Everyone said AI was going to kill the app store, but it turns out the exact opposite is currently happening. App launches are up 60% year over year and over 100% so far in April. So we're going to break down why and spoiler alert is because of AI. OpenAI is shutting down Sora next Sunday, April 27th and the head of Sora, the product chief and the enterprise CTO all walk out the door on the exact same day.
Something is happening inside of OpenAI and I want to talk about it. The next thing is that Sarah brass, the AI chip company just filed for an IPO, $23 billion valuation and a $10 billion deal with OpenAI and a new agreement with AWS. The CEO is openly saying they took the fastest inference business at OpenAI away from Nvidia. This is quite a big claim.
The next thing is that Stanford's new AI index just dropped and the performance gap between the best US and Chinese models has collapsed to a 2.7% lead in favor of the United States. Also in the same report, AI agents jumped from 12% success on real computer tasks to 66% in one year. So we're going to get into all of that and personally, I'm seeing a lot there. The other story that I want to talk about is that anthropic just passed OpenAI in revenue, $30 billion in annualized versus $25 billion for OpenAI and they did it while training their models with way less compute.
This is probably the most insane thing I've seen in a long time once you start looking at these charts. And if you can try to figure out how to actually use AI for your work, not just chatting with one model, but actually building little automations to save you time, check out my own startup, AI box.ai. It's a tool I have personally used a ton. You get access to something like over 80 plus AI models inside one account.
So if you're already paying for chat, GPT or Claude or Gemini separately, then you stack them all together and get it for $8.99 a month over on AI box. You can save all of those subscriptions. The part that I love is that we have an automation builder. You literally describe what you want in plain English and our AI builder will link the other different AI models, input prompts and build the tool of the automation for you.
You can go check it out at AI box.ai. There's a link in the description. So there's been this whole narrative for the last year that AI chatbots and agents were going to kill apps. Why would you want to open a weather app when you can just ask Claude or why would you download a fitness tracker when an AI assistant can handle it?
So you've heard this take from Karl Perry over at Nothing. You've heard this from Johnny I've working with OpenAI on whatever AI hardware thing they're building and basically every other tech writer with a sub stack. So it was pretty shocking to me after seeing all of this app figures just put out new numbers and it is the exact opposite worldwide app releases in Q1 and 2026. So 60% year over year across both Apple's App Store and Google Play and on iOS specifically it's up 80%.
In April so far the month that we're in total app releases are up 104% compared to the same window last year. Apple's Greg Josswiak put it really well. He said the rumors of Apple's stores death have been greatly exaggerated and basically productivity apps are the top five categories for the first time utility jumped to number two. This is what I think personally is happening.
I think vibe coding is working. Claude Code, Repllette, all of these tools have gotten so good that people with ideas and no engineering background are shipping apps. Right now I believe it truly is the Gold Rush and the App Store is basically benefiting in a huge way. And by the way if you want to learn how to grow and scale your own business and take courses and learn how I personally am vibe coding.
I actually have a school community called A Hustle School Community. There's a link in the description. We talk about everything that we're doing with AI to grow and scale our businesses, how we're vibe coding, how we're putting stuff out there. You go check it out.
Okay, the next thing I want to talk about is opening eye and Sora. There's actually a huge story about opening eye right now. Basically they are discontinuing Sora. We've talked about this a ton.
The app officially closes on April 26th. So six days from me recording this and the API follows in September. So basically this is their text video product. They launched this with a massive amount of hype last year and it peaked at around a million users.
Then it collapsed under 500,000 while they were burning somewhere between $15 million and $15 million a day in compute, depending on which report you believe. I've seen some eye watering reports on the costs of this. Either way, it was not sustainable. Then on basically the same day, the shutdown timeline locked in, opening eye lost three of their major executives.
So they lost their product chief, Kevin Whale. They lost Sora head, Bill Peebles and their enterprise CTO also was going. So this wasn't fantastic. They had three exits in the same 24 hours.
What I think is happening right now is that opening eye is making a very hard pivot. They're calling, you know, they're basically getting off of these side quests. They're focusing on coding and enterprise, which makes a lot of sense. That's where entropic is absolutely eating their lunch right now.
We're going to get into this on the deep dive later on. But killing Sora, I think is a really good signal in this direction. And the reason why is because entropic for the first time, which I'll talk about later, has officially passed opening eye in annual recurring revenue, run rate, and that is something I did not see coming from a lab that is raised less and, you know, I mean, basically there's way less users. Okay.
The next thing, though, is that Sarah Bress has just filed to go public. This is the AI chip company. They're based in Sunnyvale and they're basically, they're pitching themselves as the fastest hardware for training and inference. This is basically an actual competitor to NVIDIA.
And I think on the part of the market that NVIDIA currently owns completely, they're, you know, competing with them directly. So as a bit of a background, they tried to IPO back in 2024 and they had to pull it because of a federal review over an investment from G42 and Abu Dhabi. They came back with a, you know, $1.1 billion series G last year and then they did another billion in a series H in February at a $23 billion valuation. They did $500, $10 million in revenue in 2025 and they had a net income of about $237 million.
So although that number has, I think, some, you know, one-time items in it, it's still pretty impressive and they're shooting for mid May for their IPO. So the thing that I think was the most impressive in all of this, they've announced a deal with AWS to put these chips inside of Amazon data centers and have a deal with OpenAI that's reportedly worth more than $10 billion through CEO Andrew Feldman went on record saying, quote, NVIDIA didn't want to lose the fast inference business at OpenAI and we took that from them. So I think this is a big claim, right? They're saying they're basically stealing business from NVIDIA.
OpenAI's compute strategy has basically been the single most NVIDIA dependent story in AI for years. If even a slice of their inference workload is moving to Sarah Bress, I think that is, you know, the first real crack in NVIDIA's monopoly at the frontier. Definitely worth watching where these IPO prices go. I'm excited about this.
Samford's new AI index is out and there are two numbers in it that I want to call out specifically. First, there's a performance gap between the best American and Chinese AI models. It is now 2.7% back in May of 2023, depending on what benchmark you looked at. The gap was somewhere between 17 and 32 points and China is doing this in, you know, for way less money, US private AI investments was about $286 billion last year in China.
It was about 12.4 billion, roughly, you know, 23 times less. And their top model is under three points behind. So that is pretty insane. The second thing I want to cover is that AI agents went from about 12% success on real computer tasks a year ago to 66% now.
So this is agents actually navigating software, you know, they're clicking through forms, they're pulling data, they're finishing multi-step jobs in real systems. You know, 66% is definitely not good enough to let something completely loose on supervised on anything that really matters. But it's good enough for giving an agent a really narrow bounded job and checking its work, making sure that it's getting stuff done faster, you can scale it up and having, you know, more of these doing things than yourself. If you haven't tried giving an agent a real job in the last couple months, I would definitely try again this week.
It feels like we've moved forward a ton. I've been talking about it a lot on the show. I mean, as we're speaking, I have a monitor off to my side that's running cloud co-work. It is setting up databases, migrating stuff, importing data for projects I'm working on.
And it's honestly, these numbers were Stanford saying like 66%. I'm curious the model. I'm curious the timeframe on this. I'm getting like 85% of tasks completed completely.
Some tasks that just struggles with, but like the task is, I know, you know, building websites, gathering data, parsing data, sorting stuff. It is just insanely impressive. So, you know, I think this is basically something that everyone should be trying right now. Okay, something that shocked me, absolutely shocked me is that anthropic has just passed opening an annualized revenue.
The run rate is $30 billion. Anthropic has about $25 billion over opening eyes. So the comparison between these two companies, I mean, anthropics have $5 billion. And for full context, because I think this is, there's a couple different pieces to this versus the growth curve.
And anthropic was at roughly $9 billion in annualized revenue at the end of 2025. So, you know, just in December, they hit $20 billion in early March and 30 billion in early April. So March to April, $20 to $30 billion. That's a $10 billion jump.
It's taken years to get here. They were at $9 billion in 20, you know, in December. So since December, till now, we go from $9 billion to $30 billion. Guys, this is mind-boggling.
This is insane. This is, you know, the absolute greatest run ever. And, you know, to be honest, I'm not shocked. I use, I've completely stopped using chat GPT.
I use anthropic for absolutely everything right now. And the thing I was recently talking to my wife about with this is, you know, with chat GPT, sure, you can chat with it. It's great, whatever. But anthropic is literally doing things for me.
Like, as we speak, I have agents on my side panel that are gathering data, building out websites, doing reports, doing things that there's no API for, they scrape websites for me and gather information. They help me put together these briefs. Like, so much is happening. And it's not like, you know, oh, there's like a little custom GPT that I can go to and use over and over again, like I was doing at opening eye on chat GPT.
I have scheduled tasks every day at 6 a.m. I have a scheduled task, get up, log into all of my podcast distribution platforms, gather all of the data that there's no, no API for uploaded to a spreadsheet, uploaded into, uploaded into like a dashboard that cloud code built for me. And I get to see data I've never been able to see before. It's incredible.
It saves me, I mean, it's just something I would never do. So I'm saving so much time. It's just amazing, right? Okay.
So enough from beyond that. Okay. The next thing I want to talk about is a customer mix, specifically about 80% of enthroughts revenue comes from business customers. Opening eye is much more consumer heavy.
They still have 900 million plus weekly active users on chat GPT. This is, you know, this is a massive moat. But just the fact that so many people are using chat GPT. But I think the enterprise side is where the actual dollars are.
And that's the side to anthropic absolutely owns right now back in February. They had over 500 customers each spending more than a million dollars a year on cloud. But April, that number is over 1000. It doubled in about two months.
The other thing that I think is really underrated is that anthropic is doing this while spending roughly four times less on training than opening eye. That is insane. They're making more money with smaller training runs. And if that holds, basically the economics on the anthropic side look way better than it for anyone else on the frontier.
I think basically that's what investors are going to keep pulling on, right? And then in addition to all of that, you can look at all of the compute deals anthropic has going on right now, right? They have Google Broadcom about about 3.5 gigawatts of next gen TPU capacity. And that starts in 2027.
And on top of the roughly one gig of Google compute, they already have lined up for 2026. They have 3.5 gigawatts. This is, you know, one of the largest publicly announced AI infrastructure commitments ever for scale. That's roughly the output of three plus nuclear reactors dedicated to running cloud.
So what does this actually mean? This is what I think is the most interesting with all of this a year ago. You know, the working assumption in AI was that opening I had basically this insurmountable lead. They had the brand.
They had the compute, you know, the consumer scale. I mean, even today, 900 million weekly actor users is mind-boggling. You know, they had the Microsoft partnership as well and a lot of distribution. And that assumption now is broken.
I mean, just look at Microsoft and Thoropic has gotten in there with all of the Microsoft products and has cloud running inside of all of them. And I hear people using that all the time. It also turns out, I think having the best coding model and being the kind of enterprise default is way more valuable than being the consumer default, at least while enterprise is other ones actually paying, you know, full freight for AI. Personally, I'm paying $200 a month for cloud code and I never really paid that for opening up because I didn't get the value out of it right now.
I'm getting, like, you know, I would say I'm getting $30,000 a month of value out of cloud code if I had to pay someone to do all of the projects that I'm working on. So it's just the ROI is so insane to me. And I think we also have, you know, probably a moment for opening eye where they're shifting, they're shutting down a bunch of side projects, side quests, like even Sora so they can focus on this. So I do think we're going to see a pretty strong, I don't think opening eyes going down without a fight.
I think we're going to see some really strong competition in the near future, but it seems like anthropic is really pulling ahead in a certain, in a few certain areas of one of those is just pulling in pure money. That's the show for today. Thank you so much for tuning in. If you could do me one massive favor that would help the show out so much.
If you could leave a review on Spotify or Apple, it basically helps the show get found out by more incredible people like yourself. Also, a quick reminder, go check out aibox.ai. If you haven't yet, there's over 80 models building your own automations. It's 899 a month.
The link is in the description. All right. I'll catch you guys all in the next episode.