Semiconductors Are Gushing Cash… Here’s What’s Next in The AI Trade | Ben Pouladian
Speakers
Ben Pouladian
— Analyst, BEP Research
Summary
Ben Pouladian explains why the AI semiconductor boom is not a dot-com repeat, highlighting Nvidia's software-hardware moat, the shift of the true bottleneck from GPUs to powered land, and the global AI race between the US and China. He shares his top stock picks including Nvidia, Apple, and Bloom Energy, and warns against chasing capacitor bottlenecks.
- Nvidia is not a commodity; its software-hardware co-design and CUDA ecosystem create a defensible moat, and the AI capex cycle remains early.
- The real shortage is not GPUs but energized land and tradesmen to build data centers, making fast-deploy power solutions like Bloom Energy critical.
- Apple's privacy-focused on-device AI and its potential to collect a cut of AI service revenues make it an attractive AI play, despite current product shortcomings.
- High-bandwidth memory (HBM) is structurally supply-constrained due to manufacturing complexity, favoring Micron and other memory producers until at least 2028.
- Chasing low-end semiconductor bottlenecks like MLCC capacitors is risky; commoditization and Chinese competition will eventually crush prices.
- Super Micro remains a speculative turnaround story after raising $7 billion, with its liquid-cooled racks essential for Nvidia and AMD deployments.
- The AI boom has barely penetrated the broader economy, suggesting massive remaining upside for semiconductor demand.
- Geopolitical trust issues mean Western frontier models will maintain pricing power against Chinese open-source alternatives, benefiting US AI leaders.
Ideas
Nvidia dominates AI compute with software moat.
Nvidia is not a commodity company; it creates intelligence through software-hardware co-design and has a massive moat with its CUDA ecosystem and Neotron model. The AI capex cycle is still early, and Nvidia's backlog is measured in years. Despite a recent share price surge, valuation multiples have remained flat or declined because earnings growth has been explosive. He has been long since 2016 and considers it his top pick.
Semiconductor super cycle from AI capex.
The semiconductor sector is in a multi-year super cycle driven by 20 years of underinvestment in hardware capex and the massive demand from AI. Earnings have exploded while valuation multiples have remained flat or declined, meaning the bull market is earnings-driven, not multiple expansion. The complexity of the supply chain and long lead times mean the capex cycle has further to run.
HBM memory demand strong until 2028.
High-bandwidth memory (HBM) is special and not a commodity like traditional DRAM. AI models require vast memory for context, and the manufacturing complexity (stacking layers with nanometer precision) creates supply constraints. The three dominant players (Micron, SK Hynix, Samsung) will see strong demand for years, with capacity catching up only by early 2028, making a small allocation attractive despite cyclicality.
Bloom Energy solves data center power crunch.
Bloom Energy's fuel cell technology provides quiet, modular, and fast-deployment power for data centers, solving the critical bottleneck of energized land and grid delays. The company has found product-market fit after 20 years, and its solution is being used by data center developers to get power quickly, making it a direct play on the data center buildout.
Apple wins from AI privacy and services cut.
Apple will bring useful, privacy-focused AI to consumers, and its App Store will capture a cut of AI service revenues (like 15–20% from AI apps). Although current Apple Intelligence is underwhelming, Apple's ethos of on-device data privacy and the potential for AI to drive a super cycle in services make it an attractive AI play.
Super Micro rebound play on AI hardware.
Super Micro is a fallen angel with valuable liquid-cooled server rack technology essential for AI data centers. After a recent accounting scandal and stock plunge, the company raised $7 billion from JP Morgan and others, suggesting institutional backing and an order book. Nvidia and AMD depend on them, and the stock may recover from depressed levels as the business rebuilds.
Capacitor stocks are overhyped and commoditized.
Chasing capacitor bottlenecks like MLCCs is a mistake; they are low-end commoditized components that will eventually see supply catch up and price crashes. China can flood the market, and the same pattern played out during the iPhone super cycle. Stocks like Vishay that have run up on this narrative are likely to suffer.
Up Next
This Monetary Matters video, published July 14, 2026, features Ben Pouladian discussing NVDA, SMH, MU, BE, AAPL, SMCI, VSH. 7 trade ideas extracted by AI with direction and confidence scoring.
Speakers: Ben Pouladian · Tickers: NVDA, SMH, MU, BE, AAPL, SMCI, VSH