원문 보기
Welcome to this explainer. Today, we're diving straight into a staggering $7.7 billion Q2 stock portfolio from David Tepper's Appaloosa Management. But, we're not just reading off a list of stocks here. We're looking at a massive, incredibly concentrated bet that actually lays out a hidden map of the whole global AI economy.
Let's just jump right into it. A jaw-dropping 40% of Tepper's entire disclosed stock portfolio is riding on exactly three companies. This isn't some broad play-it-safe index approach. This is an absolutely huge level of concentration from a legendary opportunistic investor. So, who are the big three?
Well, it's Amazon, making up 15.4% of the portfolio, Micron Technology at 14.6% and Taiwan Semiconductor Manufacturing or TSMC sitting at 10.2%. Put those together and they are the undisputed core of his public market strategy right now. Section one, the 40% mystery. When you look at the 13F filing, you see Tepper treating these three AI-adjacent investments completely differently.
It's honestly a major paradox. Over in Q2, he aggressively bought more Amazon and TSMC, boosting those stakes by almost 16% and 24%. But, right at the exact same time, he slashed his Micron shares by a massive 41% and he completely dumped another memory chip maker, SanDisk, just walking away from a $179 million position entirely.
So, if this is just a simple all-in bet on artificial intelligence, we have to ask, why the sharp split? Why is he handling these specific stocks so differently? What exactly is his team seeing that the rest of the market might be missing? Let's move to section two, Amazon, the downstream cloud layer.
See, the analysis here points to Tepper viewing Amazon as way more than just an e-commerce giant. To him, it looks like the downstream infrastructure layer of the AI economy. During the quarter, he scooped up another 680,000 shares. That brings him to a conviction level stake of 5 million shares, cementing Amazon as his top holding worth nearly 1.2 billion dollars.
And the crucial point is hiding right there in the fundamentals. Amazon Web Services, or AWS, is seeing just massive enterprise demand. Their Q2 earnings report pretty much validated this thesis, with AWS revenue hitting 42.2 billion dollars. That's up nearly 37% year-over-year. But even more staggering, a 496 billion-dollar backlog.
That right there proves that Amazon's colossal 200 billion-dollar AI capital expenditures are already translating into locked-in contracted demand. On top of that, Amazon is taking steps to reduce its reliance on buying merchant GPUs by designing its own custom chips, like Trainium. And they've already locked down multi-gigawatt commitments from heavyweights like OpenAI and Anthropic.
It looks like Tepper sees AWS as this dominant platform, basically a toll booth that's going to get paid no matter which specific AI software model actually wins the race. All right, section three. TSMC, the fabrication gatekeeper. Appaloosa has been strategically building up this TSMC position for six consecutive quarters.
Why? Well, it really just comes down to absolute dominance. Look at this. A whopping 77% of their wafer revenue now comes from advanced nodes. We're talking chips that are 7 nanometers and below. Meanwhile, demand for their highly anticipated next-generation 2-nanometer chips is already completely outrunning supply in its very first ramp quarter.
The beauty of TSMC is where they sit in the market. Literally every major player, Nvidia, AMD, Apple, Broadcom, and yep, even Amazon's custom silicon we just talked about, they all have to use TSMC to physically print their chips. So, by holding this stock, Appaloosa secures direct, highly concentrated exposure to the raw volume growth of AI silicon, but they stay completely insulated from the cutthroat design competition between the individual chipmakers themselves.
Which brings us to section four, the memory bottleneck paradox. This is where we hit the most counterintuitive move in the entire filing, the massive reduction in Micron. If memory is so absolutely vital to AI, why on earth sell 690,000 shares? Well, the key to the mystery is right here. Micron's stock surged an unbelievable 242% during the second quarter alone.
It was an absolute rocket ship, totally driven by surging demand in AI data centers. Analysts are interpreting this move as a total masterclass in risk management. Tepper sold off 41% of his shares, basically locking in massive profits on a highly volatile cyclical hardware stock. But here's the kicker.
Because of that explosive 242% price surge, the remaining position he held on to, it still doubled in value sitting at over 1.1 billion dollars at quarter end. He kept Micron as his second largest holding because high bandwidth memory or HBM is still a critical physical bottleneck for AI data centers.
These generative AI models require massive continuous access to physical memory pools. And since Micron is locked into long-term strategic contracts with minimum price revenue near a hundred billion dollars, it looks like Tepper booked the parabolic gain, but brilliantly kept that core structural exposure.
Now, section five, anticipating the next bottleneck. Honestly, perhaps the biggest takeaway from this explainer is what Tepper actually did with those profits. It tells us how he might be positioning capital for the next phase of the AI buildout. If you track the money, you see a very deliberate rotation.
He's stepping out of those volatile chip components that live and die by pricing cycles and funneling that capital directly into power utility stocks like Vistra and Energy Energy. This sequence really reveals the architecture of the strategy. It suggests Tepper isn't just blindly buying some generic basket of AI stocks.
Instead, he seems to be methodically anticipating and rotating through the structural bottlenecks of the whole digital economy. Think about it. Going from the foundries making the chips to the memory bottleneck, up to the cloud infrastructure, and finally down to the actual physical electricity required to power these massive compute workloads.
And to further support this deep infrastructure thesis, the filings showed he just opened a brand new $107 million position in CoreWeave. This is a private turn public AI cloud computing provider that completely relies on standing up energy dense data centers. It's just another layer of the infrastructure cake that collects revenue regardless of which specific software app ends up dominating.
Now, of course, to be completely objective here, we have to step back and look at the macro risks to a strategy that is this concentrated. The most prominent threat, a potential capital digestion phase. Basically, a pause in spending if hyperscalers like Amazon and Meta decide that enterprise software adoption just doesn't justify their massive CapEx budgets anymore.
Plus, there are obvious geopolitical threats to TSMC's foundries in Taiwan. Any disruption there could totally halt the global advanced tech supply chain. And lastly, those premium valuation multiples across the tech sector leave this kind of portfolio vulnerable to a broader market correction if inflation stays sticky.
Which leaves us with a final thought to ponder today. The analysis of this portfolio suggests David Tepper has mapped out a sequence of bottlenecks, taking profits from volatile silicon and moving them straight into the physical power grid. So, as visionary capital continues to flow down this path, where do you think the next great bottleneck will emerge.
Comments 0