Snap’s 16% Workforce Layoffs, Allbirds’s AI Pivot, Anthropic’s New Opus 4.7 Model, SpaceX IPO Index

0:13Stephanie Palazzolo:Welcome, everyone, to The Information's TI TV. My name is Akash Pastracha. It is Wednesday, April 15th. Before we kick off the show, I want to do a special shout-out to a new billboard up now in Times Square, thanks to our sponsorship with the New York Stock Exchange. Check it out if you are in New York City. It'll be up for the next two weeks. First up today on the show, we have got Snap Layoffs and Allbirds AI Compute Pivot. We'll dig into both of those stories with our newsroom. We'll then discuss exclusive reporting from the information around how Anthropic is changing its pricing structure.

0:46Stephanie Palazzolo:We're also tracking news that an Anthropic executive stepped down from Figma's board. The timing of that is not so coincidental. We'll talk about that shortly. We've got our crypto reporter coming on to talk about her story. Our senior finance editor is coming on to talk SpaceX and the business of stock indices. And we will wrap the show with a special roundtable with our partner, Adobe. It's going to be a busy show, but a fun one. So let's get right on into it. Snap is cutting 16 % of its workforce, about 1 ,000 people citing AI efficiencies. I want to bring on co-executive editor Martin Pierce for his thoughts.

1:22Stephanie Palazzolo:Martin, welcome back to the show. What do you think? I think this is performative. Snap, this is not going to help Snap. This company just does not have the scale to survive. I mean, just to give you an example, Evan Spiegel in the post announcing this says that this will cut costs and enable or it will clear the path for Snap to achieve or to improve or achieve maybe net profitability. profitability. Why that is important is SNAP is already profitable on the only metric that really counts, free cash flow. Net income is a accounting fiction. So this is just ridiculous. It doesn't mean anything.

2:08Okay.

2:09Stephanie Palazzolo:How does this align or not align with this whole spectacles push that they're making? Did we get any details about that? No. I mean, and also, I mean, the thing to remember is that at the end of last month, an activist investor which has bought into Snap called on the company to cut costs to reduce the number of employees and also to spin off that Spectacles business. So Snap has done the easiest thing that it can do, which is to cut employees. It is not given any indication of its plans to spin off the Spectacles business, which is Evan Spiegel's, you know, prize, you know, prize thing, prize.

2:51project uh until unless he does something like that he's not really serious about fixing snap's

2:58Stephanie Palazzolo:problem so okay last question for you martin i mean you know we have talked on the show about how meta has used ai to very much um reinforce its advertising business do we have any sense uh here if ai is helping snap sell more ads at all is that a story that's happening um i don't know whether AI is helping them. Maybe. I mean, advertising is the part of their operation, which is not doing that well. They did say today that AI will help them operate without as many people, which is what everybody is talking about. So I think that's where AI is really helping them. Great. Well, Martin, I want to thank you for coming on.

3:42Stephanie Palazzolo:That is Martin Pierce, co-executive editor of The Information here on TITB. Allbirds is pivoting to AI Compute. This is the shoe company whose stock has taken a beating over the past few years. Shares are up more than fourfold this morning. I want to bring on Anissa Gardese, our cloud and compute reporter at The Information to break it all down. Anissa, welcome back to the show. Did you ever think we would see the day i didn't um this is not something that i probably could have predicted um but maybe now we have to look at distressed public assets and wonder whether they're going to pivot to gpu cloud um rentals so i guess it will be on my radar from this point moving forward so i mean let's just make it clear for people what exactly is all birds trying to do here what's its strategy?

4:33Allbirds announced that after essentially selling its core shoe business, that it is going to raise money and pivot the public shell company to a new firm called New Bird AI. And what they're planning to do with those funds, they're planning to raise around 50 million is to get into the GPU cloud game. So they might buy GPUs and rent them out to other people and, you know, completely pivot what the public company does.

5:01Stephanie Palazzolo:So this is basically like a smaller version of a core weave or a nebbiest type of NeoCloud offering. Is that the idea? You're right. So a lot of the existing GPU cloud providers were in a different business before they pivoted to AI cloud. But I think the big difference here is that, you know, the pivot from Bitcoin mining and renting out servers to Bitcoin miners to then renting out servers to AI developers is a little more similar than this pivot, which we're seeing right now, which, you know, this new bird AI company has really nothing to do with the original Allbirds public company. And I mean, you've written about all these NeoCloud companies at length.

5:44Stephanie Palazzolo:You know, is there any differentiation strategy here for Allbirds, let alone anyone else in the industry? As far as I can tell, this NewBird AI company doesn't seem to have a particular special sauce that I think will catapult it ahead of the existing NeoClouds. The main thing that I can point out is that they're a public company, and so it might be easier for them to get financing, and they might just become a GPU financing company. But I don't think there are too many similarities between the original company and what they're planning to do now. So maybe this will just be, you know, okay. Let's be clear.

6:23Stephanie Palazzolo:Let's be clear. I mean, there's going to be people on the floor building these facilities and they're probably going to be wearing the shoes. So, I mean, there's that, right? They might be able to move faster in between the rack. Did you have the shoes? I did not consider that. Do you have the shoes? I don't have the shoes. Okay. I'm wearing them. Before my time. I might consider selling them now that they'll be clearly a relic. Okay, well, Anissa, I want to thank you for coming on. It is certainly a fun story, and I'm waiting for the CoreWeave Allbirds acquisition. I think maybe that could be in the cards, but we'll have to see.

7:03Stephanie Palazzolo:That is Anissa Gardizi, our cloud and compute reporter, here at The Information. Anthropic is making a big change to its pricing model as usage of its products ramps up. My colleagues Kevin McLaughlin and Aaron Holmes broke that story. I want to bring on Kevin to tell us more about it. Kevin, welcome back to the show. It's great to have you here. It's great to be here, Josh. Okay, so what do we know about this new pricing model? So it's a little bit nuanced, and I see some little bit of misconceptions in the response to our story. So to be clear, Anthropic started moving some of its subscription plans to a consumption-based model from subscriptions last November.

7:43What we reported on was specifically about Claude Enterprise, which is a formerly a subscription-based offering for companies with 150 or more employees, which included advanced security controls, administrative controls, things like that. And so earlier this year, Anthropic changed its policy so that when customers of Claude Enterprise would come off of their contracts, they wouldn't be able to sign up or continue on subscriptions. they would have to switch over to the consumption-based model, which is using the Anthropic API. So Anthropic didn't announce the change. And so it's sort of becoming apparent to some CIOs now as they have these discussions with Anthropic and seem to be suddenly realizing, hey, we can't buy AI the way we've been buying it in the past.

8:33Stephanie Palazzolo:So I just want to get a couple of details clear here. So when you say Claude Enterprise, this includes Claude Code? Because I know there's the whole API business, and then there's sort of the classic Claude chatbot. What are we exactly talking about here? Sure. Claude Enterprise is basically everything but the kitchen sink. It's got Claude Code. It's got co-work. And as I mentioned before, enterprise versions of products typically include things that companies need to do to adhere to regulatory compliance. and other factors and to be more secure as they're operating. So this is specifically about the Cloud Enterprise product.

9:13Stephanie Palazzolo:Okay. And so in the end, I mean, this is likely to be more expensive for most of these customers or what's our guess? Yeah. So essentially what's happening is the subscriptions that Anthropica's often previously used to come with a bundle of compute. and in many cases, you know, companies would be able to just use and operate. That was enough for their needs. But in other cases, you had power users that were coming up and hitting the limits on those subscriptions. And so what Anthropic told us is that they want to make it sort of equivalent for everyone. And so, you know, there won't be cases where power users are getting error messages saying, hey, you got to pause your work for two hours because you're out of capacity.

9:58And similarly, in some cases, there were users that were not using all the capacity that they were allotted. And so, you know, that was sort of a situation for customers as well. So Anthropics' stance is that having it be based on the API makes everything more predictable. There's no more surprises, and customers are getting exactly what they're paying for.

10:19Stephanie Palazzolo:Now, that's how it's supposed to work in theory. Do we have any sense for how customers are actually reacting to this on the ground? You know, I actually talked to several, it's kind of interesting, ISVs that are sort of customers of Anthropic and in some cases competitors doing AI coding, they were already on the API. So it wasn't a case where they had a jarring change. I think the change is really sort of specifically about, say, you know, insurance companies. If you're a CIO in an insurance company, you would be on cloud enterprise and you are equipping your developers with AI in this fashion.

10:57And now you're having to sort of switch to this new model. I don't think it's like the costs are going to go up, but in reality, Anthropik has been subsidizing its compute in order to build market share. And, you know, Anthropik is no longer operating at a startup upscale this is a major enterprise software vendor and you know as it grows up so to speak uh anthropic is just sort of shifting and and maturing and pricing its products in the same

11:27Stephanie Palazzolo:way as these other companies do right I mean I mean that's sort of my reaction to your story was this was kind of inevitable right I mean this is how most of these products go is that you get a lot of venture funding and you subsidize it you get traction and then you realize how much it actually costs to run the thing profitably. And so I think a lot of people expected this to happen. I guess just extending this a bit though, Kevin, you know, we are at a point here where AI adoption is not what people thought. And that was with the subsidized products. And now we're at a point where the products maybe aren't as secure, as good as people hoped.

12:03Stephanie Palazzolo:I'm not talking necessarily about Claude here. I'm talking about AI altogether. But maybe the products weren't as good as we thought. They're probably going to get more expensive. I mean, this doesn't seem like great news for the whole AI adoption story altogether. I guess you could say that, but in some sense, I think everyone knew this was coming, and it's certainly not surprising. I think if you're anthropic, the time to rip off the Band-Aid is now because you've got wildly popular products like Cloud Code and now Cloud Cowork, which the latter is opening up AI to users that are not developers and other types of users within organizations.

12:42And so their usage is skyrocketing. If people like the product, they're not going to run away if the price goes up. I don't think they will. And we spoke with a licensing expert in the story who said exactly that. People are not running away because they like the products and they think they're getting value. Okay.

13:01Stephanie Palazzolo:So let's now turn to other competitors in the story here. Is Is there any way that you think OpenAI could leverage this at all and gain market share? Sure. OpenAI recently introduced a new$100 per user per month subscription. I think it was last week it was announced. It is pretty clearly an attempt to sort of capitalize on any anxiety that Anthropik customers, enterprise customers might feel. But in reality, OpenAI is going to eventually have to do the same thing. There's no way that they're just going to be able to forever say, okay, well, you know, you can keep buying subscriptions for us forever and it comes with a set amount of compute.

13:43It won't be long before OpenAI is doing exactly the same thing.

13:47Stephanie Palazzolo:And last question for you, Kevin. I mean, we've seen consumption-based pricing play out with public enterprise software companies and enterprise software at large. as you sort of think about how that shift from seat-based pricing to consumption-based pricing has gone broadly in the software sector. Are there any lessons you think that Anthropic and eventually OpenAI can learn from those older companies per se? Yeah, I think the lessons, I mean, in some cases, the lessons are very hard to learn and you have to learn it after there's a backlash from users. I don't think that that's going to be the case here because, again, as I mentioned before, if companies are getting value from AI, they're going to pay more and they're going to keep buying those products.

14:33And they might complain about it on Twitter, but in reality, or X, but in reality, they need AI and AI is helping them transform their businesses. And cost is obviously an issue, but I don't think it's going to be a deal breaker for most customers. Great.

14:51Stephanie Palazzolo:Well, Kevin, I want to thank you for coming on. That is Kevin McLaughlin, our enterprise software reporter here at The Information. In other Anthropic news, there is a new model on the way, according to an exclusive report from my colleague Stephanie Palazzolo. I want to bring her on to share with us what she knows. Stephanie, welcome back to the show. I want to talk to you very quickly. We saw the news that an Anthropic executive has left the Figma board. Is that a coincidence here? Walk me through what you made of that. Yeah, so we don't know what the exact connection is, but I would say the timing is definitely a little bit suspicious.

15:28So that exec was Mike Krieger, who is actually a co-founder of Instagram and originally joined Anthropic a couple years back to be its chief product officer. But then at the beginning of this year announced that he's kind of shifting positions within Anthropic to lead its labs project. And so essentially, you know, we learned this morning that he has stepped down from the board of Figma, which is a design startup. And yeah, I think that the interesting connection here is that this is coming pretty shortly after we broke news that Anthropic is working on its own design tool for startups. Sorry, own AI-powered design tool for its customers that it can release as soon as this week or even today.

16:12And so obviously, I think there's a potential risk of competition there for companies like Figma. And so I think that could explain why Mike decided to step down from the board yesterday.

16:23Stephanie Palazzolo:Well, and it's also making me think about how many more board changes we could see, given just how interconnected the AI sectors and sort of the old school enterprise software sectors are in terms of board members. I mean, you know, there are many different connections. But, you know, I do want to talk to you about the design tool. So you broke the news that they're releasing this new tool, and then they have a new model coming up. What do we know about the new model? Yeah, so the new model is going to be Claude Opus 4.7, which, you know, at this point, it's very hard to keep track of all the what specific point we're on with every single new model release.

17:07But this is essentially going to be Anthropics' next flagship model, although this isn't the same as their Claude Mythos model, which is this kind of secretive, super high-powered model that they've been talking about for a little bit that's currently in testing phases with early partners. So again, this model could be released as soon as this week.

17:29Stephanie Palazzolo:Okay, and so it's not the Mythos model, but do we have any idea, you know, what this 4.7 iteration is supposed to be better at? Yeah, so, you know, we're not exactly sure, but I would guess that based off of the timing of this model being released around the same time as the AI design tool, I imagine that it'll probably be better at kind of coming up with designs and editing presentations and websites and landing pages and that sort of thing. So I think there'll probably be a lot of improvements there, along I'm sure with things like coding, which obviously has been a really big focus for the company.

18:06Great.

18:07Stephanie Palazzolo:Well, Stephanie, I want to thank you for coming on. That is Stephanie Palazzolo, our AI reporter, here at The Information. Crypto exchanges are looking to bolster their defenses from possible AI threats as companies race to create the most powerful models. Our crypto reporter Yueqi Yang wrote about this story and she published that piece this week. I want to get her now to explain to us what she learned. Yueqi, welcome back to the show. It's great to have you here. Hey, Akash. What was the biggest takeaway in your reporting, Yueqi? So my biggest takeaway is that crypto companies, including the major crypto exchanges, custodians such as Coinbase, they're rushing to prepare for the arrival of powerful AI models, especially coming out of Anthropic, because these models are expected to increase cybersecurity risk.

18:59And they are getting in touch and seeking access to preview the Anthropic mythos model because they wanted to be able to prepare for it.

19:09Stephanie Palazzolo:And so who is getting access to this mythos model? Because we've talked about this on the show. I mean, it's not available for public release. It's only very specific institutions that are getting access to it. Is it crypto firms? Is it banks? I mean, who is it? Yeah, so from my research and reporting, crypto companies are not part of the first group of companies that are getting access to Anthropik's model. And Anthropik seems to have prioritized the biggest tech companies, the big banks. We know that Amazon, J.P. Morgan, Goldman Sachs, These are the kind of companies that are already getting access to it, and they're testing their model to try to find vulnerabilities in their systems.

19:56But crypto companies are also trying to get access to it, but they're not having success yet, at least from the reporting that I've done. And it is important for crypto companies to try to stay on top of this because they guard billions of dollars of client assets. And they're also very concerned about any potential threat that could come from these powerful AI models.

20:23Stephanie Palazzolo:But you did write that Binance is testing some new models. Yes. So some of the companies that I talked to, Binance included, Fireblocks is another one. They're testing commercially available models from Anthropic. So these are the earlier models. And Fireblocks told me that even just on the basis of the prior anthropic model, they're already seeing huge implications from it. The model was able to detect vulnerabilities that no other security audit firms were able to find for Fireblocks. And Fireblocks has already fixed the vulnerabilities detected by the model. You know, I have to do some research on this, but I'm sort of wondering, you know, we hear so much about the attacks.

21:10Stephanie Palazzolo:against crypto exchanges, crypto wallet companies, for example. They're very highly publicized. I'm sure that banks see the same number of attacks, if not more, just given the scale of these banks. We don't hear about them as much, probably because their defenses are probably a bit better and it's not as shiny as crypto. But I mean, I've got to imagine that given the extent of the hacks that we've seen in crypto, I mean, I'm not saying that they should get access to mythos, but it would do them some good to have some of these cutting-edge AI defense systems to protect against this stuff, right?

21:46Yeah, that's exactly right. Crypto companies are high-profile targets for hackers, especially nation-state actors. Even just this month, there's been a few high-profile hacks in crypto. There's this one decentralized crypto platform called Drift that was hacked by people that they believe were acting on behalf of North Korea. And this is actually done through social engineering. So the bad actors were people who met in person with the Drift team at crypto conferences and built relationships with them over the course of six months and eventually got into their system. So yeah, so crypto companies historically have a problem with hacks.