5:38
Amazon says it is abandoning non-disclosure agreements for data centers amid growing public discontent. In addition, startups are entering the race for personal AI agents: one of them announced the sale of computers for $ 3,500, created specifically for personal AI. And of course, we're keeping an eye on several deals of the week, including Uber's Welcome to TechCrunch's Equity podcast about business and startups. Today is Friday, October 9th. With you is Anthony Ha, along with TechCrunch Senior Special Correspondent Sean O'Kane and TechCrunch Senior Reporter Rebecca Bellan.
Rebecca, you were on vacation. So first of all- welcome back, and sorry. Tell me, did you see the news that Trump announced the creation of new " superintelligence" forces? Yes, of course, it happened while I was gone. But, judging by what I've learned, it now looks like just another pacifier. It looks like plans to create even more plans in the future, perhaps. I don't know. Do I understand correctly? Yes, it seems like an AI agent could solve this. Now we call that superintelligence, right? Did you see that during the recording of this episode, Trump wrote on Truth Social that if you call it
artificial intelligence and not superintelligence, then you are the enemy? That's it. Although it seems Elon said that the XAI division within SpaceX will actually become XSI or something. SpaceXSI, yes. Easy to pronounce, right? Yes, we talked about this a little bit last week. Not so much about the forces themselves, but about the terminology. And I think part of the bigger story here, aside from " superintelligence forces" that are unlikely to do much, is that the outrage and backlash against AI, coming from many directions, has become so loud that the Trump administration feels the need to do something. So it's a kind of strange rebranding: a pretend action, a nod towards
security. Look, a lot of it is just a " signage," these " superintelligence powers" -- they're just another shiny title for people to add to their LinkedIn or whatever. Like, okay, cool. This seems to be an integral part of the big AI rebranding we're seeing, because everyone is starting to panic about it. So they say, "No, no, don't be so scared. This isn't artificial intelligence, they say. This is superintelligence. And look at these cute personal AI agents that will help you get your money back by looking through your email, which we'll talk about later." But it's interesting. And also, apparently,
this week, at the same time as this-I don't know how much of this has to do with the power of superintelligence- there was a science summit, right? Where they announced $6 billion initiatives for the use of AI in science, as I understand it. That's, you know, 2.4 billion in tools and computing power from companies like Nvidia, AMD, OpenAI, Anthropic, and Google. What do you think about this I was just reading the announcement before starting the recording, and it seemed to me, first of all, that the language of the announcement was too arrogant. Like, "This is the largest investment in science in American history." Although, I think it's fair to say that the Trump administration has not been the best friend of science in American
history. But I'm sure we'll continue to discuss many of these larger AI issues. Actually, this will be a red thread throughout the entire issue, spoiler alert. But there are likely, so as a final reminder: Disrupt is happening next week as we write this. On Tuesday at 9:00 AM we will have a live recording of Equity, and if you use the code equity 25, you will receive a 25% discount on your ticket. So this is equity 25. Let's continue talking about AI and the backlash against it, specifically data centers, which I think has become one of the main areas of resistance in the AI community, and Amazon put out this blog post late last week where they were trying to make the case
that data centers are great, and buried deep in the text was an announcement that they would no longer use non-disclosure agreements when negotiating with local governments. Sean, what do you think about this? I found this interesting. First of all, it's good that you noticed this. It's interesting that this was kind of hidden in the blog post as part of a larger story that Amazon is obviously trying to tell, but I think it's very important. I understand why they didn't bring this to the forefront, but to me, it's the root of many of the problems the tech industry is currently facing with AI in general and data centers in particular.
You know, there's a lot to talk about whether people should be worried about the impact of data centers on their property values, electricity bills, water consumption in the area, whether these problems are real, exaggerated, etc. I think it all comes down to the fact that many of these projects are discussed virtually in secret. Sometimes they negotiate about them knowing that it is a data center, but not knowing who exactly will use it or occupy it in the future. I think this is a long-standing problem in the technology industry, right? I always remember the mid-2010s when Uber was expanding very quickly. Uber was one of the companies that actively used this tactic, especially in cooperation with local
authorities. I remember I was working at The Verge, and our freelancer wrote an article about a city near Orlando where Uber was trying to negotiate tax breaks to subsidize people's trips to public transportation after an Uber ride. And all this was discussed in secret. And not only was this unpleasant in itself, but because of the secrecy, it was clear how brazen Uber was, demanding a bunch of other things because they were sure no one would find out. The city was literally asked not to place the Uber logo on trash cans or even toilets, because it could be associated with something that is thrown away or flushed away. Just crazy stuff. So there needs to be more of this. I'm glad Amazon is doing this.
We'll see if this has any positive effect, but people are pretty determined right now I wonder what this will lead to and whether it will help communities successfully delay or stop the construction of infrastructure that these companies need, right? Look, I think from what I've heard from people who work with data centers every day, a lot of these moratoriums that are being passed now-and there are a lot of them, including the one I was in San Francisco recently, right? Which is quite ironic, but they only last a year or two. But these data centers still won't be operational in the next 2, 3, 4 or 5 years. All of these development projects will be significantly delayed for a number of reasons: from zoning to financing. Of course, if the resistance
continues, it may have an impact, but I often wonder how big the real impact of this political resistance is I mean, of course, it's worth noting that the moratorium that I'm most aware of is the one in New York, where you and I live, Rebecca, and I believe it's a one-year moratorium on permitting for major projects. So obviously there's room for maneuver here, and often it's presented as, "Hey, we need to think about this, that doesn't mean we think it's bad. Let's not do it." It sounds like, "We need time to study," which gives them the opportunity to say later, "Actually, we thought this was great. We like it." But you can see that the people who are trying to build these data centers are worried, which is why
we see these, you know, somewhat inappropriate tweets from people like Trump or David Sachs. Only time will tell how long and deep this resistance will be, but I really think they are worried. It's also worth mentioning that Microsoft made a similar statement earlier this year about abandoning non-disclosure agreements (NDAs). You know, there are all these other criticisms, right? I don't think that simply lifting the NDA will automatically make the people who oppose these data centers say, " Great. Let's do it. I'm okay with that." But I think the beginning of such a
conversation is like, "We don't even know what you're doing there, because you basically agreed to all of this in secret." So this seems like a good first step towards winning people's favor. I mean, the bigger problem here is just a communication problem, right? This is a problem, in my opinion, inherent in the entire technology industry, but now it is especially acute, because we live in an era of direct communication, when you can do without communicating with the media. They said the media is dead, etc. And I'm not trying to beat a dead horse, because we ourselves are the media and we believe that we still remain relevant. But I think that through the many controversies and backlash
against AI and data centers, we have seen that the industry is very bad at explaining its actions. And they are, of course, constantly irritated by the fact that most people can't imagine a world in which these things would be useful to them. This is partly because there are often simply no opportunities for this. And partly because they just don't know how to explain it well. For example, last week I wrote a piece about how it had been a year since we heard Tesla boast about their progress towards what was originally called "sustainable wealth" and has now become " amazing wealth " in their mission statement. It's just amazing to me that a year has passed and the company still doesn't have a clear definition of this concept, especially
considering they released an entire multi-page PDF document when they announced it. Even Tesla's biggest fans criticized Elon Musk and company for the vague wording, which read like meaningless AI-generated text. And they still don't see the need to explain this to people, which I think is just crazy, because we're starting to see real possibilities, like OpenAI, which this week, you could say, solved 500 math problems in one go in different ways. There are things that these models are capable of now, and you could try to explain that to people by drawing parallels. But I think the fact that there were no opportunities, that there was no communication, and that at first it was all literally shoved down our throats like Copilot.they're really going to have to make
some kind of excuse for that. Yeah, I mean, one of the things that this whole situation has made me think about is the narrative and the feeling that for a while, as you said, we've really been trying to force this idea on us that this is "inevitable," meaning an AI-driven future, and that you have to get on board or you'll be left behind. And I think that as that future continues to rapidly approach, obviously in many ways, there have been some obstacles at a few key points. I really like the realization that no, you really have to justify this. You can't just say "this is the future, shut up, " although I think that's what, again, the Trump administration and other people are trying to do-just shout "China, China, China, China, China," don't
argue with me. Ahem. Yes. And regarding Tesla and the idea of abundance, Sean, I really liked your story, although what struck me was not so much that they didn't define abundance, but that they made such a promise at all. This is wild, and to me, it shows how strange our relationship with Tesla is, especially with the people who invest in it and support Elon at all costs, because it's not about whether the company sells a lot of cars. It's really about the vision of restructuring society, at least that's what they're trying to sell them. Yes, that was always the vision. Yes, that sounds like "we're going to destroy money."
Yes, it's always been like this: when I talk to people who support Elon no matter what, and I tell them about the numbers, they say, "No-no- no, you don't understand, this is his vision of the future," and I say, "Sure, cool, cool, cool." The future is already here, right? I mean, there are all these companies now that promise to give us personal assistants, and they've been talking about how AI is going to help with everyday tasks for a long time, and now some of that is starting to come to fruition. Instead of the general resistance that people had to big language models because of low-quality content generated by AI, they're saying, "Oh, wait, it can find money in the mail for me, it can book me tickets." Oh, wait.
Okay, let me look at this. So, I don't know yet how good it all is. I'm testing several options. But I do n't know what you think about it Yeah, last week, or the week before that, Sean was telling us that Muse actually found some money for him. But he wasn't as grateful as I expected. He said, "That's cool, but, you know what you've done for me lately?" How much money? I didn't hear. Let's see, let's dig into this. He found me, found me a considerable amount, you know, in some unclaimed property in Texas, and it was something I should have thought of on my own, but he did it for me, and it was fantastic. But this is a one-time thing that I
can't repeat, you know? It's just, well, I think we're talking about agents now. This is the first real step that I think this industry has taken in trying to correct what I was talking about and explain to people how this technology can be useful, other than being a better Google or, you know, a psychosis generator. And I still think they're struggling with that, right? I don't know about you guys. I understand why so many people in Silicon Valley think that this agency stuff is for those who don't use it at the corporate level? But for most people, I think they'll have a similar experience: "Oh, I can immediately identify a couple of things that he's offering me." That's useful, but will there be enough business around a more passive use of something
like Muse, where it's not a commercial story but more of a search for information or something like that. I don't know. Well, I also think this could be a good transition to our second big topic, which is of course related to artificial intelligence, because it's 2026. But the point is, there are, you know, a number of different startups that are trying to take advantage of this, well, find opportunities that OpenAI and Anthropic haven't yet taken. And we can talk about any of these new companies that interest you, but maybe we should start with Ghost-a startup that was covered on TechCrunch, and, at least in my opinion, that story really blew up for us. It seems like it really touched a lot of people. This is a startup that is building a
$3,499 computer for personal AI. Yes, that's $3,500 for a computer. This is its foundation, and its main purpose is to perform AI tasks for you. So, there's an Nvidia graphics processor inside. AI models are already installed. It seems they have their own software, browser, file system, and it can run continuously without the cloud. So, yeah, I guess you know how it's different from the Mac Mini? I don't know. I was thinking: is this just an external hard drive with a little extra? Do you still need a phone and an app to make this thing work? And then, I imagine, one day they'll say, "Well, you can't have two devices." Let's combine everything into one, and then you just, what, recreated a phone and a computer? I, I just, it's
interesting. I mean, I think people like it because it's "hardware," and people like technology and specialized equipment. We have seen that creating technology is so difficult. It's hard, but people like the ideas that go into it. This is, you know, an interesting design. It looks, you know, like a mix of the PS5 and the monolith from the movie 2001: A Space Odyssey. But again, we are facing a similar problem when, as one of the developers said, this is an example from our article. It 's supposed to, you know, maybe proactively send an alert that you might be sick because it sees an elevated resting heart rate, sees from your computer history that you've been working late
on a project, and has the seasonal context that it's flu season in San Francisco. It's like, do we need a $3,500 box to tell us this? Well, I think for some people, yes. Some people are looking for something that will allow them to express what they are thinking subconsciously. I'm not saying there won't be other uses, but I think we're still very early in the phase of agent apps on our phones, and they haven't really caught on with people outside of Silicon Valley yet. And this is clearly aimed at them, I know that many use open models, for example, Qwen. And I know people who have tried to assemble their own systems in their apartments or houses that resemble something similar, but without the proper quality. So I'm sure they'll find a market for
it among people in the industry. Yes, in my opinion, this is definitely not a mass market product. Ordinary people are not going to spend $3,500 on an AI computer, which, judging by our review, we tried to find out at least something about the characteristics, but did not get many detailed answers. It was just something like, "Look, an AI computer." You know, I think there's an audience for this. I mean, most of them live in Silicon Valley, spend a lot of time at X, and are very passionate about it. You probably raised, I don't remember exactly, I think it was about 11 million dollars. Not a huge round, but at some point you have to understand that the market has to be bigger than just ardent AI evangelists from Silicon Valley. Well
, you know, that's not true. I think we're seeing one thing because there are a few other startups that we'll mention that are working in the personal AI assistant space. And one thing they all have in common, to me, is the ability to run something locally on your own device, and also a focus on privacy, right? Because these applications, these agents, won't work unless they have full access to everything, right? If you, I don't know if any of you have ever been an assistant. I was an assistant, an executive assistant. I had documents about, you know, what my boss liked, what her regular orders were, where she liked to sit on the plane. I knew all this, honestly, it was all in the documents that I gave to someone else. I was an agent, and I had
her SkyMiles and her credit card. I even had her children's social security numbers, right? In short, I had it all, right? And so I think that if you're going to entrust this to an agent, you need some privacy guarantees. And it's difficult because , on the one hand, do you really want to give all of this to Meta, right? Do you really want to do this to give back to Muse? Do you want to give all this data to Muse? On the other hand, do you want them to go to a company that has the infrastructure to protect your data, or to a startup that is a year or a month old and uses a shared cloud? So that's interesting. I mean, there are others, like Tab. They just came out of secrecy mode. The company's valuation is $300 million.
They haven't disclosed funding details, but they're focused on that as well. And one of the great quotes, written by Dom, one of our reporters-I think the final sentence is great: "Capabilities are basic requirements. " Trust is a product." Which, I thought, was pretty cool. Yes, and it seems like trust has become a big topic for a lot of these startups right now. There's Underdog, which also supposedly emphasizes privacy because it runs entirely locally, and I'm curious what that actually looks like, how powerful the computer needs to be, and how powerful the model can be if it runs entirely locally.
Maybe they will merge with Core. Yes, that's right. And there's Hark, which is also focused on privacy. So it really seems like a cross-cutting theme. Ahem. You know, I would say that for me personally, the idea of giving an agent access to my hard drive or my credit card-I'm probably a little behind most of our listeners, but I'm not ready for that yet. I won't do that. And part of the problem is that's where they see a lot of the value, right? Oh my god, once you do that,. Every now and then we'll delete an important file or buy something very expensive that you don't need, but that's okay. Don't worry about it. So, uh, I, yeah,
I'm not really sure that they're going to help me at least bridge that gap. Yes, I mean, Rebecca brings up the most important point here, because from our position on implementing these things, everyone is at a dead end, right? more out of these agents, the way companies want you to think about their best opportunities, you have to give them too much; and you will have to trust either a company that has the infrastructure and, to some extent, regulatory oversight, but a bad history of data abuse, or a company that only came into existence a few weeks ago. And that's. that's just not a decision that I think most people want to make right now.
Another thing, I know that in the case of Underdog, in particular, their main feature is not only that everything is local and based on trust, but also that they are changing the business model around this. And I think we see elements of that in Meta's Muse, but the idea is that they don't want to use information about you for targeted advertising or anything like that. It's more about getting small commissions, uh, and working with Stripe in particular, which I think is one of the investors in Underdog, to take a percentage of the transactions that you commission that agent to execute through Stripe's processing system, which is interesting and somewhat reassuring because they don't want to use or abuse your information. But on
the other hand, if this is your business model, then the agent you create will have some interest in constantly pushing me to make purchases because you want to make money from these transactions. And I'm not sure I want to use agents for that. We don't have as much money as programmers in Silicon Valley, right? That's just not true, and again, there's a certain skew in what all the incentives are, starting from step one. So, I wish you success. And I also think: well, good luck creating a business model without advertising at this stage. You know, everyone tries to avoid advertising, but then they still have to implement it, as we see in the example of OpenAI, because it's very expensive. I hear from people that these models, well, they're
agents, right? They spend hundreds of thousands of tokens on one simple task, or even no task at all. Therefore, it seems that this will be a very expensive pleasure for these companies. I doubt if they are anything more than just shells for agents that do not belong to them. I'm not sure. I don't think any of them create their own models. I would say no, they don't They probably just use whatever is at hand, whatever is open. So yeah, I think it's going to get very expensive for them, and I'm not sure that a small fee from Stripe will help them achieve any kind of profitability. But I also wanted to mention before we change the subject that the timing of all this is very interesting.
It's really ironic, is n't it? We still haven't recovered from the problems with OpenAI and Hugging Face, from " rebel agents", from the idea that agents can perform actions that will lead to disaster, which you definitely don't want to see. And then they say: "But look, let the agent take over your personal life." How to understand this? This seems to be the riskiest moment for such steps, when we don't even know if they are safe. These companies sell privacy solutions that they simply cannot guarantee because they do not own the models themselves. They don't know how these models are set up, right? And it feels, to me, cunning. This timing seems like a thief's genius, because when we reach the peak of convenience, I think many of those who
fear AI and demand it be regulated or banned will say, " No, that's fine." Like, well, it made my life a lot easier, so I guess it's normal. Well, we'll see. I mean, another point worth mentioning in this thread is that when we talk about trust, it's not just about gaining the trust of consumers or even businesses, but also that there are partners who also need to, in quotes, work with agents in a certain way. And so, last week, news broke that Apple was restricting access to disks for AI agents , and a number of websites were making it difficult for them to do their job, or I don't know if they were creating the difficulties or if people were just more likely to encounter obstacles when trying to buy airline tickets through Muse, and United
said, "No, we do n't allow that." We do not support this. And so I think there's probably going to be some changes, but it's also fair for companies like United to say, "Hey, you can't have unlimited access to buying tickets on our site." Half of them will eventually end up being refunded because people did n't really want it. So I think there needs to be this trust-building process, this period of negotiations. I was just going to say that Amazon did this. That's right. Amazon did this with Perplexity and others. They stated, "No, you cannot use your agents to shop on our platform."
Everyone has to use Rufus, or now Alexa for shopping, but still. We really need to move on, that's why. Yes. But no, it's a great reminder, of course, that safety is important. Safety is often an excuse for other types of protective behavior. But let's talk about other startups. First up, we have Lambda, which I believe is one of the neo-cloud startups that we talked about earlier. Rebecca, tell us what's going on with them Yes, this is a startup in the field of neo-cloud AI computing. Wall Street Journal was the first to report this. This could be a round of up to $4 billion at a preliminary valuation of the company of
$14.5 billion. And this could be their last private round before their planned 2027 IPO. I'm personally skeptical that they'll go public, because they were supposed to do it this year, and now it's maybe even late 2025, so it looks like an attempt to buy time. But yes, this is a fairly significant fundraising effort I think especially for companies like this, although in general we talked last week about how Aura canceled its IPO. But I think we're waiting for Anthropic's results to see how many more companies will dare to take this step. On the other hand, OpenAI appears to be preparing to close a $30 billion funding round; and if there is still so much liquidity in the private market, it could deter many from going IPO next year.
Yes, I'm also curious about what the appetite will be for "neo- cloud" companies, because on the one hand, they have real revenue that many other AI startups don't have. But on the other hand-what are their long-term prospects? It is obvious that intensive construction of data centers is currently underway. Although with some resistance and a slowdown. So this question is still open, perhaps they are selling something that is valuable now, but may lose relevance in a few years. Oh my God, you hit the mark. I think there's still a lot of interest in them in the short term because they have the capabilities that everyone needs, right? But in the long term, it's difficult, because the publication studied a letter to investors, which
indicates that the order book grew from 15 billion in June to 50 billion in September. This looks serious, but $35 billion of that amount appears to have come from just one client-Anthropic, which signed a deal with Lambda in late August. In many ways, the problem with such "neo- clouds" is that their entire business depends on one or two large clients. This is not revenue, it is just a backlog of orders. This money will not arrive soon, and the contracts themselves may be terminated. If demand doesn't meet expectations for whatever reason, right? So yes, it seems like a pretty risky thing to do. And also, I just want to point out that Lambda is probably raising this round of equity, but that's in addition to four separate debt raises in the last
few months or maybe the last year, I don't even know. Data center construction is largely financed through debt, and you see this trend, or I see it among non- cloud services as they prepare to go public: they raise capital through equity to look serious, and then they take on a lot of debt to finance the construction. You see this with Enscale, which also recently raised, I think, about $2 billion earlier this year, and then did two or three more debt rounds, including the one that TechCrunch reported last week. So, yes, there are a lot of moving parts here. A lot can go wrong. Yes, we will keep an eye on this. In the meantime, we have one final deal that I'm happy to say is, to my knowledge, unrelated to AI, and that's Uber entering the
catering market. Sean, what's going on there? I can bring AI in here if you want. I.I won't No, please. Actually I.yeah, I mean, no, I won't. Uber is buying a company called ezCater for $2.3 billion in cash. Good for catering business. I'm sure they use a bit of AI because they're not a catering company per se. They don't come and set out food on your two-meter table. They are more like a marketplace that connects catering companies that prepare food with businesses that want to order it.
The idea is that Uber is making this acquisition so that.well, Uber is trying to grow not only its core business but also Uber Eats, and it has done that, but it has also been achieving both goals through numerous acquisitions, partnerships, and investments. So this is a way to expand Uber Eats into a market where Uber is currently barely represented-it's something like office catering or event catering. You know, the way Dara Khosrowshahi put it in the press release about the acquisition is really about - and I think if you read between the lines, which I will now - there are a lot of restaurants on Uber Eats and other services (DoorDash or whatever) that don't always feel like they're getting the best deal and are kind of under
pressure from the platforms and their incentives and the commissions that they charge and how hard it is to operate in the food delivery business when you already have low margins. So this is a way to potentially increase volume, it gives these restaurants another sales channel for their food, right? For example, you can now provide food for catering events, which was previously unavailable to you. So my question is, does this change, and how much, the way a restaurant perceives its business? Does this make it more viable than usual, as it is usually a very complex business? Will this just be a way to get a little extra money for commercial kitchens that already exist? And I have to quickly say that this is a cool exit story, in my opinion.
I was looking through the archives and saw that we wrote about this company back in 2014. They were founded in 2007 and developed on their own for about 7 years before attracting their first investment. So getting to the point where you exit for $2 billion in cash in 15-20 years is a completely different story than what we see today with many AI startups. Yes, that's pretty much what I was going to say too. This story takes me back 10 or 15 years ago, when there were many different catering startups that were again saying, "We do catering, but we're not a catering company, we're a platform, we're a marketplace." And you know, this excitement somehow hasn't survived to this day. So it was a very pleasant surprise for me that they not only stayed afloat and
not only sold, but sold, as you said, for $2.3 billion. Uber likes to buy companies. They like it. I mean, I won't go into it, okay? But Dara Khosrowshahi ran Expedia before coming to Uber, and I think it's worth thinking about when you look at how he's guided Uber's business at all levels. It seems logical to me that he sees great value in being not only what the company is known for, but also a place where there are a multitude of options from other companies and services. And he certainly grew the business that way, making it profitable for the first time. So yeah, that's how he achieves this engagement, right? I wrote about this a few years ago, about the " Amazonization" of Uber, you know? You just plug into different platforms until
Uber becomes the one that transports people and goods, right? Well, I think we can end it there. We're almost out of time. Equity will return next week. In the meantime, you can find Equity under the nickname @EquityPod in X and Threads. Thanks for listening, see you next time. Equity is hosted by TechCrunch senior reporters, producer-Teresa Lo Consiglio, editor- Kell. Subscribe on YouTube or wherever listen to podcasts, and stay tuned for upcoming events at techcrunch.com/events.
Thank you very much for listening to us, until the next meeting.
