David Tepper Puts 40% of Appaloosa Portfolio in Three AI Stocks

Key Takeaways

• David Tepper concentrated roughly 40% of Appaloosa Management’s portfolio in Amazon, Micron, and Taiwan Semiconductor Manufacturing at the end of Q2.
• The positions span cloud computing, memory, and advanced chip fabrication, giving Appaloosa Management exposure across the AI infrastructure supply chain.
• Strong demand supports the thesis, but semiconductor cycles, capital spending, valuation, and geopolitical exposure remain material risks.

David Tepper’s three largest AI positions amount to more than a collection of popular technology stocks. At the end of the second quarter, Amazon represented more than 15% of Appaloosa Management’s portfolio, Micron accounted for over 14%, and Taiwan Semiconductor Manufacturing represented more than 10%.

Together, the holdings form a broad wager on the physical and cloud infrastructure supporting artificial intelligence. Amazon supplies computing capacity through Amazon Web Services. Micron produces the memory required by AI accelerators. Taiwan Semiconductor Manufacturing, also known as TSMC, fabricates many of the advanced processors designed by semiconductor companies.

The macro backdrop helps explain the concentration. Gartner forecasts worldwide AI spending of more than $2.5 trillion in 2026, following about $1.5 trillion in 2025. Generative AI spending alone was expected to reach $643.9 billion in 2025, with roughly 80% directed toward hardware, including devices and servers.

That mix matters. Enterprise generative AI applications attract considerable attention, but much of the near-term spending continues to flow toward data centers, accelerated servers, networking equipment, memory, and cloud capacity. Tepper’s portfolio appears positioned around those capital-intensive layers rather than around a single AI application.

Amazon is the largest of the three positions, and Tepper added shares in Q2. Amazon Web Services is Amazon’s fastest-growing business and its largest contributor to profitability. AWS is investing heavily in chips and networking, with reported payback periods of two to three years while some customer agreements extend for five years or longer.

There is plenty of capacity risk in that model. Still, Amazon’s $496 billion backlog provides visibility, while greater use of internally developed chips could improve the economics of its infrastructure. Amazon has also projected that AWS could eventually become a $1 trillion revenue business.

AI is not confined to AWS. Amazon’s e-commerce operations are using AI and robotics to improve fulfillment efficiency, while its sponsored advertising business provides another high-margin growth engine. That combination gives Amazon several ways to translate AI investment into revenue or operating leverage.

Micron is a different proposition. Tepper trimmed the position during Q2, but it remained Appaloosa Management’s second-largest holding. The investment rests heavily on tight supplies of high-bandwidth memory, or HBM, which is packaged with GPUs and other accelerators to move data quickly enough for demanding AI workloads.

Producing more HBM can constrain other parts of the memory market. Micron, SK Hynix, and Samsung have emphasized HBM capacity, while HBM consumes substantially more wafer capacity than conventional DRAM. Competition with advanced logic chips for scarce extreme ultraviolet lithography equipment adds another bottleneck. The resulting imbalance has also lifted ordinary DRAM and NAND prices, which can benefit Micron because HBM represents a smaller share of its revenue than it does for some competitors.

Micron’s forward price-to-earnings ratio was below 6.5 times, but that apparently low valuation deserves context. Memory profits can move sharply when supply catches demand. Is this cycle structurally longer because of AI, or merely another unusually strong semiconductor upturn? That is the central question for investors.

Taiwan Semiconductor Manufacturing provides the broadest semiconductor exposure of the three. Tepper increased Appaloosa Management’s position by about 24% in Q2. TSMC benefits whether customers prioritize GPUs, custom AI application-specific integrated circuits, or advanced CPUs, provided those designs require its manufacturing capabilities.

Scale, process expertise, and high production yields have given TSMC substantial pricing power and made it closely connected to customer roadmaps. IDC projects AI infrastructure spending will exceed $1 trillion by 2029, with accelerated servers accounting for more than 94% of the market. That outlook supports TSMC’s opportunity, though geographic concentration around Taiwan remains a portfolio risk tied to geopolitical factors.

Enterprise adoption also depends on governance. The NIST AI Risk Management Framework gives organizations a structured way to assess reliability, security, transparency, and other AI risks. Wider use of NIST AI RMF 1.0 and ISO/IEC 42001 could influence which cloud and infrastructure vendors earn long-term enterprise commitments.

Tepper’s concentration therefore looks less like three separate stock selections and more like one connected thesis. Amazon, Micron, and Taiwan Semiconductor Manufacturing occupy different points in the AI production chain. If infrastructure spending remains elevated, all three can participate. If cloud demand slows, memory supply loosens, or chip capital expenditure outruns monetization, the same concentration could amplify losses. The opportunity is substantial. So is the exposure.