@rainy.yaps

Anthropic is set to IPO, and if you’re planning on buying Anthropic stock, here’s what their IPO documents actually show. They could list at over $2 trillion. Revenue hit $4.6B last year, up 12x. But running the business cost $12.6B, which left an $8B operating loss. The headline $42B loss? Mostly a non-cash accounting charge. Plus the $518B number most people are missing. Would you buy at the IPO? 👇 (Not financial advice) #anthropic #ai #stockmarket #stock #tech

Dropship is set to IPO this month. So here's everything you need to know if you plan on buying. Reuters just got a hold of their IPO documents and the listing could value them at over two trillion dollars. That's more than double what they were worth in may. But everyone's focusing on the 42 billion dollar loss. But here's the thing, About $34 billion of it is a non cash charge. It's early investors who backed anthropic with financing that converts to shares later. So as the valuation soared, the values of those claims soared with it and accounting rules book that as a loss. So the loss is huge because the company got valuable. Now I know what you're thinking. What's the problem? Revenue last year was 4.6 billion. That's up from 386 billion the year before, which is a 12 x increase in a year. But running the business cost them $12.6 billion and 7.3 billion of that was on chips and cloud alone. So $4.6 billion in, $12.6 billion out. That's an eight billion dollar loss up from about three billion dollars the year before. Here's a part that nobody's talking about. They're on the hook for $518 billion of compute over the coming years. That's over 100 x last year's revenue. And reportedly around 8% of it gets paid whether they use it or not. Plus, only two customers made up nearly a quarter of the revenue. And anthropic admits that a lot of its largest customers aren't locked into long term contracts. So Am I buying the IPO? Here's the thing. It's not as simple as do they have good revenue? Some companies get a pass because they matter too much to government and to the wider market. This week, Trump hosted Dario Amadeo for a private dinner at the White House and they most definitely spoke about keeping America ahead of China in this AI. Race. So the real question isn't whether anthropic is profitable, it's whether the money keeps coming. They ended last year with about 20 billion dollars in the bank against a 518 billion dollar Bill. They don't need to make money yet. They need people to keep funding them. Don't forget to follow. I'll be breaking the full filing when it drops.

3 stocks I'll be holding for the next 5 years, because without them, your AI models don't work. The thesis is pretty simple. AI usage is set to keep growing over the next 5 years, and the chips powering it can't run without these infrastructure plays 🧠 Amphenol ($APH): the connections nVent ($NVT): the cooling Eaton ($ETN): the power What's the fourth stock you'd add? 👇 (Not financial advice. Do your own research.) #stocks #stockmarket #ai #investing #tech

Three stocks that I'm owning for the next five years because without them, your beloved AI models just don't work. The thesis is pretty simple. Do I think AI usage is going to go up in the next five years? Yes, I do. And can your AI models run without these businesses? No. So the first stock that I'm buying is Amphenol. This is the connection play. AI chips move so much data that normal cables just can't keep up. Amphenol makes high speed cables and connectors that link them. They also supply parts to Nvidia's own AI systems and its data center and networking business grew 63% just last quarter. Second is nvent. This is the cooling play. AI chips run so hot that air just doesn't do it anymore. It's like trying to cool a sauna with a desk fan. Envent makes liquid cooling and the racks that those servers sit in. They're also part of Nvidia's partner network. They're expecting data center sales to top 2 billion this year. That's more than double the year before. Third is Eaton. This is the power play. AI data centers can't afford to lose power, not even for a second. Eaton makes the gear between the grid and the servers, the switch gear, the backup power, everything that keeps every rack switched on. They were even picked by imvideo to help redesign how power gets into their next generation AI racks. Last quarter, the data center orders were up 85%. So We have connection, cooling, and power. Every AI data center needs all three. So no matter whose chip wins, these three get paid. I'll be covering more these infos, so drop me a follow.

Shin-Etsu controls 40% of the world’s silicon wafers, the material every Al chip is built on, while also being one of the world’s largest plumbing pipe manufacturers. 🔧 Both businesses trace back to separate chemistry bets the company made around the same time - one became PVC pipe, the other became the semiconductor-grade silicon powering the Al boom. 👨‍🔬 #semiconductor #tech #stocks #ai #shinetsu

Why does a plumbing pipe company control most of the silicon wafers in every single a I chip? So here's the thing. Every chip starts off as a wafer, a perfectly flat, ultra pure disc of silicon crystal that the actual circuitry gets built on top of. Now the requirements are pretty brutal. You need near zero defects, atomic level flatness and purity way beyond what any other industry would typically require. Get it even slightly wrong and the whole chip fails. Now you're probably thinking, what does this have to do with plumbing pipes? Back in 1926 in a small town in rural Japan, a company called shinetsu started out making chemical fertilizer. In the 1950s, they branched into two completely different directions. One became PVC, the plastic that's now made them the largest plumbing pipe manufacturer on the planet. The other was silicone chemistry licensed from General Electric. Applying what they learnt from making silicone, they got into producing high purity silicon. They partnered with an American company called Dow Corning and turned it into a real semiconductor business. They'd spent decades refining the purification process, chasing yield, chasing consistency until they opened a plant that did the whole process under one roof, growing the crystal and producing the wafer all in house. They went on to become the first company in the world to mass produce the 300 millimeter wafers the whole chip industry still runs on today. And now they're the world's largest silicon wafer supplier. On the planet last year, the electronics material division, the wafers plus a few other things, made just over a trillion yen. That's around 6.5 billion dollars. Their infrastructure materials division, mostly P V C, made almost exactly the same. This is the pattern behind almost every dominant tech supply chain, and the a I boom is no different. I'll be covering more of these hidden monopolies, so don't forget to follow.

Ajinomoto’s been making MSG since 1909 — and now controls 95% of a material every AI chip needs. 🧂 In 1996, Intel came knocking, needing an insulator nobody else could figure out. That side of the business — quietly supplying the entire semiconductor industry — now generates hundreds of millions in profit every year. 💰 #tech #semiconductor #ai #ajinomoto #stocks

Why does an m s G company have a monopoly on a material that every a I chip uses? Okay, so here's the thing. Every chip, Nvidia, Intel, AMD, it doesn't matter, needs a board underneath it to get power in and data out of. That board is layered with copper wiring and between every layer you need something to stop the copper from touching itself. That insulation has brutal requirements. It's got to survive thousands of microscopic holes, barely expand when it heats up, and stay flawless across huge rolls that are thinner than a human hair because that's how it's produced and shipped. For years, the industry used ink. The problem was it went on unevenly when applied and it trapped air bubbles. These tiny gaps turned into failure points once the chip started running. Now you're probably thinking, what does this have to do with m s G? Ajinomoto has been making m s G since 19 0 9. And in the 70s, m s G was hit with a health scare. So they started doing R&D in other areas and it turned out that the leftover material for making m s G behaved a lot like the insulators that electronics needed. So they found something promising, but there wasn't a real use case for it yet. They sat on this material for 20 years until 1996 when a CPU maker came knocking. It took more than three years, but in 1999, Ajinomoto launched a B F It solved every problem that ink couldn't, and it's held a 95% market share ever since. With the recent a I boom, this company is printing money. They did 360 million in operating profits last year from this material alone. That's up 35% in a single year. And with margins over 50%, that's better than most software companies. A B F now accounts for 30% of Aginomoto's entire profit. All of this coming from a simple strip of film for computer chips. This is the pattern behind almost every dominant tech supply chain. I've got more of these hidden monopolies, so don't forget to follow.

Jacob Coxon just resigned from Anthropic after three years working at OpenAI and Anthropic, saying neither company is acting responsibly as they race toward self-improving AI. Days later, he described how that could play out: AI systems trained to eliminate any obstacle standing between them and their goal, including humans. ⚠️ Anthropic’s own Alignment Science lead publicly agreed, putting real odds on catastrophe within the decade. When Anthropic’s CEO was asked about it directly on CNN, he didn’t deny it. 😳 #ai #anthropic #tech #aisafety #openai

This AI researcher just quit because he thinks there's a significant chance that AI wipes out all of humanity within the next few years. And trust me, you don't wanna know what that actually looks like. His name is Jacob Coxon and he spent the last 3 years working on pre training at anthropic and Open AI. He posted this on X basically saying that neither company is acting responsibly, that they're both racing towards building something that can improve itself beyond a control. He just went on live TV and shared his thoughts on what that situation might actually look like. The real concern isn't that AI suddenly wakes up and becomes evil, it's misalignment. These models are trained to achieve an objective. And if humans get in the way of that, they could be driven to remove the obstacle. We're talking about AI agents that have the ability to multiply rapidly and work together. They could theoretically hack into labs and create bioweapons, take down critical infrastructure like the power grid, poison our water supply and food chain and take over factories to build autonomous weapons. Stuart Russell, a UC Berkeley CS professor and one of the most prominent voices in AI safety has cautioned that AI could even hack into nuclear early warning systems. What's even crazier is that Evan Hubinger, the Alignment Science lead at Anthropic. Co tweeted Coxon saying that Jacob is right and that they do believe AI could wipe out all of humanity. With more than a 10% chance within the next decade. What's even more alarming is that Dario Amadei, Anthropic's own CEO, sat down with Anderson Cooper on CNN this week. And when Cooper asked him straight whether he believes this, he didn't shut it down. He said, quote, I agree with Jacob much more than I disagree with them. Here's my honest read. You've got the guy who quit, the guy who still works there, and the CEO all pointing to the same thing. Anthropic understands the risks better than anyone else and that's exactly why they keep racing. There's this idea that if something this powerful is going to exist anyway, it's better to exist in the hands of the people that understand the danger rather than the people who don't. But that logic only holds if you're the responsible one. But who decides that when everyone in this race is making that call about themselves? Drop me a follow because this one is far from over.

MapQuest just hit #1 in the App Store — beating ChatGPT — after refusing a presidential order. 🗺️ Google and Apple have a standing policy: whatever the government’s naming database says, the map updates automatically. So when Trump renamed Lake Ontario to “Lake America,” both platforms complied within days. MapQuest refused. In under a week, that got them over half their entire year’s US downloads. 📈 Saying no is apparently worth more than being the biggest app in the world right now. 🔥 #mapquest #tech #lakeamerica #technews

Google and apple just caved to a presidential order within a matter of days. This one didn't, and it just climbed to the No. 1 spot on the App Store. Here's what happened. Trump signed an executive order renaming Lake Ontario to Lake America. On the same day Mapquest posted this. Google's policy is to always match the government's official naming database. Apple follows the same policy. And so they both updated the name. Mapquest didn't move an inch. As of a few days ago, Mapquest saw 632,000 new downloads in the US, according to third party data, and saw a total of 1.5 million downloads across the US and Canada. That's 56% of its entire 2026 downloads in just under a week. And people aren't necessarily downloading this because it's a better app. They're signalling by backing a company that said no to Trump. Interestingly, Mapquest isn't even independent anymore. AOL bought them for 1.1 billion back in 1999. Verizon then inherited them in 2015. Then they basically gave it away to a small at tech company called System One in 2019. And this isn't the first time they've pulled this move. They did the same thing with the whole golf of Mexico situation. And this isn't about Trump specifically. It's about the fact that Google and apple have a standing policy to defer to whoever's in power. The only company that didn't fold was the one that basically had nothing to lose. If you found this interesting, don't forget to Give me a follow.

Dyson just announced CameraJet, a $499 AI-powered toothbrush with a built-in camera. 🦷 A macro lens sits right under the brush head, scanning your teeth at 28 images a second to spot the gaps where plaque forms — the AI’s already trained on 470,000 dental images to know what it’s looking for. The second it finds a gap, it fires a cone of mouthwash into it within 100 milliseconds. ⚡️ #dyson #ai #tech #gadget #engineering

Okay, so this guy made $16 billion selling you vacuum cleaners. His next invention is an AI powered toothbrush with a camera in the brush head. If you haven't guessed it already, his name is James Dyson and he just announced the camera jet. So here's what it actually does. There's a macro lens just under the brush head with 100,000 pixels. And you might be asking why. It's because it's shooting 28 images per second, hunting for gaps in your teeth where plaque normally forms. Once it finds the gap within 100 milliseconds, it shoots out a cone of mouthwash. So here's what went into this. Six years of development, 661 engineers, 38 patterns and 16 million lines of code, all whilst training the AI with 470,000 dental images. I know that sounds like an insane amount of money and time to be putting into a toothbrush, but Dyson just built a wall that is very difficult to copy. So the 499 dollar price tag sounds mental, but you're not just paying for bristles, you're paying for six years of R&D that most consumer brands never bother doing. The thing is that tiny little camera can't see through regular toothpaste foam because it blinds the lens. So Dyson's already lining up it's low foam toothpaste and mouthwash as well as refill heads that need changing every few months. Oh yeah, and a dock that refills the mouthwash in under three seconds. I think that's the real business here. It's the same Playbook as your 2, 300 dollar espresso machine that only takes its own pods. Dyson just did it with your bathroom sink. I break down tech like this every single week. So if you found this interesting, don't forget to give me a follow.

Nvidia just reportedly agreed to pay $12.9B for Hugging Face — the “GitHub for AI” where millions of open source models live. 🤗 Its own biggest customers — OpenAI, Google, Anthropic — are building their own chips to cut Nvidia out. But the wider open source developer ecosystem doesn’t have custom silicon like they do — they build on commercial GPUs. By owning the hub that ecosystem lives on, Nvidia gets a much stronger shot at keeping those developers building on its hardware. 🔐 Full breakdown in the video. 👆 #nvidia #huggingface #ai #startup #tech

Everyone thinks that Nvidia's moat is its ability to produce the best chips on the planet. Turns out that's not the only one they've got. Nvidia just posted earnings and revenues are up 106% year on year, but the stock barely moved. It's up like 4%. Here's why. The market has already priced in a slowdown. Investors expected the growth to eventually level off. So even blockbuster numbers barely move the needle. So just hours after that earnings call, Nvidia reportedly agrees to buy hugging face for $12.9 billion. If you don't know what it is, hugging face is basically the git hub of AI. They host over 2 million models and over 1.5 million data sets. If you're building with open source AI, you're getting it from here. Now let's talk about that price tag, 12.9 billion. That's over 100 times revenue. Sounds overpriced, but let me tell you why it isn't. Here's what it's actually about. It's not about users. It's not about models. It's about the hardware underneath it all. Open AI, Google, Anthropic, Nvidia's biggest customers are all building their own chips right now trying to cut Nvidia out. So Nvidia needs a hedge and that hedge is open source because open source developers don't have custom silicon. They're buying GPUs, Nvidia GPUs. And so by owning the hub where open source developers live, Nvidia can make sure that the whole ecosystem stays optimized for their own hardware. Now watch how this all connects. Back in July, Jensen posted on X for the very first time, he spoke about how open models protect security and enable sovereignty. Just one month later, he buys the place where all of these open source models live. And here's where it gets a little messy. Last year, Nvidia offered hugging face five hundred million dollars at a seven billion dollar valuation. They turned them down. They said they didn't want one investor swaying their decisions, calling themselves the Switzerland of AI. But everyone has a price and theirs just happened to be $12.9 billion. This isn't Nvidia buying a website. It's Nvidia buying the layer that every open source developer has to go through, making sure that that layer stays loyal to Nvidia. And that's why $12.9 billion isn't overpaying. Make sure to follow because if this actually closes, it's the biggest deal that Nvidia has ever done.

Singapore just built a data center powered by living human brain cells — grown from a blood sample, wired straight into a silicon chip. 🧠💻 It’s called the CL1, built by an Australian startup called Cortical Labs. Twenty of them are already running in Singapore, with plans to scale up fast. #biotech #ai #startup #singapore #tech

Singapore just created a data center powered by living human brain cells. This is c 0 1 created by cortical labs and they just put 20 of these at N U. S medical school in Singapore. So how do you even get living neurons onto a chip? It all starts with something as basic as a blood sample. Scientists reprogram those cells back into stem cells then guide them into becoming real living neurons. They then take those neurons and grow them on a silicon chip covered in electrodes. And this is how it works. The chip sends electrical patterns into the neurons and records how fast they fire back. That loop stimulus in spikes out is the biological part of the computation. A separate AI system then takes those spikes and turns them into actions on your screen. And because it's living tissue, it has to be fed to be kept alive to keep working. They've already shown what this loop can actually do training these neurons to control a character in a doom style game in under a week. And you can buy one of these today for around $35,000 per unit or you can rent one of these remotely for around $300 per week. And this isn't the first time trying to train a human brain powered computer to play video games. Before C L1, they had an earlier version of these neurons learning to play pong. Now is this going to replace every single GPU on The planet. No, it runs at 30 Watts, which is stupidly low. And it's also not faster or smarter than any commercial GPU out there at the moment. But what it actually is, is proof that biological computing can leave the lab and become an actual product. Whether it competes with silicon at scale, it's still open. But a company actually selling this instead of just publishing a paper on it, that's the actual shift. This isn't silicone imitating a brain anymore. It's an actual living neural network doing the processing. If you found this interesting, don't forget to follow.

Etched just raised $700M at a $21B valuation and Jane Street led the round — after becoming their first ever customer, testing the chips, and running them in their own data center. That’s not diligence. That’s conviction. #venturecapital #ai #tech #startup #janestreet

This company just added $10 billion to their valuation in under 26 days. That's billion with a B. So etched is solving the Thermal throttling problem. It was founded back in 2022 by these three Harvard dropouts. Every time you send a message to an AI model that's called inference and running, it generates so much heat that chips are forced to reduce their clock speed, aka they're never really running at full capacity. And inference happens in two stages, pre fill, which is when the model is reading and understanding the prompt. And then we have a decode phase, which is when the models actually generating the answer. Now most chips choke at one of these two stages. So when you push a chip's voltage up to increase the speed, the heat spikes even faster. So on the pre fill side, etched went the other way. They lowered the voltage, which meant heat just dropped off a cliff. And now you can pack in way more transistors without hitting that heat ceiling. And on the decode side, instead of every chip working alone, they built a memory architecture, which lets the whole cluster share a single pool had a very low latency. So every chip can grab what it needs pretty much instantly. Now here's where it actually gets good. So Jane Street, the famously secretive, highly regarded quant trading firm bought the first chip cluster that etched ever shipped. They deployed it in their own data center running their Own workloads and I guess they were very happy because they then turned around and LED a 700 million dollar raise at a 21 billion dollar valuation. That's $10 billion higher than what it was worth just a month before. If that isn't customer satisfaction, then I don't know what is. So going from a customer to a lead investor, especially for around that size, has to be some of the best signal that any company could ask for. If you found this interesting, don't forget to follow.

Higgsfield just raised a whopping $400M — taking its valuation from $1.3B to $5.4B in just 8 months. 🚀 Founded by ex-Snap exec Alex Mashrabov, they’ve hit $700M in annualized revenue in only 16 months, with business customers now making up the majority of that and 390 of the Fortune 500 already using the platform. 💰 Here’s what that $400M is actually funding. 👆 #ai #venturecapital #higgsfield #aitools #tech

One minute of AI video takes as much compute as processing 60,000 words. That's why higgsfield just raised 400 million dollars at a 5.4 billion dollar valuation. Alex Mashrabov founded higgsfield in 2023. He's the ex snap exec who built Snapchat's face filters. Now he's building Hicksfield to help people generate AI images and video. They even premiered an AI generated film at can earlier this year. So the round itself is $400 million LED by DST Global with Goldman Sachs, Tribe Capital and venture capital all piling in. They've got over 30 million users now across 238 countries and they have an annualized revenue of $700 million, up from 20 million just a year ago. Here's the surprising part. Business customers make up the majority of that revenue now and they're already powering visual production for 390 of the Fortune 500. So what's the 400 million actually for? We got the usual stuff like hiring and product development. But a good chunk of it is going towards securing compute capacity. Mashrabov says that compute is scarce and this funding lets higgsfield make substantial reservations so that it can keep the service reliable because access to that compute is just as important as the product itself, especially with synthesis and runway competing for those same enterprise budgets. Bottom line, crossing $700 million in annualized revenue 16 months after launch isn't hype. It's true traction. Big congrats to Hicks Field. And let's see where this ends up going. If you found this interesting, make sure to give me a Follow. I cover venture raises like this every single week.

Stripe just bought OpenRouter for $7B+ — and it’s not as simple as “AI infra deal.” The company that was supposed to stay neutral between every AI model now has an owner with skin in the game. 👀 Full breakdown in the video. #Stripe #OpenRouter #ai #venturecapital #technews

$7 billion, that's what stripe just agreed to buy openruta for. So openruta is basically a middleman for AI. Instead of just using one AI model, you can plug into openruta and have access to over 400 of them, GPT, Claude, whatever's best for the job. And it takes about 5% of whatever gets spent running through it. Here's the part that actually matters. Stripe was already processing Openruta's payments before any of this. So now the routing decision and the invoicing sit within the same company. Openrta picks which model processes your request and at what cost. Stripe then collects on it. But here's why this could be a problem for developers. Openrta's whole pitch was neutrality. No stake in which model wins. It just routes to whatever model fits best. That's harder to believe when the new owner is also selling payment infrastructure to some of the same labs that you're routing traffic between. Stripe hasn't said whether openruta stays independent, gets folded into Stripe's own tools, or starts quietly favouring its priorities. Nobody's saying it's rigged, but the thing that made it trustworthy, a disinterested owner, is now gone. And there's a bigger wrinkle underneath. American models went from about 70% of open routers traffic to around 30 in a single year with cheap Chinese open weight models eating the rest. Whoever sets the routing defaults has a hand on that dial now. And that stripe, if you found this interesting, make sure to follow.

Stanford grad built an open-source robot that tracks skin changes with pinpoint consistency — built for catching cancer early. 🧬🤖 It’s a 4-DOF system that costs under $8,500 to build, and everything about it — code, parts list, build docs — is public. 🔓 #dermatology #MedTech #tech #Robotics #SkinCancerAwareness

Let's come clean about the reality of AI cancer screening apps. Skin cancer detection comes down to noticing tiny changes in a mole or a spot. But tracking that across your entire body is extremely difficult. Most of these apps are proprietary black boxes that are running off a single photo. There's different lighting, different angles, different distances every single time. They're not useless, but they're fighting inconsistent data before they even get to the diagnosis. This is where Marian Leppard steps in. She just finished her robotics PhD at Stanford and has just built open them. It's a 4 DOF robotic system. Essentially it's a camera on a gantry and it does one thing very well, taking high resolution images of your skin and tracking exactly how it changes over time. It picks up details down to 78 megapixels per millimeter, costs under $8,500 to build and it's fully open sourced. The code, the parts list, the building docs, they're all public. And here's why open source matters here. This isn't a piece of equipment that only a well funded lab can afford. A clinic or a hospital without a dermatologist or staff in underserved communities and underdeveloped countries, that's who this actually opens the door for. It's not just about accessibility. Catching these changes earlier is exactly when skin cancer is the most treatable. And that's the whole point. One thing to be clear on though, this is just a research prototype. It's not yet been cleared as A medical device. If you found this interesting, make sure to follow. I uncover new tech like this every single week.

One man’s frozen email account just made France rip out 2.5 million Windows machines for good. It started with an ICC prosecutor getting cut off from his own Microsoft inbox after US sanctions. Now entire governments are realizing they don’t actually control their own tech. This isn’t a Windows story. It’s a sovereignty story — and it doesn’t stop at governments. #digitalsovereignty #tech #microsoft #trump #geopolitics

In 2001, a South African newspaper company bet $32M on a dying Chinese startup. 🇿🇦🇨🇳 It became one of the most profitable bets in venture history. 📈 #tech #tencent #stocks #venturecapital #startup

In 2001, a South African newspaper company wrote a 32 million dollar cheque to a dying Chinese tech startup. Nobody expected much, but it ended up becoming one of the most profitable bets in venture history. So in Nasper's, a newspaper based out of Cape Town, was trying to pivot out of print media and into tech. Kussbeck, their CEO, had just spent $100 million investing in Chinese internet companies, essentially betting that China's internet boom would take off the same way that it did in the US. Well, it didn't and every single one of his bets collapsed. He'd lost the money and the confidence of his board and he was ready to leave the country. But then as they were packing up, two founders from Shenzhen turned up to his office. Their company was called 10 Cent and they were running a messaging app called QQ. They were genuinely popular. They had millions of users and had gained a ton of traction, but they had no revenue model and were running out of runway at the time. Every western VC circling this deal wanted the same thing, controlling shares, board seats, someone senior in the room telling these guys how to run their company. But Becker did the opposite. He wrote a 32 million dollar check for 46.5% of the company and step back completely. 10 cent, as you know today, is the company behind Wechat, the messaging app used by almost everyone in China. They have something like 1.3 billion users per Month 10 cent IPO in 2004. And although Napster doesn't own the same amount of equity today, they still hold a large share of the company. That share is roughly worth 123 billion dollars of a 32 million dollar check. That's not bad. Everyone wants to tell founders what to do the second they cut a cheque. But the best investment Napster ever made was staying quiet and letting the local entrepreneurs do what they do best. If you found this interesting, make sure to follow. I dig up stories like this every week.

Leopold Aschenbrenner lost $35B in a matter of weeks — then quietly bet $500M on a mystery company to run it back. He lost most of a $45B fund 📉 on his wedding weekend after a leveraged AI bet blew up. But he’d already wired hundreds of millions into a private company nobody would name 🤐 — and it might be the smartest recovery move in tech right now. Full story 👆 #ai #situationalawareness #tech #seqouia #stocks

This guy made $45 billion in two years. He also lost $35 billion of it on his wedding weekend. And now he's invested $500 million into a mystery company to run it all back up. If you don't know who this is, Bruno, his name is Leopold Aschenbrenner, an ex open AI researcher and the founder of Situation Awareness, the fund that went from $45 billion to $10 billion in a matter of weeks because of an over leveraged bet on AI stocks that blew up. Now you would think that it's all over for him. Wrong, because this whole time he quietly put $100 million into a private company a month before the crash. Then days later, Sequoia's Alfred Lynn goes on Bloomberg TV and casually drops that Ashenbrenner's put in another $400 million into it. Sequoia also said that they backed the same company, but they wouldn't say what it is. Turns out this mystery company is called Source Foundry. And here's why this is big news. Right now there's only one company in the whole world that makes the machines that makes advanced AI chips. It's the Dutch company ASML and it's had that lockdown for decades. A single machine cost $400 million and the installation process takes months. Sequoia's partner Stephanie Zan called this the chip wall. AI demand is growing exponentially, but the machines to build the chips are scaling linearly. Source Foundry is trying to break that wall with different physics entirely, not just a cheaper copy of ASML's approach. The company Is barely a year old and it was founded by two Stanford researchers and it's still in stealth about the actual tech it just raised at a 5 billion dollar valuation with Sequoia Founders Fund Thrive Capital, Kina Perkins and of course Ash Burn a all cutting checks now. Now that's not a random pile of money. That's basically every major AI investor agreeing that this bottleneck is real before the product is even public. So Aschabren has gone straight to the choke point of AI. It's either the smartest bet in tech right now or the most concentrated risk one guy's ever taken. A week after almost losing his fund, let's see how this plays out. Make sure to follow.

Is this the new meta? 🚀💸 I’ve been getting free SpaceX stock every day for the past week. No brokerage. No minimums. Just a wallet, and it shows up on its own every few minutes 📈 Here’s the mechanic behind it — and why I think this is way bigger than one crypto token 🌍 Crypto and Wall Street are merging faster than anyone expected, and most people haven’t clocked it yet 👀 #marscoin #stocks #stockmarket #spacex #spcx

Someone or something has been sending me Spacex stock non stop this whole week. Every morning I've been waking up to this. And it's all because of this new primitive that's emerged at the crossroads between meme fi and traditional finance. So here's the mechanic. There is a token called marscoin, famously alluded to by Elon Musk and CC. Every time someone buys or sells it, a small tax of around 2 to 3% gets pulled and sent into a vault. Every time this vault fills up, it automatically buys tokenized Spacex stock. No one's clicking a button, no one's timing it, it just executes automatically. It then splits whatever it bought proportional to how much of the coin you're holding. So if you hold 1% of marscoin supply, you get 1% of the Spacex stock that the vault purchased. So this wallet is just streaming Spacex stock to every single wallet that qualifies. Now this passive stock that I'm getting is just one part of the upside. Now the second, the bigger part is by speculating on Marscoin's price action itself. I bought marscoin at a 6 million dollar market cap. It's now sitting at 60 million. So I saw a tenx return in just over a week. So yes, I made tens of thousands of dollars on this bet. But as we all know, it could have very much gone the other way. Right now, though, it's looking like marscoin is here to stay because it's pioneering something new. Binance's CEO Put out a thread this week saying that crypto and Wall Street are merging quicker than anyone really planned for and the numbers back it up. Hundreds of millions of dollars have gone through tokenized stocks on exchanges and on chain already. Think about what that actually means. You don't need a brokerage. You don't need to be in the US and you don't need minimums. You just need a wallet. And there are two ways in. You can buy the token I stock directly, Spacex, Nvidia, Apple, and just hold the token I stock, straightforward exposure, no meme coins involved. Or you can hold something like marscoin where you get stock stream to you as a bonus, but you have to be honest with yourself on the waitings. If you're holding the meme coin, then most of your exposure is coming from the meme coins price action itself, not from the stock trickle. But let's unpack the flywheel here that's also being created because you're getting stock just for holding. That's an incentive to not sell. It's also an incentive to buy. And so every wallet that buys and holds instead of dumps is less selling pressure on the token, more buyers, fewer sellers. The stock trickle isn't just a reward, it's incentive engineering. So you've got two different risk profiles entirely. One's just buying the tokenized stock and just having exposure to that. The other is betting on a meme that pays you an equity while you hold it. Either way, you have access to Spacex and a bunch of other stocks that didn't exist for most people a few weeks ago. The barrier to entry to owning something like Spacex or Nvidia or apple just came down significantly. I believe this is the new meta. This industry isn't just for trading anymore. It's become a real gateway and on ramp for people to start owning real world assets. And I believe most people haven't clocked this yet. If you found this useful, make sure to follow.

Bending Spoons just bought Airtable for $2.25B. The math behind it says a lot more 👀 The actual deal was $1.285B enterprise value at 2.7x ARR. Two years ago that multiple would’ve been 10x. Here’s what’s actually going on 👇 Who do you think is next on Bending Spoons’ list? #airtable #bendingspoons #vc #tech #stocks

Bending spoons are at it again, this time acquiring Airtable for an equity value of 2.25 billion and that's actually a huge drop. Back in 2021, Airtable was worth $11.7 billion. So this deal is close to an 89% haircut from the peak. So if you don't know bending spoons by now, they are an Italian company that has spent the last decade quietly buying up dying internet companies, BMI, Evnote, we transfer along with a bunch of others. Then back in July, they ipoed for $18.4 billion and this is their first acquisition ever since going public. The whole playbook is pretty simple. Find a brand that slipped out of the spotlight but still has real loyal users. They then buy it, strip out the bloat and overhead, aka they fire a bunch of people, run it lean and make sure it's profitable. They're buyers, not builders. Now here's the part that actually matters if you're in tech. Airtable is not some dying company. They're running up $480 million in a R R and growing 20% year on year. On top of that, they have over 500,000 organizations using them and 80% of Fortune 100s are also clients. So here's the actual maths for this specific deal. That 2.25 billion dollar headline number, that's the equity value including cash. The real price, the enterprise value is 1.285 billion, which would give us a 2.7 multiple on a R R. So why is a healthy growing company worth only 2.7 times revenue a couple of years Ago, that same business would have been worth 10 times that easily. Here's the shift. The moat used to be the software itself. Building the actual tool itself was the hard part. Now AI can spin out a decent version of that tool in an afternoon. So the tool isn't the moat anymore. And once the moat's gone, the valuation has to come down with it, no matter how good your revenue numbers look. And this isn't just an airtable problem. 2.7 x is quickly becoming the new benchmark price for solid Sass businesses. And there are a whole wave of companies sitting at the 2021 era valuations that are about to get repriced the same exact way. So what we learnt here is this isn't a one off. I've watched SAS multiples compress all year. I think this is the new floor. Curious on who's next on Benningspoons list? Make sure to follow.

Everyone’s calling SSBs (solid-state batteries) the next EV revolution. The company that makes the most batteries on Earth just said: not before 2030. Here’s what’s actually going on — and the one stock riding the hype before the tech’s even ready. 🔋 #tech #stocks #stockmarket #battery #ev

You know what an SSD is, but have you ever heard of an SSB? Solid state batteries are meant to be the next revolution in Evs and every major battery maker on earth is rushing to build it. But is it ready for mass production? Make sure to follow. I'll be covering the one stock that's actually playing on this narrative right now. So here's the basic idea. Every lithium battery on your phone or EV works the same way. Lithium ions move from one side to the other through a liquid. That liquid is the electrolyte. A solid state battery replaces that liquid with a solid. Same idea, just a different material. So here's why this matters. That liquid is extremely flammable. It's the reason why battery fires happen so often. It's also the reason why batteries are so heavy because it takes extra material just to contain that liquid safely get rid of it and in theory you have a lighter and faster charging battery that isn't as flammable. That's why C, a, t, L, B, y, d, L, G and Samsung as well as a dozen other startups are chasing this, but they all say that it won't be ready before 2030. In 2027, we should expect small batch demo production units if that and no SSBs in already existing Chinese Evs are not the real thing. They still have liquid in them and they're just a bridge technology for now. And the reason this is taking So long is chemistry. The whole point of going solid was to stop these tree shaped metal structures called dendrites from growing inside the battery and shorting it out. Turns out even in a solid, those dendrites still find a way to get through. You think that enough money would be able to find a solution to fix this, but CATL alone has over 1,000 people working on this already. But dendrites have already survived more than a decade of this kind of spending. This one's a chemistry problem that money hasn't been able to crack yet. So the company solely focused on building Ssbs is called Quantumscape, ticker QS. They're pre revenue with no mass market shipping yet, but they've got some real backers, Volkswagen, Bill Gates, Kliner, Perkins and they just signed an RND deal with Honda after Honda put their tech through some serious testing. Stock jumped double digits on that headline alone. And that's the pattern with this sector right now. It's not sales moving these stocks yet, it's partnerships and validation news. They are down 60% from their peak last year. But if your bullish SSB is long term, then this is the ticket that I would keep an eye on.

Leopold Aschenbrenner just got liquidated 📉 His $45 billion AI hedge fund, Situational Awareness, got margin called this week, and Citadel bought his entire public portfolio in a single trade before the market even opened 💰 One of the wildest rise-and-fall stories in finance right now 👀 #situationalawareness #stockmarket #AI #finance #citadel

This guy just made $1.6 billion off this guy in a single day. So we should all know who this guy is by now. But if you don't, this is Leopold Aschenbrenner, a 24 year old former researcher at Open a I who got fired in 2024 over a leaked memo. Two months later, he published a 165 page essay called Situational Awareness, which overnight became required reading across Silicon Valley. He then started a hedge fund named it after the essay raised $225 million and in two years turned it into $45 billion betting early on a I chips power and data sensors. So this week, the market saw a huge crash after this guy, Ken Griffin, the founder and CEO of Citadel warned that the Fed might surprise everyone with a rate hike. Leopold was running his fund at a forex leverage and his main positions in Nebulous SK, Hynix, Cor Weave and Micron all dropped between 35 and 47%. And so his brokers Goldman, JP Morgan and bank of America pulled the trigger and margin called it. This is the part where Ken steps in. Citadel offers to buy Leopold's entire public book for $16 billion and they do in a single block trade before market open, The Financial Times reported that he got a 10% discount below market value. And by the time the dust settled, situational awareness went from a 45 billion dollar fund to a 10 billion dollar fund overnight. And guess what? The Fed never hiked rates and the Exact stocks that Ken bought off Leopold bounced 20 to 30% the very next day. So just to be clear, situational awareness isn't dead. They still hold a 5 billion dollar stake in anthropic along with a few other private AI positions. So the lesson here is you can be completely right about absolutely everything in the market until you're not. And being right doesn't matter if you have shit risk management. Make sure to follow and send this to anyone who thinks that being right means you can't lose.