I've Bought Every AI Correction Since 2018. Here's My List

[01:00:00:00 - 01:02:18:14] I seem to keep getting the same question right   now, and that is, is the market too expensive to  invest in right now? Let's just imagine that you   were the worst investor alive. Every single month  from May of 2024 through November of 2025, let's   say that you bought $1000 worth of Palantir at the  absolute highest price each month, at the absolute   highest peak, and the worst possible price that  you could have paid for that month. No choice of   cherry picking in the dips. Instead, you're just  buying the top 19 months in a row, making your   investment at $19,000. That horrible investment  style would have been worth roughly $43,000 today,   or a 126% return, on the absolute worst possible  timing strategy in the world. Over that same time   period, the market would have only given you a  21% return. Palantir, even though it was bought   at the absolute worst price every single month, it  returned six times what the S&P did. Now recently,   Palantir released Q1 earnings where revenue grew  84.7%, beating expectation on every single line.   But the stock dropped 6% the next day. Does  that even seem to make sense? What company can   beat every expectation in front of it and then  drop the next day? So then it makes you wonder,   has it reached its peak? And to answer that,  you have to zoom out. I mean, way out. Here's   the bigger picture. Last year, in the first half  of 2025, US GDP grew at 0.1% annual rate once you   strip out AI data center spending. That means  that data center investments drove 92% of all   countries' growth in the first half. That means  this is an AI economy. Which means that the real   question for investors isn't whether you're paying  too much for these names. It's whether sitting out   of this growth cycle ends up being one of your  biggest regrets. And the reason that I've been   able to see a lot of these trends coming from  way out is because this is exactly what I used   to do for a living. Before I retired, I was an  executive at companies like Amazon and Target,   where I managed categories like men's apparel  and toys. My team and I had to find the trends   a year in advance. And that skill set translates  straight into my investing strategies to always   look ahead. So today I'm going to walk you through  how I'm investing in this growth cycle right now   to set myself up for the next several years. And  here's a perfect example about trends where the   news dropped just a few weeks ago that Palantir  is bidding on the air traffic control program. And   here's a video clip of mine from a year ago. [01:02:18:14 - 01:02:38:03]   Now I am going to drop a speculation bomb  on this video. But I find it intriguing   that Palantir recently partnered with Archer  Aviation. What if, and I realize it's a big if,   Palantir uses this as a test bed for something  much bigger. What if this is just an opportunity   to upgrade the US air traffic control  systems with their foundry capabilities?   [01:02:38:03 - 01:06:06:12] 12 months later, almost to the day,   the FAA shortlisted Palantir to build the AI  tool that's going to modernize US air traffic.   And it's a $32 billion program. My off-the-cuff  speculation became the shortlist. And that's my   trend forecasting in action on the exact type of  stocks that we're going to be digging into today.   For those new to the name, Palantir builds  software that takes massive amounts of data   and turns it into real-time decisions. They  have foundry on the commercial side and Gotham   for Defense and Intelligence. I first covered them  back in 2023 at $15 a share. And I've kept coming   back ever since. Two years ago, US commercial  revenue was $150 million in a quarter. Today   it just printed $595 million. That's growing 133%  year over year. A four-fold increase in two years,   with commercial now growing 1.6 times faster  than the government segment. And that's the   accelerator that you're going to need to pay  attention to. And here's the margin machine.   Since Palantir is inside an enterprise, ripping  it out is essentially impossible. Foundry takes   months to deploy across financial systems, supply  chain, manufacturing, and government databases.   Once it's running, customers renew and they buy  more modules. The technical term in software is   net dollar retention. In plain English, that  means existing customers buy about 50% more   from them every year. In November 2025, the stock  pulled back about 30%. But over the same window,   operating margins expanded over 1500 basis points.  And the commercial customers count grew 31%. So   let me get this straight. The price came down,  the fundamentals kept compounding, and the gap   closes a little bit more every single quarter.  Eventually, one of two things is going to happen.   The price catches back up to the fundamentals,  or the next earnings print makes the math obvious   enough that institutional money has to chase it.  Regardless, both outcomes are going to end the   same way. In a nutshell, that's Palantir. Now  let me walk you through six other names that   are sitting at this same setup across this AI  economy. We'll start off with Applied Digital,   a specialty data center developer purpose-built  for AI workloads. Most data center companies   aren't diversified across cloud, enterprise,  and AI. Applied Digital builds just one thing.   AI factory campuses engineered for the densest,  most power-hungry compute in the world. In April,   they announced their biggest lease yet, a 15-year  deal with an investment-grade hyperscaler worth   $7.5 billion. It pushed their total contracted  lease backlog past $23 billion. They also just   spun off their cloud computing business as a  separate company called Chronoscale. Applied   Digital still owns about 97% of it, and the split  lets the AI cloud business and the AI real estate   business scale on their own terms with different  capital structures. So Chronoscale might be worth   putting on your watch list. Getting back to  Applied Digital, the primary tenet has been   CoreWeave at Polaris Forge One in North Dakota,  and the April lease adds a second investment-grade   hyperscaler at a new campus. This campus design  is liquid-cooled with ultra-high-density compute   racks built next to standard natural gas where  power is cheap and the grid has a lot of capacity.   The push is to scale this AI factory model from  600 megawatts of contracted capacity to over   1,000 megawatts in the pipeline. Applied Digital's  market cap sits at roughly $13 billion, but their   contracted lease backlog tops $23 billion. The  company is priced at less than half the value of   the contracts that are already signed. [01:06:06:12 - 01:07:38:16]   Every company tied to AI infrastructure, the data  centers, the power plants, the chip fabs, they   all depend on precious metals. And that brings us  to today's sponsor, Canadian Goldfields Discovery   Corp., a pure-play gold exploration company  in northwestern Ontario's Miminiska Fort Hope   Greenstone Belt. This region has produced over  200 million ounces of historic gold from Red Lake,   Timmins, and Hemlo. The Miminiska Gold Project sat  through two different gold cycles untouched, while   every other neighbor got developed. Mineralization  is confirmed at both ends of a 14-kilometer   system. The gold sits in Banded Iron Formation,  or BIF, ancient layered rock where gold runs in   continuous, predictable bands instead of scattered  veins. This BIF runs through the 12-kilometer   corridor between those zones, never drilled with  modern systematic methods. BIF is one of the most   reliable gold hosting formations worldwide, same  setting as Muscle White at 6 million ounces and   Homestake at 44. With no debt, 4.5 million  in cash, permits in hand, and a 10,000-meter   drill program set to launch. Canadian Goldfields  Discovery is ready for business. The team brings   pedigree from K92 Mining, Ivanhoe Mines, and BHP.  And really, ownership tells the rest. Roughly 30%   with board and management, 30% with institutions.  That's 60% of the float is in strong hands. So   let's see, they have a pure gold play, a tier  1 jurisdiction, a BIF system with 12 untested   kilometers, and 10,000 meters of drilling starting  right now. That's the setup that early stage   investors really look for. If you'd like to learn  more about Canadian Goldfields Discovery, please   check out the link down in the description. [01:07:38:16 - 01:08:29:18]   Next up is Credo Technology, a semiconductor  company specializing in moving data inside AI   data centers. This semiconductor news focuses  on the chips themselves, like NVIDIA's GPUs or   AMD's processors. But Credo doesn't make those.  They make the connections between the chips.   Active electrical cables, optical digital signal  processors, and the high-speed interconnect that   lets thousands of GPUs in a data center talk to  each other fast enough to actually work. In April,   they announced a $750 million acquisition of a  silicon photonics company called Dust Photonics.   The deal vertically integrates Credo's high-speed  signaling expertise with silicon photonics,   and management is now guiding optical revenue  alone to exceed $500 million in fiscal 2027.   Their customer base is the hyperscalers,  Microsoft, Amazon, Google, and others.   [01:08:30:18 - 01:16:45:10] Microsoft used to be 86% of revenue. That number   is down to 42% as the rest of the hyperscaler base  ramped up, which is meaningful diversification   of customer concentration. That's good. The  technology is power efficiency. Credo's optical   signal processors run at 30 to 50% lower power  than competitors at the same speed, which matters   when you have hundreds of thousands of GPUs in a  single building all needing to talk to each other.   The push is the Dust Photonics integration, the  move into co-packaged optics and ramping their 1.6   terabit optical platform, the highest bandwidth  interconnect commercially shipping today.   The revenue grew 201% year over year with a  peg ratio of 0.9. That's the highest growth   rate of this list at the lowest growth adjusted  price. Now we move on to Amphenol Corporation,   the picks and shovels layer of the AI build-out.  Amphenol doesn't make chips or run the cloud. They   make the physical connectors and the high-speed  cables that link everything together inside a   data center. With the same products serving  autos, aerospace, and mobile, but the AI side   of the business is really what's accelerating.  In April, they reported record Q1 results, with   the IT data comp segment generating virtually all  its organic growth from AI applications. Orders   hit a record $9.4 billion, up 78% year over year.  The book to bill of $1.24 means orders are coming   in faster than they can ship. In January, they  closed a $10 billion acquisition of Comscope's   connectivity and cable solutions business,  which gives Amphenol coverage across the full   data center signal path. High-speed copper, power,  active copper, passive fiber, and active optics.   They're no longer just connecting chips, they're  connecting everything that moves data inside   of an AI factory. The technology is custom  engineering at scale, which is why they keep   winning sole source contracts on these next-gen  GPU racks for specific hyperscalers. For them,   the push is owning the entire data center signal  path and capturing the growth in AI-driven cabling   density. A typical GPU rack now uses two to three  times the interconnect length of a 2024 rack,   and that ratio just keeps climbing. And now we'll  move on to the next company on the list, which is   Meta Platforms, the parent company of Facebook,  Instagram, WhatsApp, and Threads. For investors,   what matters is they're the second largest  digital advertising business in the world,   sitting in front of roughly 3 billion daily users.  Meta is on the list because the stock is down 23%   from its recent high of $796 back in March. While  the price was getting cut, the business was doing   the opposite. Revenue grew 33% year over year, and  operating margins expanded over 800 basis points   in that same window. The technology behind it is  the AI-powered ad targeting layer that they had to   completely rebuild after Apple's privacy changes  completely broke their original tracking model.   But they figured out how to predict ad relevance  using their own first-party data and on-device   machine learning, which is why advertising  revenue has accelerated, not decelerated. Meta   had pretty much cut the Reality Lab's budget  and delayed the next mixed reality headset,   pivoting hard into those Ray-Ban AI glasses. And  that bet sold more than 7 million frames in 2025.   AI glasses are the actual wedge into the next  consumer device category, and they're betting   hard on it. Other size, Meta's peg ratio sits  at 0.93. So for context, at the end of the day,   the stock pulled back 23% while every fundamental  is just getting better. Next on the list is ARM   Holdings, the company whose chip architecture  powers virtually every smartphone in the world.   As a reminder, ARM doesn't manufacture any chips.  They license the design to companies like Apple,   Qualcomm, Samsung, and Nvidia, who pay ARM a  royalty on every chip that's shipped. And that's   exactly why their gross margins look more like a  software company than a semiconductor. Well, quite   honestly, their margins are probably better than  anybody's. In March, ARM announced something that   breaks their 35-year-old business model. They're  going to sell chips directly to customers for the   very first time. The product is called ARM AGI  CPU, and Meta and OpenAI are committed customers,   with production expected by the end of 2026. To  me, the bigger story is underneath this in the   data centers. Because we have AWS Graviton, Google  Axion, Microsoft Cobalt, and Nvidia Grace all   run on the ARM's Neoverse architecture. ARM has  effectively become the standard for hyperscaler   compute. The best part of their technology  is power efficiency. ARM designed chips run   cooler and use less electricity than competing  architectures, which matters as much in your   phone as it does in a data center running hundreds  of thousands of AI accelerators. The main drive   for them is to move into AI-specific compute. ARM  split the company into three new business units.   Edge, PhysicalAI, and CloudAI, and it's launching  their first proprietary chip alongside the legacy   royalty business. They're going from licensing the  design to actually selling the product. And ARM's   data center royalties just doubled year over year  with a 97% growth margin. I mean, they're the true   definition of printing money because it's just a  royalty business and they're growing just as fast   as everyone else. And the next name on the list is  Terradine, a semiconductor equipment company built   directly into the AI chip production line. Once  again, now Terradine doesn't make chips. They make   the test platforms that every chip has to pass  through before it ships. They make the equipment   that determines whether the Nvidia GPUs or the  AMD CPUs actually work before they ship them.   In April, Terradine reported Q1 results that beat  expectations on every metric. earnings per share   grew 241% year over year with AI-related demand  making up nearly 70% of their total revenue. And   their reward for this? The stock fell 19% on the  news. It's the same pattern as Palantir, Meta,   and the rest of them that are on this list. But  of course, since then, the stocks have gone up and   down multiple times. As for their customer base?  They happen to be the entire AI ecosystem. TSMC,   Samsung, Nvidia, AMD, all of the hyperscalers  and all of the memory companies. And their CEO   describes the demand in three different stacking  AI waves. They've got general purpose data center   build out, inference optimized silicon, and edge  AI. And they just so happen to be positioned   for all three of those. The technology is test  platforms specifically designed for high bandwidth   memory, the kind of memory that sits right next to  every AI accelerator. Their Magnum memory tester   gives them a lead in HBM testing that competitors  are still trying to catch up to. Their drive is to   expand into AI data center connectivity and test  software. They just invested $157 million in a   joint venture called Multilane Test Products and  acquired a software company called Test Insight,   adding capabilities for testing AI infrastructure  from end to end. So that's seven names across   different layers of the AI economy. You've  got software, real estate, semiconductors,   connectors, chip design, and test equipment. They  all happen to be completely different businesses,   but they all work towards the same setup. Each  one beat expectations on their last earnings   report. Each one has a multi-year strategic growth  sitting right in front of it. And on most of them,   the stock either dropped after the beat or pulled  back from a recent high, even while the underlying   fundamentals just kept improving. And that's the  gap that I keep talking about. The gap between   what the market sees in the moment and what the  math says about the next five to ten years. If you   go back to the question that we started with, is  the market too expensive to invest in right now?   My answer is the same answer that I gave  you 12 months before the FAA shortlisted   Palantir for air traffic control. Look at what  these companies are going to own in five years,   not necessarily what all the pricing action  is doing today. The trend forecasting lens   is what earned its keep when I bought into  Palantir two and a half years ago because I   was trying to look at what it was going to be  in five years. And it's the same lens that I'm   using on these seven names today. But even then,  I try to find pullbacks for every single entry.   Where dollar cost averaging for a long-term  investor like me is just the same boring game   plan I go to every week. As always, thanks so  much for watching and we'll see you next time.