Neuralink Rival Raises $200M, Benable Social Media App, AWS Chip Strategy, Snap Surge | Nov 6, 2025

6 Nov 2025 · 36 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

Podcast Episode Summary: Neuralink Rival Raises $200M, Benable Social Media App, AWS Chip Strategy, Snap Surge | Nov 6, 2025

Episode Overview In this episode of The Information's TITV, host Akash Pasricha explores significant developments in the tech industry, including Snap's earnings report, a $200 million funding round for Synchron (a Neuralink competitor), Metropolis's latest advancements, the launch of Benable's social media app, and Amazon Web Services' (AWS) chip strategy.

Key Guests

  • Sasha Kaletsky: Managing Partner at Creator Ventures
  • Tom Oxley: CEO of Synchron
  • Alex Israel: Co-founder and CEO of Metropolis
  • Tony Staehelin: CEO of Benable
  • Shaown Nandi: Director at AWS

Discussions and Highlights

  1. Snap's Earnings Report
  2. Performance: Snap's revenue grew by 10%, with a notable increase from its subscription service, Snapchat Plus, which has a revenue run rate of $750 million.
  3. Partnership: Snap announced a $400 million partnership with AI startup Perplexity, aimed at enhancing their AI capabilities.
  4. Challenges: Despite growth, Snap's advertising business is still slower compared to competitors like Meta and Pinterest, with North American ad growth stagnant at 1%.
  1. Synchron's Funding and Brain-Computer Interface Technology
  2. Funding: Synchron secured $200 million to enhance their approach to brain-computer interfaces (BCIs), focusing on a less invasive method via the jugular vein rather than through open skull surgery.
  3. Clinical Trials: The funding will support larger clinical trials and the development of a next-generation BCI system capable of interfacing with multiple brain regions.
  4. Goals: The aim is to restore movement control for individuals with paralysis and potentially aid in cognitive preservation.
  1. Metropolis and AI in Payment Systems
  2. Funding: Metropolis raised $1.6 billion, combining equity and debt, to enhance their AI-driven technology focused on automatic parking payments and seamless consumer experiences.
  3. Business Model: They charge convenience fees and SaaS fees, partnering with various businesses to improve operational efficiency.
  4. Expansion: Metropolis is expanding their technology to various sectors beyond parking, including retail and airports.
  1. Benable Social Media App
  2. Concept: Benable is a word-of-mouth platform that allows users to share recommendations for products and services.
  3. Market Potential: The app aims to fill a gap in the social media landscape, focusing on genuine user-generated recommendations.
  4. Monetization: While currently without revenue, Benable is exploring partnerships with brands for future monetization options.
  1. AWS's Chip Strategy
  2. Purpose-built Chips: AWS is focused on developing specific chips like Tranium for AI workloads, Graviton for general-purpose computing, and Nitro for virtualization tasks.
  3. Market Position: AWS aims to provide various chip options for customers, including NVIDIA and their proprietary chips, emphasizing price, performance, and reliability.
  4. Future Development: AWS is committed to evolving its chip strategy based on customer needs and market demands.

Key Takeaways

  • Snap Inc. is showing signs of recovery with a strong subscription service and AI partnerships, though still facing challenges in advertising.
  • Synchron is positioning itself as a serious contender in the BCI market with innovative funding and clinical support.
  • Metropolis is leveraging AI to revolutionize everyday transactions, showing a promising growth trajectory.
  • Benable aims to carve a niche in social media by focusing on trust-based recommendations.
  • AWS is solidifying its position in the chip market by developing specialized chips tailored for various computing needs.

Conclusion This episode highlights the dynamic and rapidly evolving landscape of the tech industry, focusing on innovations in AI, social media, and the ongoing development of brain-computer interfaces. Each guest brought valuable insights into their respective fields, underscoring the importance of adaptability and customer-centric strategies in achieving success.

For more information and to catch future episodes, tune into The Information's TITV on their official website or platforms like YouTube and X.

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.

0:13Welcome, everyone, to the information's TI TV. My name is Akash Pasricha. It is Thursday, November 6th. We have got a great show lined up for you today. We'll kick it off with Snap Earnings. Shares are surging after the company announced a partnership with Perplexity. We'll break it all down with a Snap Investor. Also, one of Neuralink's rival companies just raised a big funding round for its unique approach to brain-computer interfaces. We've got an exclusive interview with Synchron coming up. We'll also talked to the CEO of an automated payment startup, Metropolis, about their new funding. And we are still in TI50 mode, bringing on another CEO from our list of the 50 most promising startups of 2025.

0:56And last but not least, we are talking all things AI chips with Amazon Web Services, who has an interesting chips strategy themselves. There is a lot going on. And so let's get right on into things. Snap reported earnings last night. Revenue was up 10%, a slight acceleration from the last quarter, but the big news was the company's new partnership with Perplexity. Joining me now to break it all down is Sasha Kolecki, co-founder and managing partner at Creator Ventures. Sasha is also a Snap investor himself. I'm also bringing on our advertising reporter, Catherine Perloff. Catherine and Sasha, it's great to see you.

1:34Welcome to the show. Thanks very much. It's great to be here. So Sasha, let's talk about the results. We got 10 % revenue growth, which was a slight acceleration from the last quarter. What stood out to you beyond the revenue growth? Yeah, I mean, it's what made up the revenue growth. So what really happened is, as you said, revenue grew 10%. Within that, Snapchat Plus turned into a real monster. It's now at a$750 million revenue run rate. almost at Duolingo levels, just for the subscription product. As a reminder, Snapchat Plus is the subscription product which users can get additional Snapchat features.

2:15Advertising revenue is turned around. It's looking good. The only exception to that, which I'm sure we'll talk about, is US enterprise ads, but everything else, the rest of the world and everything, and SME is super growing fast. And Spotlight, which is their TikTok product, is also doing fantastically. In some types of the product, 600 % year-on-year growth. So it's really, really going well in many departments. And Sasha, we were talking offline before we started taping. So you bought shares after the last earnings call. Is that right? Yeah, well, we spoke last time about the last earnings call when it dropped, you know, 15 % or so.

2:47And I said that I was bullish at the price. So I bought. So I put my money where my mouth was and I bought. I mean, in terms of what gets me excited, I mean, Snap is historically been a very tough stock to own. There's a lot of reasons not to want to be bullish on Snap. Single digits has historically been single digits revenue growth when every other big tech company is growing significantly faster. It's also very anxiety-inducing. We talked about this last time. Every time the company reports, we're talking double-digit swings up or down. Every single time. Yeah, without exception. I think it's the last seven or eight in a row who's been there.

3:24But yeah, the ad product's also not level with competition. And finally, you know, what's most often criticized on X and elsewhere is that they give up almost 10 % of the market cap every year to stock-based compensation, which doesn't feel great as a shareholder. But despite all those issues, it just got to a point where the price became low enough that it became attractive. And then some of these headwinds became tailwinds on the revenue growth side and the efficiency side that they had a great quarter. And I hope that it will continue for the next couple of quarters as well. I want to get into the perplexity deal, but Catherine, I want to come to you.

3:58Let's talk about the state of the advertising business because it's still much slower than other advertising tech companies like Meta and like Pinterest. We were talking to you yesterday about that company. How do you size that up? Yeah, I think there's still some challenges with the Snap ad business. It only grew 5 % year over year. And in North America specifically, you know, I think Sasha kind of highlighted some of these issues, but it only grew 1%. And that's a problem because those are the advertisers that typically pay the most for ads. And they also kind of describe their advertising from large customers or big advertisers as a headwind.

4:43They've reduced the share of money coming from large advertisers, which I think is good. I think a lot of smaller platforms that aren't meta and Google over rely on large advertisers. Like, for example, Pinterest yesterday, or I guess now it's two days ago, they said that only 15 % of their customer base was small and medium advertisers. And I think it's good to have diversity there because large advertisers, you know, they control a lot of ad spend, but they're not the majority sometimes of the money you can get. And it's good to get a lot of mom and pops relying on you. But, you know, it's not great that Snap sees the biggest advertisers who could pay them the most money as a headwind.

5:26So I think that that could be a challenge. So, Sasha, let's talk about the perplexity deal. Every time any tech company announces any kind of a deal with a fast-growing AI startup, we can expect the stock to surge. That is certainly what is happening this morning. Why is this not just a flash in the pan? Why, in your opinion, does this have any kind of enduring value for the company? Well, the funny part of this deal is you say startup. I mean, Perplexity has a valuation of$20 million and Snaps at the time of the deal was$12 billion. So Perplexity is theoretically almost twice as big. But yeah, the headline number is huge,$400 million just for one year, at least how it was presented.

6:09A mix of cash and equity. And it's effectively for the Snapchat AI default search to be using Perplexity. They're not going to put ads in it. At least Snapchat are not going to sell ads in it. It's going to go straight into perplexity subscriptions. That's the reason for the deal. And it's for one year. The payment is going to be over one year. It's unclear if the deal will be renewed beyond that. But it seems like it's the start of a new Snapchat strategy to start putting a window for AI agents to build products on top of, which I think could end up being a relatively sizable revenue line. Catherine, what did you think of the perplexity deal?

6:48Yeah, I think it's interesting. It's sort of like, is this a new revenue stream for Snapchat? They mentioned that they were open to doing other deals with other types of AI companies. And I kind of, you know, thought about this a little bit as sort of like, you know, Google paying Safari for distribution. And, you know, it seems like they're saying they're open to other deals. So unlike that, arrangement's not exclusive. So if Snap could sort of leverage the fact that it has a lot of users and kind of make those users monetizable to other platforms, I think that could be attractive, especially because, you know, sometimes Snap has had trouble monetizing those users.

7:32So if they could find a better way for, you know, other types of companies to monetize them and sort of be a front door, that's really interesting. But yeah, it's also, I haven't really feel like I've heard other types of these deals. You know, we saw, for example, ChatGPT being the super app kind of announcement, but that was ChatGPT being the front door to other apps versus the app being the front door to an AI search engine. So I do think it gives Snap a bit of leverage, and it's pretty interesting. Well, look, I think any conversation around Snap in the past has revolved around their pursuit of the Spectacles business, and it is interesting to see how perplexity has now come into that conversation.

8:13I'm sure there's much more news to uncover here and to watch how the story unfolds. Thank you to the both of you for coming on. We really appreciate it. That is Sasha and Catherine here on TITV. Okay, the arena of companies that are building brain-computer interfaces similar to Neuralink is becoming more crowded by the year. Synchron is one company that takes the unique approach of crawling its chip to the brain through the jugular vein so you don't have to open up the skull. It's a company that I have covered in great detail. And in fact, I wrote a magazine feature on the business last year, which I will link in the show notes.

8:49Today, Synchron announced it has raised$200 million in a new funding round from big name investors, including Coastal Ventures. Joining me now for an exclusive interview is Tom Oxley, co-founder and CEO of Synchron. Tom, it's great to see you. Welcome to TITV. Hey, Cash. Good to see you. So big news today. $200 million in new funding. I remember it was just yesterday. We were watching the Brisbane Lions to their champion. It was the grand final, I think. And I should say they won again. They're going for like a three-peat now. Going for a three-peat. Yep, they're on fire. Unexpectedly and lots of kids.

9:27Amazing thing. Great. Okay. Well, we could talk more about football later on. Talk to me about what the$200 million is for. Yeah, so very exciting. Great support from our early investors. And this sets the scene for the next three years for us. We've done two clinical trials to this point with our permanent implant of our implantable BCI. The technology is intended for people who can't move their body to restore some level of control. we've been working alongside Apple to figure out gestures that come decoded directly from the brain to control Apple's iOS and we're now we've spent the last year or so getting to mature maturing about manufacturing and now we're moving towards later stage a much larger clinical trial and taking that first generation system to commercial launch and then about about 20 % of our financing is going to a next generation system that we're extremely excited about, which we think is going to be a real generational shift in the field.

10:32When you say next generation system, what is that system? So I'm not going to talk too much about it, but the core thesis for the company is that we believe the pathway to scale for this technology is through not delivering the technology in the operating room because that's a huge bottleneck for delivery at scale so it still is through a catheter it looks different the delivery looks different to our first generation system it's not a stent um it does use the blood vessels but it's got a slightly different approach of getting to the brain and it can get to all brain regions and we've got a micro electronics breakthrough that means we can deliver um a very large number of channels across multiple brain regions and it speaks to the you know there's been a lot of the use of the term whole brain interface recently so this would be a minimally invasive transcatheter whole brain interface fundamentally though is it still going in through the jugular and then reaching different parts of the brain i know it's not a stent but it's still it still would be delivered in the place where you would get a pacemaker or a stent or have a catheter so it would not be it's not open brain surgery but it It is a new way of delivering technology to the brain.

11:50So that's a next generation device. Let's talk about the stentrode though. When I had talked to you a year ago, you were marching through the process of undergoing clinical trials and moving ahead. Where is the stentrode at the original device? And we had talked about there was a new version of the stentrode that you were working on at one point as well. Yes, yes. Yes. So we've made some modifications to the stentrode. We've made it easier to use, easier to deliver, easier to manufacture, longer battery life, better wireless function, better integration, works with Apple. And we've been doing a lot of thinking on what are the use cases that people who have paralysis and motor impairment really need.

12:35So, you know, we're really fixated on the idea that we can help people who have lost the ability to independently use devices, restore that level of control. So you might have seen, we announced a few months ago, the first use of Apple's, what they're calling BCI-HID, which is a new Bluetooth protocol. So now when our system connects, implant goes in, system connects, you basically connect the Bluetooth and the system immediately recognizes that there is a brain signal coming in. And unlike with the keyboard or a mouse or even an eye tracker, there's now two-way communication between what the brain's doing and what the cursor on the screen is doing.

13:20And Apple's put a lot of energy into making it work well. And so we're excited to roll that up in the next clinical trial. What was the valuation for the funding round that you just raised? Not disclosing that, but you see in the report, it's around the billion dollars. Okay. And commercialization timeline, when I had talked to you, I think it was something around three to five years. What is it looking like in terms of how you're thinking about when you can actually bring this to market? Yeah, we're on the scale of several years. This financing, we want this financing to take us through the large scale pivotal study.

13:56There tends to be a bit of a gap between the completion of that study and getting the final commercial approval with the FDA. typically that can take up to about a year maybe even longer there's probably going to go to panel so um it takes a while like typically a class three implantable is a 10 to 15 year journey from you know original design all the way through to final commercial approval or what the fda called pma pre-marketing approval uh so that's that's roughly the timeline we're on for the first gen system before i let you go i do want to get your take on on a on a slightly different issue which is you're a doctor and we're talking about restoring function to people with ALS, people who have undergone these diseases and disorders.

14:39In parallel, since we talked last, there's been this discussion about longevity that has gotten so much attention. And I just want to get your take on it. As a doctor, you are also an entrepreneur. There's a lot of money now going into the longevity space, not necessarily related to the work that you're doing, but what do you make of all this investment going into that longevity sphere? Well, I mean, I'm not really up to speed with all of longevity, but there seems to be some breaking of the BCI space into longevity. I think what I've been thinking about recently is that there's some pro-transhumanist sort of, you know, people talking about how BCI takes humans up to some other level.

15:27I think the reality for the next several decades is that, I think for this first wave, it's about movement control for people who've lost movement. The concept of this whole brain interface, I think, gets to multiple cognitive domains. My view is that before we get to some transhumanist future with BCI, there's still going to be the potential for BCI to preserve cognition. And so there's a whole range of conditions where there's cognitive decline, there's mild cognitive impairment. I think BCIs are going to give you a stability over what your brain can do and preserve any decline and help you continue to perform at a level where the body might begin to fail over time.

16:08I think, well, then after that, Merge Labs is the new company. This concept of the merge is that there's this symbiosis of brains to computers. that's gonna it i'm not saying that's not gonna happen but we're talking on a very long time frame for that sort of a big cataclysmic cultural event i think for the next 30 or 30 years this technology is going to be about preserving uh human function and preserving really like self-expression human human expression right well congrats on the funding ron ron it's great to see you and uh tom thanks for coming on the show. We appreciate it. Thank you. Okay.

16:49Many AI companies are built to make everyday life a little easier, and that's exactly what Metropolis is doing. The company uses AI and computer vision to make parking and payments automatic. Think driving out of a parking lot without tickets or cash because the technology recognizes your license plate and charges you automatically. This company just announced$1.6 billion in additional funding that is a combination of equity and a whole lot of debt. Joining me now to discuss that is Alex Israel, founder and CEO of Metropolis. Alex, it's great to have you on the show. Welcome. Thanks, Akash. It's great to be here with you.

17:25Well, congrats on the big news. Lay out the company for us so that we understand just how complicated the business of parking and payments can be. Happy to. Happy to. So you're absolutely right. Today, we announced a landmark financing,$1.6 billion of new financing,$550 million of Series D led by Liontree and a$1.1 billion senior term loan led and lead arranged by JP Morgan. So very exciting day for our 23 ,000 employees and across our 4 ,200 locations in the United States. To answer your question, Metropolis was founded in 2017, and our business is entirely focused on artificial intelligence in the real world, and how do we leverage computer vision and artificial intelligence to drive seamless checkout-free commerce experiences everywhere you go.

18:18We started with parking, but we're moving past parking into quick-serve retail, gas stations, charging, tolling, and then moving past the mobility environment everywhere you go. And you're also doing stuff with airports too, right? Correct. So we spend a lot of our business, we operate north of 80 airports across the United States right now. And we're deploying this technology to facilitate, once again, seamless experiences. So for us, most of our lives are riddled with wasted time, whether it's ordering the same coffee over and over again, checking in in the airport, getting into a class A office building.

18:54How do you leverage computer vision and AI to make those experiences seamless? us. And so how do you make money? So we make money based on our partnerships, both on a B2B basis and a B2C basis. So whether it's a transaction, we often charge a convenience fee or to our partners, which include thousands of businesses across the United States, charging them a SaaS fee. Okay. And these deals are what? These are probably multi-year deals, I imagine with these large, I don't even know, the airport companies, the parking lot companies. Yeah, governments and real estate owners. So often they are seven plus year long contracts.

19:34Okay. And so talk to me, debt is obviously a huge component of the real estate market. Why was it that you took on all this debt? And what exactly is the debt being used for as opposed to the equity? So we have a remarkably profitable business at this point. And what we really want to be focused on as an organization is fueling growth. So it's both the physical deployment of our technology into the real world, as well as scaling into new verticals. At this point, we're probably across the largest applied artificial intelligence company in the world. We spend most of our time talking about generative AI and LLMs, really services that serve or facilitate experiences through the palm of our hands or on phones or on desktops.

20:16What we're focused on is how we actually affect the day-to-day lives of millions of Americans. Are you developing your own models? Are you working with these hot and high-flying AI startups? Who are you working with under the hood? We have a lot of partnerships, but right now our core technology is all proprietary. At this point, we're onboarding somewhere between one to two Americans onto our platform every two to three seconds, 24 hours a day, seven days a week. Almost, actually as of later today or early tomorrow, we'll have crossed 20 million members onto our platform. And how big is the business now in terms of valuation?

20:52How much are you doing in revenue? So we just crossed$5 billion in valuation and we don't really share externally our revenue figures, but they are north of$2 billion. $2 billion annually. Mm-hmm. Okay. And so I wonder as you've dived deeper into AI, how have your costs changed? We've talked a lot on this show about how expensive AI can get. You're obviously developing your own models, which means that you have a little bit of control over the cost. You don't have to look outside. But has it affected your margin structure at all? It's an interesting consideration. So we founded the company in 2017, really early days in the construct of how we talk about AI today.

21:38And when we think about lowering cost structures, what we're hyper-focused on is actually driving what I'd qualify as revenue synergies. So how can we leverage AI not to reduce simply the costs to our partners, but how can we drive more revenue to their locations? We're really focused on driving irrefutable value to our partners, which is really all about revenue synergy. How, when you deploy Metropolis's technology, do not just amenitize your locations, whether it's a car wash, a gas station, or a coffee shop, but more importantly, how does that drive incremental revenue to your business, changing the underlying value of your business.

22:12Great. Well, it's a fascinating business. And Alex, I want to thank you for coming on the show. It's great to meet you. Thank you, Akash. It's great to be here. Okay. This week, we've released our list of the information's 50 most promising startups that our sources cannot stop telling us about. Every day this week, we are previewing a company on that list. Today, we are talking about Benable. The company runs a social media platform that is built around everyday people recommending brands and products. Joining me now is Tony Stalen, founder and CEO at Benable, and Anne Guillen, the reporter who picked the company for our list.

22:48Tony and Anne, welcome to the show. It's great to have you. Great to be here, Akash. Thanks so much. Hey, Akash. So, Tony, let's talk about your company. What is this new social media platform that you're building? Yeah, we're Benable, the word-of-mouth platform. So it's an app where people find and share trusted recommendations, all types of recommendations. So yes, products like you were talking about, but also local services and restaurants and travel recs and resources, like everything. Our tagline is find and share your favorite everything. And now every two seconds, someone in the world adds a recommendation to Venable.

23:22How old is the company? We started about four years ago. Four years ago. The company doesn't have any revenue yet. When do you think you're going to start charging? Who will you start charging? Will it be a subscription, an ad-based platform? Talk about that. Yeah, great. We're careful kind of about what we talk about publicly. What I'll say is there's a lot of excitement from our brand partners and the things being recommended on Venable. You know, Ulta Beauty, great brand, great partner of ours. Their products have been recommended over 50 ,000 times on Venable already. Viore, Etsy, you kind of go, Hotels.com, all of our brand partners.

23:56So there's a lot of excitement from them on, wow, these people in the world love our products and services. How do we partner deeper and grow it that way? Okay. And so translate that for us in terms of business model. How do you see yourselves ultimately monetizing that in the future? Yeah. There's a few different ways businesses like us monetize. Right now, we're in pilots with some of these companies building an offering for them. So we'll have a lot more to come on that front in the next six months when we reconnect. So, Anne, you picked this company of all the social media platforms that are getting started up, not the least of which are the Sora-esque AI generative video social media platforms.

24:37Why did Benable really stand out to you among this category? Yeah, no, it's a great question. And I think now in the age of AI, e-commerce and consumer companies are in a really interesting spot. So I think, you know, something that I've heard from brands over and over again this year is they really want to understand how AI search tools are talking about them, what kinds of sources they're pulling from. And something just, you know, anecdotally that a lot of brands have found is that whenever some of these tools can find reviews or input from real people, they find that really valuable. And so brands, of course, are trying to figure out how can we get more people talking and sharing online about our products.

25:26And so I think that was one of the reasons that Benable stood out is that it's kind of an interesting new avenue for that. A lot of people talk a lot about Reddit, but I think just from talking with Tony and talking with investors that Benable has a lot of potential to help kind of fill that gap. You've been working on these lists for a couple years now over your tenure at The Information. I wonder how this year's category of e-commerce and consumer startups differed from years past. And as you were researching things, what sorts of observations you made about what VCs are interested in funding at all?

26:05Yeah, well, it's interesting. Consumer and e-commerce companies are definitely in an interesting spot. But this year, I think a lot of the companies that had a lot of energy or interest or momentum around them, a lot of them were related to AI and agentic commerce and shopping through ChatGPT and Perplexity and some of these other new tools. So along with Benable, we included a couple of other companies that are more working on the brand side, as well as with AI developers to help kind of build out that experience from payments processing to helping brands and merchants share their product information with AI search tools.

26:50So I think that definitely is an area that has a lot of interest and excitement right now. It's obviously still very early days, but I think an aim with this list is to try to pick some winners early on. And there definitely is a lot of opportunity in that area for startups to really break out. Tony, anytime anyone takes on the mountain of a task of building a social media platform, there are tons of risks that come with it. How are you thinking about the risks and what risks are top of mind for you? Yeah, before I ever started, you know, wrote the first line of code for Benable for the prototype, we tracked down all the past founders who tried something similar.

27:28It's not a new idea, right? This word of mouth platform. One of the biggest risks of why this company hasn't existed before is going too early into one single vertical, right? becoming just about books or just about restaurants or just shoppable things. And so it's really important. I think one of the biggest risks is not to get sucked into that, tie yourself to the mast, resist the call of the siren and build out this horizontal recommendation platform that can become an integral part of billions of people's lives around the world. So that's how I think about the biggest risk. Great. Well, Tony and Anne, I want to thank you for coming on the show.

Read the full transcript

28:01Tony, it's a fascinating business and look forward to speaking with you both soon. Thanks so much, Akash. Thanks, Anne. Okay. Our next segment is with our presenting partner, Amazon Web Services. AWS's Tranium AI chip has been getting a lot of attention lately. It was very much one of the things that executives focused on in the earnings call last week. Joining me now to discuss the company's AI chip strategy at large is Shao Nandi, a director at AWS. Shao, it's great to have you back on the show. Welcome. Ash, it's good to see you again. So let's talk all things chips, because that's the thing that AWS is very excited about right now.

28:37Can you just give us a little bit of an overview? I think Tranium is the chip people might have heard a lot of recently, but there's also Graviton. There's also Nitro, which is another chip that I hear about. Just set the stage for us and explain to us what is what. Yeah, a little context. I mean, we've been excited about this for a long time, even though it's all the news right now, thanks to AI. right? Maybe a little bit of framing on the concept of purpose-built chips or purpose-built infrastructure because people are like, why do you need that? Why is that a thing? And I'm a heavy traveler.

29:08Pranes, trains, automobiles, I do it all. And when I think about my methods of transit, like they all have different value props, right? Planes get you there fast, but they're not energy efficient. They're not flexible with the location. Trains are super efficient, but a little slow, right? And cars, you control everything. Before this sort of revolution of purpose-built infrastructure, chips were general purpose. They did it all. That means they're rarely optimized for one mode of computing or operating. And when we invented Graviton, an ARM-based capability, back in the 2018 era, by the way, on the fourth generation of Graviton, it was all about price, performance, and energy consumption.

29:44How could we take x86 chips and make them more efficient targeting general purpose computing workloads? In Graviton, by the way, 20 % better price performance. That means for the same performance, it's 20 % cheaper and 60 % more energy efficient. Good for the environment, good for your bottom line. And so Graviton's been in play for a long time. We also layered in Nitro, which was taking the challenge in clouds because we operated at such scale of how do you access virtualized infrastructure efficiently. So we took these unique tasks and built a specialized chip that could do a lot of those low-level tasks automatically, making them more secure, more private, and most importantly, more performant, more cost-effective.

30:28And then you've got Tranium. Yeah, so Tranium has been super exciting. We took this same mental model and we said, what's up next? Well, obviously it's AI and ML. Machine learning, AI workloads. We've seen all the success from the GPU providers in helping accelerate those workloads. We wanted to build chips from ground up that were initially optimized for AI training and then AI inference, which is the running the actual models in production. And we've announced up to the third generation of Tranium. We've launched Tranium 2 to production last year. It's just an incredible business. And we see such outcomes for customers with Tranium.

31:05And I'll tell you why it matters to them. Because customers care about price, performance, and reliability. And by building a purpose-built ship for Tranium, taking all the learnings we have from Graviton for training and inference, we've seen that benefit come right in. So let me jump in here because around Tranium, I know there's a third iteration of the Tranium chip coming as well, which Andy Jassy talked about on the call last week. But as it relates to this concept of price performance, I want to understand really what that means in the context of Tranium. Because the way I understand it, there are still scenarios where customers can come to AWS and actually run their workloads on NVIDIA's GPUs.

31:45use. And then there are different use cases where they will use Tranium. So how do you square that then with the price performance thing? It's all a little confusing. Break it down for us. Yeah, 100%. So look, first of all, one of our core tenants at AWS is all about choice. We want to enable both business users and technical users to get the blocks they need and get the service they need to go build. And for some companies, let's take a model provider. Like we just announced a big partnership with OpenAI, who's going to be leveraging NVIDIA's latest generation chips right here on AWS. For those model providers, they're often looking for the biggest, most powerful computing element possible to go train that model and get to market as quickly as possible.

32:28Not necessarily the cheapest. No, not necessarily the cheapest. And, you know, they also may have built targeting a specific chip architecture. or they may have been working with NVIDIA chips for years and years and have really optimized for it. And so for those customers, for example, we're always going to have the best-in-class NVIDIA products. We were one of the first to market with Blackwell. We're going to continue to see those launches happen. On the other side of it, we've had a huge announcement with Anthropic. We have a massive computing cluster for them with almost 500 ,000 Tranium 2 chips in it.

32:59It's built to build the latest cloud models. They have optimized for that, and they're taking all the price performance benefits. But when you asked about use cases, I'll tell you what I really see. When we think about customers who are running models for business outcomes, so I'm not building a model, I'm not fine tuning a model, but I'm actually using it for a business case. I'm perhaps helping improve outcomes in call centers through automated, you know, translation or sentiment analysis. They tend to care a lot about price performance because ROI matters. And so they want their workload running on the most efficient infrastructure possible.

33:31And most customers don't want to think about it. And so when you think about services, we have like Bedrock, I can tell you the majority of usage in Bedrock, token usage, is actually already running on Tranium because customers are a little isolated from the model via Bedrock, right? They use Bedrock as a way to get to these models. And our mission is to run it as cheaply, but performatively as possible. And that's where Tranium comes into play, running these workload, these inference workloads. So as it relates to AWS's chip strategies, then what is the three to four year vision for not just Tranium, but the chip portfolio as a whole and how that's balanced with all the other chips that customers can choose from while also running AWS.

34:12Yeah. First, I'm going to repeat what I said before. Choice is powerful for us. You will see us offering not just Intel Silicon, AMD Silicon, NVIDIA Silicon, and our own Silicon, but I suspect others in the future as well. So we always want to empower them. And you'll see us working on the supporting chips like Nitro that really make those work the best in AWS. us. For our own purpose-built infrastructure, you can look to the learnings we had from the database world or other areas where we've released additional purpose-built databases when there are niches and gap areas that have to be filled.

34:45And I think for areas like this, you'll probably not just see new versions of Tranium, but you'll see adjacent space chips over time as those needs open up. Inference and training, you'll see continued evolution. But we tend to focus in on where there are gaps. And the one thing customers tell us very consistently is they are looking for price, performance, and reliability to be at the top of the food chain for demand. Right. Great. Well, Chown, it's great to see you. Thanks for coming on, and we'll see you again very soon. Absolutely. Good seeing you. Well, that does it for today's show. A reminder, we are on this stream Monday through Friday at 10 a.m.

35:20Pacific, 1 p.m. Eastern. I want to thank Amazon Web Services, who is our presenting sponsor for this production. And I want to thank you for tuning in. We really do appreciate your viewership. I am already excited for our next show tomorrow. Have a great rest of your Thursday. Bye-bye for now.

From the publisher

Managing Partner at Creator Ventures Sasha Kaletsky talks with TITV Host Akash Pasricha about Snap earnings and the company's new $400 million partnership with Perplexity. We also talk with Synchron CEO Tom Oxley about the Neuralink-competitor’s $200 million funding round and its next-generation whole brain interface, and Metropolis Co-founder Alex Israel about Metropolis' $1.6 billion capitalization and their goal of being the largest applied AI company in the world. Next, we get into Benable's word-of-mouth social platform with its CEO Tony Staehelin and The Information’s Ann Gehan. Lastly we talk to Shaown Nandi about AWS's purpose-built AI chips strategy, including Trainium, Graviton, and Nitro, and how they balance with NVIDIA.


Articles discussed on this episode:

https://www.theinformation.com/articles/introducing-informations-50-promising-startups-2025


TITV airs on YouTube, X and LinkedIn at 10AM PT / 1PM ET. Or check us out wherever you get your podcasts.


Subscribe to: 

- The Information on YouTube: https://www.youtube.com/@theinformation4080/?sub_confirmation=1

- The Information: https://www.theinformation.com/subscribe_h


Sign up for the AI Agenda newsletter: https://www.theinformation.com/features/ai-agenda

More from The Information's TITV

All 304 episodes
Neuralink Rival Raises $200M, Benable Social Media App, AWS Chip Strategy, Snap SurgeThe Information's TITV · 36 min
Listen in VO