Podcast Radio News.

0:11Bloomberg Tech is live from coast to coast with Caroline Hyde in New York and Ed Ludlow in San Francisco. This is Bloomberg Tech. Coming up, we go to AWS Reinvent in Las Vegas to discuss the cloud company's new chips, new models and new AI agents. Plus, Michael Dell donates an unprecedented $6 billion aimed at jump-starting the investment accounts for 25 million children in America. And Warner Brothers' Discovery receives a new round of bids with Netflix flashing mostly cash for its offer. We'll have the details, but first let's check in on these markets, which are flashing green. We actually have a bit of a reprieve after yesterday's sell-off.

0:52We're back into risk-on mode, tentatively so, across the benchmarks. We're looking at the Nasdaq 100 up 9%. Of course, NVIDIA leading the chart in terms of points, but all of some of the MAG7 really dominating on the day. You're looking at Bitcoin even getting a bid. Yesterday, it was woeful. Today, we find some sort of stability. We're up more than 5%. In fact, largely crypto is in the green. We also turn our attention to some of the key MAG7 that we want to look at in terms of their own events, their own announcements, and then looking at what's happening with Amazon. We're currently trading up 1.3%.

1:22We're getting a little nudge higher on some new news coming out about its own large language models, about updates, of course, to its own AI agentic focus. And key has got to be Tranium 3, all about its chips. Let's get straight over to Ed Ludlow. You're in Las Vegas at the reInvent event. Yeah, it's probably the Tranium 3 headlines that move the needle right. This is a push forward, an acceleration of bringing the latest generation accelerator to market called Tranium, but useful in both the training and inference use case. and much as Amazon has done in prior generations of Tranium, it's talking up the cost and performance metrics of it.

2:02Forex, the prior generation on price per performance. It was interesting when the headlines hit. I mean, we're up 1.5 % on Amazon. A small move lower or pairing some of the gain on both Google and NVIDIA. And this is the story, right? Amazon wants to expand the use of Tranium servers beyond one core customer, which is Anthropic. and they're giving evidence in this release. We'll get to it later in the day when we speak to Matt Garman in particular of the savings that those customers are made using it while also making sure they have capacity that is based in NVIDIA GPUs. And of course, the open AI deal that AWS has, $38 billion of it is predicated on availability of NVIDIA.

2:43I mean, the eye, the focus must be so trained on what's just been seemingly a positive thrust for Google when it all comes down to its own chips, right now? Yeah, in recent weeks and months when there have been reports about Google's TPU and the idea that Google Cloud will have a big third party use it, it's moved the needle on those shares. What universally right now the hyperscalers are experiencing are supply constraints, right? Supply constraints where they're basically saying this is out there in the real world. On Trinium 3, it is out there in the real world, by the way. General availability from today.

3:23And they've also talked about the pipeline through to Tranium 4, but you're right. It's all about TPU. It's all about chips, but it's also about their own innovations when it comes to large language models. Lest we forget, yes, they've got an integration of OpenAI and Anthropic Claude, but they've also been moving the needle on Nova 2. Yeah, Nova 2, an updated year-on version of its first kind of in-house foundation model, a variant of which Omni can take inputs, multimodal inputs of text, video image, responding kind. But that's going to be the question of the story today. You know, AWS is number one in cloud and cloud computing through scale, through leveraging infrastructure.

4:03But when does it become number one in AI? You know, not just being a place that hosts others' models, but has evidence that its own foundation model is resulting in something tangibly useful, particularly for enterprise customers. And I'm pretty sure that that's what those that have joined the show today want to talk about. Yeah, we've got so many conversations come up in this show, but you've got to be sticking around because we thank you, Ed. We're coming back to you because throughout the show, we'll have interviews. But at the end, we'll have a sit down interview with AWS CEO Matt Garman at 3 p.m.

4:32Eastern time. Meanwhile, look, this is all a story of infrastructure about big tech companies such as Amazon, but also Alphabet, Meta and Microsoft spending heavily on AI with expected capital expenditure of over$380 billion combined in their current fiscal years. That's going against big tech's mantra for the past two decades, let's call it, which is more about delivering growth while keeping really tight lid on spending. It's a real flip-reverse. Let's talk about it. Bloomberg's tech equity reporter, Carmen Reineke. And sort of you've all been assessing how investors take this sudden wild capital expenditure.

5:06Yeah, I mean, it has been the talking point and the thing that we've been focusing on the most with some of these big companies. And as you said, it really represents a change from how they've operated over the last two decades. They really had very capital-light businesses and were able to be very profitable, grow exponentially in some ways because of that. And so now having this shift, we're seeing investors reward it on some ends and then really punish it on others. So, you know, Microsoft, we saw, you know, jump after its latest earnings report. But Meta sort of got hit because Zuckerberg, you know, didn't explain enough about the return on investment of the spending.

5:41The other thing that's interesting is that this capital expenditures as a portion of revenue have jumped to a level that's not normal for tech companies. So for Microsoft, its capital expenditures are now 20 % of its revenue. And in addition, you know, Alphabet and Amazon, these spending to sales ratios are some of the highest in the market. Okay. So we now have to perhaps digest a period of higher capital expenditure. We're meanwhile worrying about some of the financing that goes on within it, some of the circular financing that's been the narrative. But today the market jumps. Why? What are you hearing from the investors you're talking to?

6:16Yeah, I mean, I think that last week was very interesting. I think we're seeing a little bit of debt buying, some more optimism and enthusiasm coming back into the market. There's a, you know, NVIDIA's CFO was speaking this morning. There's just a lot of tech news sort of coming back into the market that people are getting excited about. Well, we'll see if it holds. Santa Raleigh upon us potentially, Carmen Reineke. We so appreciate your time. Let's talk about the crypto markets because they too are recovering today after a heavy sell-off yesterday when Bitcoin slid as much as 8%. Let's take a look at strategy as well.

6:49The artist formerly known as MicroStrategy, heavy investor in Bitcoin, gaining some support after scrambling to calm its own investors, creating a$1.4 billion reserve to fund dividend and interest payments rather than having to potentially sell Bitcoin as a last resort. Newberg Senior Digital Finance Editor Anna Herrera is joining us once more from London. Anna, why today the dip buying? It's hard to know with crypto, as I'm sure you're aware of. And it's crazy to think that yesterday we were down 8 % and thinking of all the macro condition that might have been leading to that drop. So, you know, we were speaking to traders today and investors, and it seemed that although the price is up, the sentiment is still a bit cautious and volumes aren't as up as they were before.

7:36I think people are sort of assessing that crypto in reality has been on a downward spiral a bit since October 12th, the 11th when the price crashed. And so while today we're a bit up, in reality, we're still pretty lower than we were around a month ago when Bitcoin reached its record highs. I mean, negative Bitcoin funding rates still in anxiety, extreme fear levels with coin market cap sphere and greed index, there's still a lot of comfort needed to an investor base that's been beaten up, particularly the retail trade, who've tried to gain exposure, not just by owning crypto, but by owning crypto related assets, such as what leveraged bets on micro strategy, now called strategy.

8:16And talk to us about how much retail have hurt here as well as institutional. Yes, obviously, we know that people were trying to retail investors were trying to get exposure into Bitcoin and crypto through buying ETFs that were exposed to strategy or strategy stock itself. And that's fallen dramatically over the past year. And so, you know, they've been hurting. But at the same time, we've been trying to wait to see if more institutional buyers will come in and offset that. But it seems to be, from what we're hearing, a sort of wait and see mode, partly waiting until the Fed decision next week.

8:45So it's kind of in a way, same old with crypto. One day it goes down 8 percent, the next day it's up 5 percent. The asset class seems to be maturing and there's more ways to invest in it, but maybe not much has changed. Bloomberg, Anna Herrera, always great to have you. Thank you very much. And stay with Bloomberg because there's a conversation coming up you won't want to miss if you're in crypto. Fong Li, strategy president, CEO, coming up in the next hour on Bloomberg Crypto. Elsewhere in that ecosystem, not all is gaining support. We're watching shares of American Bitcoin Corp. Look at that, down 44 % at one point, down 50%.

9:18The Bitcoin miner, which counts Eric Trump as a co-founder and chief strategy officer, seeing multiple trading halts amid the volatility. Now coming up, we'll be joined by Colleen Aubrey, Senior Vice President of Applied AI Solutions over at AWS. We are live from the company's re-invent conference in Las Vegas. This is Bloomberg Tech.

9:52Good morning from Las Vegas. Let's talk about some of the announcements out from AWS. Colleen Aubrey is Senior Vice President of AWS Applied AI Solutions. And Colleen, you basically work with the full spectrum of the smallest startups to the biggest enterprise customers that AWS has and the public sector. And distilling it as far down as I can, you basically help them to take technology and innovate, make whatever they do better. Yeah, the mission I'm on is really to put AI into the hands of the business on a day-to-day basis and really make it work for their business in production every day, helping them to deliver better experiences to their customers.

10:29We have a new generation of AWS Silicon and server design. We have a new generation of Nova model. Yeah. And we have frontier agents. That's right. And I think you probably heard me at the top of the show. But I think the big question about AWS reInvent is, what is it in this year's offering that takes AWS beyond being the sort of number one cloud computing platform for capacity to giving something tangible, useful, that can improve a company's own technology? Yeah, there's probably two areas that I'd like to talk about specifically. For me personally, one of my products is Amazon Connect. and this started as a contact center application, but today, you know, I see it really as an agentic product that's actually looking at whole customer experience.

11:14And so for me, this is where we're putting our infrastructure, our silicon, our models, and now today, like, bringing that all together in a product that can work for businesses. So we launched 29 new features on Sunday, we announced that. And I would say there's four key agentic capabilities that come with that. The first is actually AI voice and allowing customers to interact in a very natural way using Novasonic and having agents actually resolve issues for them on their behalf and in the background. The second is actually putting AI to work as a teammate next to customer service representatives, helping them to actually get tasks done, complete the paperwork, the processes, provide recommendations and help them really have a better view of the customer so that the conversation they have with customers is much, much richer.

11:58The third area is actually combining what I call clickstream, which is the paths that Customers will take through websites and profiles to be able to present a much more specific recommendation on what a next step might be for a customer. And finally, one of the big issues for companies is sort of confidently putting AI to work in their business. And in this case, we've added observability where you can inspect how an AI is reasoning, how it's thinking, what tools it's using. So companies can really observe AI in the same way they would think about the people in their business working with customers.

12:33I want to go to point number two. You used the word teammates. Yes. But you've in the past talked about a hybrid workforce, people and agents. A lot of what's come out since Sunday night seems to reflect that. You want to increase the number of useful agents that also internally, we can talk a little bit about what you're encouraging Amazon teams to do. But the idea is that workforces of all types should get comfortable working alongside agentic AI. Yeah, I see a future where actually everyone is managing a team of AI agents. And I think our frontier agents with software developments, DevOps and security is sort of one of our first moves in that direction, where you actually have a teammate with a developer who's able to address complex problems, able to work over hours and days, be able to solve for whole objectives and to scale with that person.

13:25that the role changes. I think we all are managing AI teammates, a team of people that are all of AI people that are out able to, we can delegate to, we can inspect what they're doing, we can iterate with them, provide feedback. And for me, that's where I think we end up going. I think we're clearly moving that way in the customer service direction. And I think the developer experience is also clearly moving that direction. You're in a leadership position at AWS, but you have been with Amazon broadly entering your 21st year. You're part of the S team, the leadership team generally, internally within Amazon's many arms, is this like, particularly in the context of the technology you've released here at reInvent 2025, is this at a stage where Amazon teams are dogfooding this or it's just inherent the use of technology?

14:11You want your own teams to be, I think, AI native. Yeah, that's true. And I think what I observe within Amazon, certainly we're all very keen to get our hands on our new frontier agents. And that is something we've had in beta internally. and I think teams are very excited to use that in production. We use Connect within our own customer service teams, our seller support teams, our AWS support teams, for example, and so we continue to iterate and learn from that. And I would say broadly across Amazon, we've really tried to sort of embrace the chaos, to put AI in the hands of every person in the company and to see what they can do, how they can transform, how they work, what they learn.

14:50And so you have a lot of teams, sort of grassroots, sort of trying to solve for what is the new way of working using AI. And some of those experiments don't lead anywhere. And some lead in a place that's really meaningful. And then you have sort of this social contagion that happens that people learn from each other. Okay. AWS, number one in cloud computing in the sense of scale, infrastructure, deployment. Is it number one in AI? I think we have all the pieces in place. And we're well on our way. We're very, very focused. And what I really like about our strategy is that we're working all the way, all the way the full stack, you know, from clean energy, chips, data centers around the world, bedrock with real great selection, including our own models.

15:34And I think Rohit's making really fast progress on that. And I think for Swami and I really trying to put those, all of those pieces to work, we're out of the box that a business can get value. We've got a conversation with Swami coming up in a bit. Colleen Aubrey, SVP of Applied AI Solutions. for AWS. Thank you so much. Carrie. What a great conversation. Let's stick with Amazon as well because the company, which is vast, it plans to offer deliveries of hundreds of household items, including some fresh groceries or over-the-counter medicines within just 30 minutes. The test program has begun in Philadelphia.

16:07We understand it will begin there and it's home city of Seattle. Now coming up, the Dells make an unprecedented$6.25 billion donation to the Kids of America. More on that next. This is Bloomberg Tech.

16:34Michael and Susan Dell are donating$6.25 billion, gifting 25 million children in America$250 each in an effort to jumpstart their investment accounts for the future. The proceeds build on the Invest America Initiative, or Trump accounts, as they're known, which will cede$1 ,000 for every child born from 2025 to 2028. Let's talk about all of it with Bloomberg's Tom Maloney. And how unprecedented is this, this size of philanthropic gift? It's pretty unprecedented. I mean, there have been big philanthropic gifts before, but something going to cover 25 million children in the cover really is pretty unheard of.

17:11And a gift to go through the federal government. I mean, he's donating the money to the U.S. Department of Treasury. That's pretty unusual, too. So it's quite an unprecedented gift. Let's go back to what all of this is sort of sitting alongside, which is invest in America. We sort of saw that roundtable alongside other CEOs, a lot of them tech, who have been thinking about how to get money into the hands of children to invest in the longer term, Tom. That's right. Yeah, those were created as part of the one big, beautiful bill. As you mentioned,$1 ,000 for kids born between 2025 and 2028. And what Dell is doing is kind of covering the gap for children 10 under who aren't covered under that program and giving them$250 to be used for things like college or starting up their own business or a home deposit when they're 18 or older.

18:02now Michael Dell and his family and he's the 11th wealthiest person in the world but it's other tech moneyed people who've been putting this initiative to work and I think about really Brad Gerstner who has been the driving force initially behind what is this sort of invest America outreach he set it up back in 2023 of course we know him from Altimeter yeah that's right he's been working on this for a couple of years as you mentioned it predates Trump obviously you know, the name Trump accounts certainly got his attention and I think helped put it into the one big beautiful bill. But yeah, it's kind of an issue that Democrats and Republicans have been interested in something like this, a product like this that gets everyday Americans invested in the stock market from a very young age.

18:53It's kind of a bipartisan issue. Yeah. And Brad over at Altometer, he's been saying that this is a platform for every company in America to think about how they reward employees' own children. But it doesn't need to just be a company. It could be mums and dads, churches, synagogues, many to be putting it aside in a sort of tax efficient way, Tom. What's interesting is, of course, Dell stock has moved on all of this. And this comes after a truth social post. Yeah, that's right. Dell stock was up about 4 % last time I looked. Trump obviously singling out Michael and Susan Dell for their donation this morning.

19:27He was seemingly very happy about it. And, you know, when Trump tweets about something, it tends to move stuff in the market. So we're kind of seeing that. I don't know if it's going to last, but there you go. And of course, that contributes to the net worth of the Dell family on the back of that Bloomberg's Tom Maloney. It's been fascinating talking to you. Thank you very much indeed. Let's turn now to Warner Brothers Discovery. The company has received a new round of bids, with one from Netflix consisting of a mostly cash offer, according to sources. For more, Bloomberg's Michelle Davis, who covers M &A in media, joins us now.

19:59So can we break down how these are starting to look? Because they're meant to be sweetened. They've been sweetened, yes. So the bidding is heating up. Yesterday, Warner Brothers had asked for sweetened offers from all of the companies. So we know they received offers from Netflix, mostly cash. That was a big surprise because people were expecting it to be stock. You know, Netflix's stock has been doing pretty well. They have great currency, but they're also an investment-grade company, so they can't afford to raise a lot of money. We hear they're in the market talking to banks about tens of billions of dollars for a bridge loan.

20:29So there's Netflix, there's Comcast, which has also bid. It only wants the streaming and the studios of Warner Brothers, similar to Netflix. We're still trying to glean the details of that offer, but we have heard that it likely includes a little more stock than the cash offered by Netflix. And finally, Paramount, the one who kicked all of this off, they're the only bidders who have actually made a play for the whole company, we're hearing that their offer is also completely cash. They have backing from Larry Ellison, David Ellison, you know, some of the rich, wealthiest people on earth, and Middle Eastern money, as well as Apollo, are putting in some of the financing for their offer.

21:03And it feels as though that's been the narrative, that the Ellison bid might be one that is more approved of by the administration. But are there any blockers, are there any issues from a regulatory perspective for Paramount Skydance just getting any bigger, or any of these players getting any big at? Yeah, so the Warner Brothers board is really, I mean, they are in a position where they have lots to assess, but it's not apples to apples because they're evaluating, you know, a bid for the whole company in cash, a bid that might include stock and cash, but only for part. And then, you know, other structures that are unclear.

21:35For Paramount and Comcast, there's clearly lots of synergies that they could achieve by combining the companies. But there's a view that doing that, you know, to achieve those synergies, they would cut a lot of jobs, maybe thousands of jobs at the studios, by combining these studios. Netflix, their argument might be, we're not going to do something like that. And so it'll play better, you know, with labor unions and regulators. But Netflix also already has a lot of power. So it's a big wild card to see how the regulators are going to view that. Yeah, many are going to say, is YouTube being counted in the numbers, how we think about this from a regulatory perspective, from a state and federal perspective?

22:10Timeline, any sort of idea? We're hearing that they want to get this sorted before the end of the year. And that makes some sense because for the banks who are underwriting the loans, they don't want the loans on their books at year end because it's going to cost them a lot more money, which means the bidders won't be able to pay as much for these assets. So we've heard that there's a deadline of before the holidays. A decision could be made as early as this week. You know, the binding offers are in, so that means Warner Brothers could move quickly. But they're also giving themselves room to ask for more bids, you know, later on this month.

22:40Michelle Davis will be across the story. We appreciate it. Now coming up, we're going to We're going back to AWS re-invent. The conference is on in Las Vegas. We're going to sit down with Swami Sivasabramanian, vice president of Agentec AI for that company. This is Bloomberg Tech.

23:01Welcome back to Bloomberg Tech. We check in on these markets that are losing a bit of their earlier steam. We're still holding on to three-tenths percent higher, but remember we're up more than a percentage point in earlier trading as NVIDIA perhaps loses some of its edge in terms of points perspective. Apple leading the charge right now, Intel as well. I'm looking at Bitcoin, though, still holding on to its gains. Best day since May. We're now at the highest level, 90 ,299, since the end of November. So, a few days. But we're up 4.7 % as we see a little bit of shifting in sentiment. Let's look at Keyname that has shifted in sentiment.

23:32It was beaten up in November. Now we're still up 2%, let's call it. We're rolling over somewhat. But Palantir and other AI names have seen a bit of love today. It seems as though Palantir could be a trillion dollar company. That's what Dan Ives has been saying on the day. But really, he's looking at big tech and Mag7 to add 20, 25 percent in terms of share growth for 2026. But for now, we return to one of the key Mag7 in that pile. And it is Amazon. And its AWS event is upon us in Las Vegas. Ed, take it away. Yeah, thank you very much, Cara. Let's talk more about Amazon's agentic plans with Swami Siva Subramanian.

Read the full transcript

24:08He's vice president for Agentic AI at AWS. And you've basically brought out and released three, what you call frontier agents. That's right. Across an autonomous agent, security, DevOps. And considering where AWS sits in the market and everyone that attends reInvent, probably the question most people would have is, if I'm a large enterprise and I have multiple teams, multiple desks doing multiple things, how do I deploy three of these frontier agents? Yeah. Let me start out with saying, here at AWS, we are building the foundation for billions of agents, and that's an understatement. So what we launched with these frontier agents is essentially a new category of agents, because while so much of activity is happening in agents, there have historically been more assistance to humans.

25:01And while we deploy it even within Amazon across tens or thousands of developers to help with software development, we were actually wanting to push the envelope and being able to actually do a lot more where they are actually autonomous, where you don't need humans to constantly steer them. And they're massively scalable and they can run longer. To be autonomous, they need to have an end goal. That's right. How does that work? So it's a case of saying, by the way, you use the term AI assistant. What we're now really talking about is AI co-worker. That's right. And that is the key thing here. It's just like what these frontier agents are doing is completely transforming how software is getting done in these teams.

25:44So much activity happens primarily in like software development. But now these agenting teammates, they show up as another teammate for software development. They pull up actually jobs from like GitHub task and start coding and clearing backlog. And if you're in the middle of the night actually getting page to deal with an issue happening with your website, this agent first takes the first page and says, oh, let me look into what's happening and unpack and then find out the root cause so that you can actually quickly resolve it and go back to bed. Like I've been on call and I know I could use that help and be proactive so that those issues don't happen.

26:26And third is actually a security teammate. Right. So it shows up again. and most of security has always been about afterthought. One of the things we are changing the game here is actually meeting developers even before they write a single line of code when they're designing software. We catch them right there and say, here are the things you should be aware of. And when they write code, we say, oh, don't do that, do this, and actually do penetration testing all before shipping. So now you can view it as like you have the best-in-class developer ops and security engineers showing up as virtual teammates for every software.

27:03You say best in class. What evidence does AWS have that in real world deployments, running these three agents in parallel autonomously has a clear benefit? Either efficiency, productivity, what data can you share? I'll give a couple of examples. One is in the DevOps agent that we launched, We ran it across thousands of escalations that happened this year, alone on incidents and so forth. These agents actually correctly identified the root cause 86 % of time. Right. That is really impressive. 86%, 86%. 86%. And the second one is Commonwealth Bank of Australia. They actually have a huge cloud infrastructure running across 1 ,700 accounts.

27:49and they put this DevOps agent test to test and for a very complex networking stack to debug it. And what they said wanted to take them like hours to debug, this was able to do it in minutes. And these are just few simple examples. Same thing what's going on with security agents. SmugMug is a great photo company and I'm a big customer and they are completely automating the security operations altogether using AWS Security Agent. And these are all the beginning of how this is going to change the game in a big way, because what you want these agenting teammates to be is always be there and you delegate some of the boring drudgery associated with ops and security and backlog on development so that developers are empowered to be extremely creative and you will start seeing 5 to 10x improvement in productivity in a big way.

28:48To recap, it's three agents. Kiro is software development. You have the AWS security agent, the AWS DevOps agent. They are designed to run for hours autonomously. To some, that's disconcerting, right? And I go back to my earlier question on, you know, how much concern is there? In the last segment, Colleen talked to me about observability, understanding in real time what it's doing. But that sounds to me like it's human supervision. It's a great question. I mean, this is the nuance and the art of balance we had to do. One, you want these agents to be autonomous and massively scalable, but you also want them to not go off the rails.

29:28So first, we are built it with the right guardrails. And second, even as part of development workflow, these agents can go actually take a goal from a developer to say, hey, go work on upgrading this piece of code to the latest SDK, and then start upgrading. But then as a final step, it might send it for a code review to the human and say, take a look and make sure I'm okay. And once they say I'm okay, I think it's fine. Or they can even configure it to say, if it passes all these tests, you're good to ship it so that you have built-in guardrails, either through human or automated tests. And these are the kind of things now we are able to actually innovate with the customers as well.

30:09And even there, we have done some pure innovation like introduce automated, mathematically provable property-based testing. So that means now developers can specify their goal and we can mathematically prove with things like Kiro that whatever it is generating can be tested exactly that way and it's accurate. So that is like another game changer. Swalmi Siva Subramanian, Vice President for Agenda KI at AWS. Three new frontier agents, Caro, that we're talking about here in Las Vegas. Back to you. Loving it. Back here in New York, Ed, it's time for Talking Tech. First up, Apple's head of AI, John Gianandrea.

30:48Well, he's stepping down with plans to leave the company entirely in the spring. The move caps a pretty tumultuous tenure that included a fumbled entry into generative AI. Apple won't be directly replacing Gianandrea, instead opting to break up the AI team. Plus Samsung has unveiled its first tri-fold smartphone. It's called Galaxy Z Tri-Fold. And the device contains two hinges that allow it to transform into a larger tablet-like device. The phone is set to be released first in South Korea on December 12th with a price of about$2 ,450. And the US Commerce Department has agreed to invest as much as$150 million into X-Lite.

31:26It's a chip startup where former Intel CEO Pat Gelsinger serves as executive chairman. It's the latest move by the Trump administration to bring chip making capabilities back to the United States. Now coming up, guess what? We go back to AWS in Vegas. This time it's a conversation with Sanjay Bhakta, Kone Nas, Chief Product and Technology Officer. This is Bloomberg Tech.

32:04Luma AI, known for its flagship-created product, Dream Machine Video Generator, is taking a step forward in global expansion with opening its first international office, and it's in London. This follows its recent$900 million Series C investment, which was led by Humane. Amit Jain is with us, co-founder, CEO of Luma AI, and so pleased to join you. I can't wait to get into reasoning video models and to the future of world models. But first, Amit, why London? we have a huge pipeline actually by the way thanks for having me i'm very excited to be here um we have a huge pipeline of of researchers engineers from uh like you know europe and and also deep mind that want to join luma and um also london is is the gateway for the business in in europe as well as middle east so it seems to be the the right place to to have our second office outside of Palo Alto and Bay Area.

32:57So yeah, today is our launch of the London office. I have a feeling Demis Hassavis over at DeepMind and the AI labs at Google will be having his ears pricked by that, Amit. So why would people currently working or having been trained at DeepMind and the like want to jump ship to help build Ray3, your reasoning video model? Yeah, I think that's a great question. So Luma already has actually a deep bench of researchers from NVIDIA, from DeepMind, from schools like MIT, Stanford, and Berkeley. And the reason that these incredibly exceptional people actually joined Luma is because, one, while it is a very, very well-resourced AGI lab, we are only about 150 people.

33:41So we get to do and get to have resources per person that is unheard of in the rest of the industry, actually. And second, Luma is so ultra-aligned on this one goal, which is to build multimodal AGI. There's actually no second, there is no second project happening at Luma. This is what we do. So people, researchers, engineers, and product people who believe strongly in this mission, Luma is the best place in the world for them to be. Looking at some of the amazing video that you create, I can see how it goes into the creative sphere and to advertising and the like. But Amit, it feels as though this model, these world models, this move to perhaps more physical AI applications must open up different industries.

34:28Yeah. So, I mean, video and video is actually the path to AGI. And let me tell you just very briefly why that is the case. Language gives us reasoning and language gives us the human abstraction that is necessary. Video shows us the entire universe, right? Like, you know, how things behave, how water behaves, how the laws of physics work. When you combine them together, then video, audio, and language together is basically a chance to build a universal simulator. So one application of that, as you point out, is obviously for entertainment and to generate video, to automate or to make digital the act of creating video.

35:04But in addition to that, it is the gateway to be able to build actually general purpose robotics. That is the second area that we are very deeply focused on and now starting to build out a team and the research area of that. And if we want to have any shot at building general purpose robots, then the only way that happens is by giving them this level of general understanding of the universe so that they can reason in their head. They can actually simulate every scenario in their head. Like, okay, what would happen if I do this? What would happen if I do something else? This level of learning and reasoning is essential.

35:40So video, as it advances and as we scale these models, they will not only get better and better, they will get more accurate in simulating physics. And this is the path to building physical intelligence. And that's why Luma's mission is to build multimodal AGI that can generate, understand and operate in the physical world. And that is the end goal. Amit, what is the biggest blocker to that or the thing that keeps you up? Because at the moment you've got the money, boy, do you have the money, 900 million raised. But you've also got promise of compute coming from Saudi Arabia. Is it about compute?

36:09Is it more about the talent that is what you need to push through to get to your mission? Yeah. So there are two really important things for Luma. Over the last couple of years, we worked tirelessly to solve the research problem for, like, you know, really, really great video. Now, the next frontier for us are these omni models that are able to, like, you know, reason in audio, video, language, text together. So the research is, like, you know, the first one that we pay a immense amount of attention to. Now, to solve the research, of course, you know, people and really, really brilliant people is one of those bottlenecks.

36:43So we hire from some of the best. See, we don't need thousands of people. We need like, you know, maybe 200 or 300 really, really brilliant people to work on these problems. So, yeah, that's the first one, the talent and great aligned people. That is the first one. Second is obviously compute. So alongside with the Series C announcement, we announced that we will build with Humane and collaboration with Humane a two gigawatt compute cluster. That is the largest compute build-out in our space of world models and video models and this kind of physical AI. And it is also one of the largest compute build-out in AI, period, right?

37:20So compute is one of the other things that is a big limitation. Ultimately, multimodal AI will be a superset of LLMs will be a superset of the AI we have today. It will require more compute than LLMs do right now. So compute is the second very important input to our business that we are working to not only shore up, but solve in a way we can serve these models economically to a lot of people. Amit Jain, come back and tell us how you're continuing on that mission. Go found a CEO of Luma AI. Thank you. Coming up, we're talking more about AI and infrastructure. We're going to AWS's event in Vegas.

37:57We've got a conversation with Sanjay Bhakta, Kondé Nast, Chief Product and Technology Officer. This is Bloomberg Tech.

38:18Sanjay Bhakta, Kondé Nast, Chief Product and Technology Officer, joins us from AWS reInvent in Las Vegas. So you're kind of bucking a trend that we've been talking about throughout the year in the context of AI. Actually, on-prem has kind of been back when AI is being run at the edge. You're going the other way from on-prem relying heavily on AWS. The rationale? Well, we are actually users of AI. We are not like some of these LLM companies that are creating and training their own models, although we have trained models in the past, but at a much smaller scale. So for us, investing in infrastructure and building out data centers for the size of our operation doesn't make sense.

38:57The work with AWS is focused on delivering personalized content, right? And is that at the foundation model level, you know, AWS Outward Nova to today, or it's just simply that you're leveraging their scale, the multitude of data platforms they have? Yeah, for that particular use case, we're actually just leveraging their infrastructure and scale. We built all the personalization models ourselves, and we trained them in-house. And we've been doing that for the last probably three, four years, long before LLMs became a thing. We've had our own data science team and we've been training models. So we use our own homegrown models for most of the personalization and recommendation work that...

39:34So herein lies the question of AWS reInvent, AWS number one in cloud computing. But they want to do more. They want to be number one in AI. Their own foundation models, the agentic tools that they release today. What would it take from Amazon in terms of the utility of that technology for you to rely on them more heavily? Well, actually, for some of the generative use cases that are LLM-based, we are using Amazon Bedrock. We have a contracts management rights clearance system that we just launched. We are using Amazon's Bedrock capability for some of our moderation, AI-based moderation of user-generated content.

40:14So we're looking more and more now towards using out-of-the-box capabilities that Amazon provides rather than build and train our own models, which we used to do in the past. I think the need for that is becoming less and less. Conde has a relationship with OpenAI. Conde was one of the first to make a deal with OpenAI. Early days, it seems, but how is that progressing? That's going well, actually. We're starting to use ChatGPT quite widely within the enterprise. Internally? Internally, yes. And for external use case, we've actually launched an AI-based recipe search on Bon Appetit on our website.

40:53And also it's going to come out in our app, which allows customers to go in and do natural language search and also be able to modify the recipes according to their taste. So that's the first use case. We are looking at others with OpenAI as well. The broad theme of the program today here at AWS has been about companies of all sizes moving from using AI assistants internally. to what AWS would call the AI co-worker, a hybrid workforce of people and energetic AI. Conde has done layoffs in the past two years, financial years. How much has that been about AI tools changing productivity, eliminating certain roles, changing certain roles?

41:33Well, I don't think most of it has been about AI, actually. Some of it, especially in tech, we use extensively, we use AI for our work on a day-to-day basis, which eliminates the need for some roles. We can do more things with fewer people. But other than that, across the organization, we've not really had any AI-based impacts. I've got to ask, in a different part of the Amazon universe, I've been experimenting at home with Alexa and Alexa Plus. Do we get some kind of Condé-Nas Alexa integration? Is there work going on there? We have actually already integrated with Amazon Alexa. How does that work?

42:11It's available. So, you know, you can query and you can get content, you know, read out to you from some of our publications, not all. So, but I think it's pretty cool. We should try it out. At the heart of that question is something a bit more existential. Conde sees itself more as maybe entertainment. And if you look at the revenue streams, very different from saying you're a news organization. Just explain how you see this company transitioning right now. Yeah, I think we're definitely in the entertainment space because if you look at our brands, you We mostly cover leisure, fashion, lifestyle.

42:42We're not a daily news outlet, so people don't come to us on an everyday basis. So we are competing with other entertainment outlets, whether it is streaming media like Netflix or Hulu or Amazon Prime or social media, TikTok and Instagram and others. So people have limited spare time, so we are competing for a slice of that. So we definitely have to do way better in terms of our personalization, our user experience, and also our great journalism that we have, I think, is what is going to take us forward. Sanjay, in the year ahead, what's the one big change you want to make on the technology side, the one thing you're still yet to do, be it AI or something else?

43:18Yes, I think it's mostly definitely around AI. We have a pretty solid data infrastructure that we've built on Amazon with Databricks. and our next task is really to figure out how do we vectorize all of our content and make it readily available in real time to LLMs. I think that is probably our number one challenge. Sanjay Bhakta of Condé Nast, thank you so much for joining us here in Las Vegas. Caro, back to you. Fascinating set of conversations across the gamut of all things AI and entertainment. Ed, we thank you. You've got so much more coming up. That does it for this edition of Bloomberg Tech, but do stick around.

43:52Ed's got a sit-down interview with AWS CEO Matt Garman. It's 3pm Eastern 12pm Pacific First Right here on Bloomberg Television You all don't forget To check out our podcast You can find it on the terminal As well as online On Apple, Spotify And iHeart From New York From Las Vegas This is Bloomberg Tech

From the publisher

Bloomberg's Caroline Hyde and Ed Ludlow break down Amazon's new AI products from its AWS Re:Invent conference. Plus Michael and Susan Dell make a $6.25 billion donation to the future of children in the US. And, Netflix makes a nearly all cash offer for Warner Brothers Discovery in its sweetened bid.

See omnystudio.com/listener for privacy information.

More from Bloomberg Tech

All 342 episodes
Amazon Debuts New AI Chips and Models at AWSBloomberg Tech · 44 min
Listen in VO