Big Tech's AI Plan Has Failed

Hey guys, it's Sasha. Apple launched Siri back in 2011, long before the world went bananas for AI assistance. And over the next 11 years, Apple worked tirelessly to make Siri progressively more useless. Then Chad GPT arrived. The entire tech industry decided to have a massive bonfire with their giant piles of cash and Apple sat around sucking their thumb. Eventually, they announced a product called Apple Intelligence. It is now 2 years since that announcement and Apple Intelligence has broken Siri's record for the most useless innovation in history. As usual, Apple nailed the marketing. They just forgot to build the intelligence part. Then there was Project Titan when Apple decided to try and build a self-driving car. They spent 10 years doing it. And in 2024, Apple fired the entire team and shut down the project when they realized that it was easier to just keep pretending their iPhones are getting thinner. And now, for the past 3 years, the tech world has been laughing at Apple because every single other tech company is incinerating hundreds of billions of dollars to build their own AI project. At the earnings call last week, Mug Zuckerberg has announced that Facebook is planning to spend up to $145 billion on AI this year. Google said that they're spending $190 billion on capex for AI. Microsoft also plan to burn $190 billion. Amazon have gone a little bit higher. They are planning to spend $200 billion. Most of this money is being spent on building giant data centers and infrastructure for training new AI models and then using those data centers for inference afterwards. Apple's total capital expenditure last quarter was $1. 9 billion, which is down 36% from the year before. Google is spending 190% of last year's profits on building out data centers for AI. Facebook is burning 240% of last year's profits. Amazon is going for 260% of their profit. Apple is spending about 5% of their profits. And that's across the board on all capex, not just on AI. Apple has spent over a decade getting everything about AI wrong. They were early to a party nobody else went to with Siri. And then they were so late to the actual party that it's now too embarrassing to try and go. It's past the time when you can claim to be fashionably late. It would be like turning up sober to a house party at 2 a. m. when everyone is already naked in the pool. And the thing with this party is that although it costs a lot of money, it really costs a lot on data, specifically your data. Because when everyone is naked in the pool, nobody cares much about privacy. You know who does care about your privacy? Cyber GhostVPN, the sponsor of today's video. In a world where companies constantly collect data about what we do online, protecting your privacy has never been more important. Think about how often you connect to public Wi-Fi on a train, in a hotel, in a cafe, or anywhere while you're out. Sometimes your phone even connects automatically when the signal drops. But when you use public networks, the network owner can potentially see the data your device sends and receives, including the sites you visit. That's where Cyber GhostVPN comes in. With just one click, Cyber Ghost encrypts your internet traffic and routts it through secure servers in over 100 locations worldwide, helping protect your online activity and keeping your browsing private. This is why no one will know about your online activity. Cyber Ghost also lets you unlock georestricted content from more than 40 streaming platforms, including Netflix, Amazon Prime, and Disney Plus. By changing a virtual location, you can access content libraries that may not normally be available in your country. It's no surprise that Cyber Ghost has grown to over 38 million users worldwide. Personally, I like using it not only for privacy, but also to reduce tracking from websites and keep my online activity more secure. Cyber Ghost works across all major devices and operating systems, including Windows, Mac, Android, and iOS, and you can protect up to seven devices at the same time, which is great if you want to share it with your friends or family. If you're interested, check the link in my description or in the pinned comment to get 84% off plus four months free. And if you try it and you decide it's just not for you, Cyber Ghost offers a 45day money back guarantee. So there's no risk. Go and check it out using the link below. So I recently drew this diagram to show how big tech is passing around the same money in a circle to try and pretend that there is commercial value in building out these huge AI data centers. This of course is definitely not roundtpping because that would be illegal. No, it's just money moving in a roundtrip sort of way around all of

these companies where each company counts the same money as their own revenue. But one thing that stands out on that chart is that there is precisely one tech company missing and that big tech company is Apple. Elon Musk is building the Colossus data centers in Memphis. Amazon is building data centers in Indiana and Mississippi. Google and Oracle are converting the entire Texas panhandle, it seems, into one giant data center because, well, I guess there wasn't anything there anyway. Facebook is building data centers in Louisiana, in Ohio, and presumably somewhere in the metaverse. Microsoft is building data centers of their own and giving money to Open AI so that OpenAI can then give that money to Oracle and Coree to build their data centers. The same story about AI generating infinite growth and infinite money and universal high income, whatever that is, keep being told every year for the last four years already. Old people who have been in their jobs for a long time are becoming very concerned that AI is going to take those jobs. But then we have the reality. In the last few days, OpenAI announced a new model GPT 5. 5. Anthropic released Opus 4. 7. Twitter or X or XAI or SpaceX or whatever it's called this week released Grock 4. 3. Each of these models cost billions of dollars of compute and required many hundreds of thousands of graphics cards to run non-stop for months churning data. And yet the US economy growth rate is currently in the latest available data at just 2. 0%. Instead of collecting unimaginable amounts of money and distributing it out to the people, the US government is busy paying back tariffs to US companies that had to pay those tariffs for imported goods last year. As with every tech innovation, the expectations of the tech bros do not align with how the real world actually works. Instead of costs coming down over time, models are becoming more expensive to train. trying to get small improvements in performance now cost billions in training and two to five times more in inference costs. A few days ago, Enthropic announced that they now have a run rate of $30 billion in revenue. This means that if you multiply their revenue from the last month by 12, you get to 30 billion. This revenue apparently increased from just 9 billion at the end of 2025. So revenue for AI companies is growing insanely fast on paper. And in recent interviews, Dario Armade has said that Anthropic is already close to being profitable. Apparently, every model that Anthropic has released is profitable over the lifetime of that model. According to them, it's just while you're collecting revenue on the last generation model, at the tail end, you're also then spending model to train the new generation model, which costs a lot more than the previous one did. But that doesn't count towards the old models metrics. And the theory is that eventually Anthropic will get to a place and every other AI company where the revenue just keeps going up in the vertical line, but the development cost will stop increasing and then the company will start being profitable. And so there's all this talk of insane profitability just around the corner. But 2 months ago, Anthropic went and collected $30 billion in funding in their series G funding round. Series G, by the way, comes after A, B, C, D, E, F, and the seed funding round to start the business in the first place. It's extremely unusual for a company to be going through a series G funding round. It normally doesn't get that far. It's even more unusual to be doing it while expecting to do series H and the rest of the alphabet just a few months later. So, two months ago, Anthropic collected $30 billion in funding. This is a company that is apparently making over $30 billion a year in revenue. Now, recently they went and collected another $5 billion from Amazon with a deal to get 20 billion more in the future. And then Google gave Anthropic another $10 billion with a promise to pay $30 billion later. The obvious question is where is all of this money going? There aren't enough data centers in the world for Anthropic to burn through $30 billion in 2 months. Somebody somewhere is not being fully honest with the numbers because apparently in the whole of last year, Anthropic only lost apparently $4 billion. If you only lost $4 billion last year and your revenue and profitability are going up like no tomorrow and you're almost pretty much profitable, why do you need $95 billion in funding a few weeks ago and apparently another 50 billion in a few weeks time? The truth obviously is that the real cost of training and operating these models is much higher than what the companies would like to admit. There is a lot of creative accounting where the cost of training the models the main cost of the entire operation doesn't count towards profitability when these

companies do their adjusted earnings reports. It's a bit like a company selling wooden fences saying that the cost of buying the lumber should not be included in their profit and loss statement. And while every big tech company is tripping over each other to build the biggest and baddest data center, Apple has accidentally become a winner by doing absolutely nothing. Apple have struck a deal with Google to pay Google $1 billion. $1 billion per year for Gemini to power Siri. One huge data center with 500,000 Nvidia GPUs inside costs something like $20 billion to build. So Apple is paying the equivalent of about 5% of one of those data centers per year instead of building five of them data centers themselves. When the dotcom crash hit, the poster child of that collapse was Cisco. Back then the internet was the big new thing and Cisco was the biggest baddest company building the networking, all of the hardware and software to provide the internet to the masses. In March 2000, at the height of the dotcom bubble, Cisco became the most valuable company in the world, overtaking Microsoft because the market presumed that whoever is building the pipes is going to benefit the most from the internet age. It turned out the actual connection itself, the provision of internet as a service, that bit became a commodity. Nobody cares about the wires. The companies that benefited most from the internet were those that then built products and services on top of the infrastructure that became commoditized. Today, everyone is banking on large language models to be the thing that drives value in the future. Everyone except Apple. I know Apple also talks a lot about AI on their earning scores because that's the age we live in. You can't not mention AI 100 times. But watch what Apple is doing, not what they're saying. In March this year, Apple released the MacBook Neo. This is a radical departure from everything Apple has ever done in the last 30 years. This is an Apple laptop with all the same software that you can buy brand new for $599. Instead of being powered by one of the big expensive M series chips, this laptop is powered by an A18 Pro chip, the same processor that powered not this year's, but last year's iPhone. And here's the thing, that laptop is good enough to run any of the AI models via the same online web interface that you use on any other device, including your phone. But on that laptop, you can also start a business. You can use it to study. The educational version, by the way, of the laptop costs $100 less. And which specific LLM will people use on that laptop? Well, it doesn't really matter. It's about as important as which specific fiber optic cable connects your house to the internet. You don't know and nobody cares. Apple might have fumbled with the release of Apple intelligence, but on my iPhone, I have a folder which has every AI model that I use in there and I can use those whenever I want. If one of them starts being better than the one that I'm currently using, I'll go and switch. It takes me two seconds. I use all the main LLMs all the time. And when you use them a fair bit, you get to know which one is better for writing code, which one is maybe better for generating text, which one is better for generating some kind of ideas or knowing technical details about system architectures. I have four of those open in different tabs on my computer right now as I'm recording this video. But that's not a strength for AI. LLM models. That's a sign that LLM itself is a commodity. Three and a half years after Chat GPT, most people are still not prepared anything to use an LLM. Out of the small number of people who do pay, the vast majority are not prepared to pay more than $20 a month for a product that probably costs several hundred per month to provide. But while the AA companies are losing more money every year, the Wall Street Journal reported that Apple collected $900 million in 2025 from people paying for third-party AI subscriptions on their phones via the App Store. And Apple makes a 77% margin on their services business. We've just found out that Tim Cook is stepping down the CEO of Apple this year. and Apple have picked a man to replace him. And who do you think did they pick? Was it a software guy? Maybe, you know, the CFO. Maybe it was some kind of high-profile AI expert because that's what you have to be today, right? No, it was John Turnis, the guy in charge of hardware engineering, the guy responsible for engineering the MacBooks and the iPhones. That tells you everything about Apple strategy in that one decision. Apple screwed up on AI and accidentally missed out on incinerating hundreds of

billions of dollars on AI. Everybody else is so deep into burning money on AI, they can't stop doing it because it'll be a pretty bad look with the investors and might lead to the valuations crashing. The only solution is to keep going, keep burning the cash. You can't throw enough good money after bad when it comes to building gigantic data centers. Yes, Google, Facebook, and Amazon all have huge amounts of cash, and they do print more cash every quarter with their core business, but we're now hearing warnings from every big tech company that the capex spend in 2026 is going to mean reduced or negative free cash flow. Amazon just came out and said they'll probably have negative free cash flow. And at the same time, while everyone else is burning money, Apple now have the biggest amount of cash and marketable securities of any company in the world, except Warren Buffett's Berkshire Hathaway. Remember to go and get your special 84% off deal plus 4 months free from Cyber GhostVPN. Link below. Thank you for watching and I'll see you