The three AI stocks that Wall Street loves are not about the most advanced models; they are about the infrastructure of trust. In a market where the code is the law, the real value lies not in the algorithm that predicts, but in the systems that truthfully execute. BofA, JPMorgan, and Oppenheimer have each named their favorite AI stock, and together they form a triad that spans the entire AI stack: Palantir for the application layer, AWS for the cloud layer, and Lam Research for the physical layer. As a decentralized protocol PM who has seen the ethical and technical pitfalls of centralized systems, I find this triad both fascinating and deeply concerning. The bullish narrative is clear, but the hidden risks – from valuation froth to ethical blind spots – demand a closer look.
Context: The Three Layers of AI Infrastructure
The AI market is no longer about who builds the biggest model. It is about who delivers the most efficient, reliable, and scalable infrastructure. Palantir represents the application layer, where AI is deployed into enterprise decision-making. AWS represents the cloud infrastructure layer, where compute and storage are provisioned. Lam Research represents the physical infrastructure layer, where the semiconductors that power AI are manufactured. Together, they form a chain of demand: Palantir’s success drives AWS usage, which in turn drives wafer fabrication equipment orders for Lam. This triad is not a random selection; it is a bet on the industrialisation of AI.
BofA’s pick is Palantir, with a $255 price target. JPMorgan chose Amazon (AWS), with a $365 target. Oppenheimer selected Lam Research, with a $400 target. Each analyst is a five-star rated on TipRanks, and their reasoning is based on measurable growth signals. Palantir’s U.S. commercial revenue surged 149% year-over-year, with a 35% increase in customers and a 76% increase in revenue per customer. AWS posted a 37% revenue growth and a backlog of $496 billion, nearly 2.5 times the previous year. Lam Research reported that its NAND revenue doubled and raised its 2026 wafer fab equipment (WFE) forecast to $150 billion, calling 2027 “exceptionally strong.”
Core: The Technical and Commercial Verification
Let us dissect each layer. Palantir’s growth is not just a spike; it is a pattern. The 149% commercial revenue growth, combined with a 134% guidance raise, suggests that management sees sustained demand. The customer count of 653 U.S. commercial clients, with an average revenue per customer of $3.5 million, indicates a “land-and-expand” strategy that targets high-value, high-stickiness contracts. Based on my experience auditing smart contracts, I know that high per-customer revenue often correlates with deep integration into the client’s core operations. Palantir’s Ontology architecture, which maps data to decision-making, creates a switching cost that is analogous to a protocol’s network effect. However, the $3.5 million average is a double-edged sword: it means that losing a single client could cause a noticeable revenue dip.
AWS’s $496 billion backlog is a milestone. As a PM who has worked with cloud providers, I know that a backlog of this size indicates multi-year commitments from enterprises. The 37% revenue growth is amplified by the fact that AWS’s operating margin is improving, even as it invests in custom AI chips (Trainium and Inferentia). These chips are designed to reduce the cost of inference, which is the dominant workload for AI applications. In my work with Aave governance, I learned that efficiency is not just about speed; it is about accessibility. By lowering inference costs, AWS is democratizing AI, much like how DeFi protocols lower barriers to financial services. The self-designed chips represent a vertical integration play that could weaken NVIDIA’s pricing power in the long term.
Lam Research’s NAND revenue doubling and the $150 billion WFE forecast point to a structural demand for memory and storage. AI servers require high-bandwidth memory (HBM) and large SSDs, which in turn require advanced etching and deposition equipment. Lam holds a leading position in NAND etching, making it a direct beneficiary. However, the $150 billion figure is a forecast, not a guarantee. As someone who lived through the 2022 bear market and the FTX collapse, I know that forecasts can be overly optimistic. The semiconductor industry is cyclical, and the current upswing is partly driven by the AI narrative. If AI demand slows, the equipment orders could be among the first to be cut.
Contrarian: The Blind Spots in the Triad
The bullish case for these three stocks is compelling, but it ignores several critical blind spots. First, valuation. Palantir trades at 80-95 times forward sales, a level that is historically reserved for the most speculative of growth stocks. BofA’s $255 target implies a forward price-to-sales of over 100x. While the company is growing quickly, such a valuation leaves no room for error. If commercial revenue growth slows to 100% next year, the stock could be cut in half. In my audit of the Parity Wallet, I learned that a single vulnerability can destroy billions in value. Similarly, a single earnings miss can destroy Palantir’s premium.
Second, ethical risks. The article makes no mention of AI ethics, data privacy, or regulatory compliance. Palantir’s origins in government surveillance (Gotham) raise significant concerns under the EU AI Act, especially in high-risk categories like policing and border control. I have seen how ethical failures can destabilise protocols, as when the FTX collapse shattered trust in centralized exchanges. Palantir’s lack of transparency around its government contracts is a vulnerability that could erupt at any time. AWS is not immune either; its data localisation challenges in Europe and China could drive up compliance costs. Lam Research is exposed to export controls on semiconductor equipment to China, which could shrink its addressable market.
Third, the assumption that the AI demand cycle will continue indefinitely. The current infrastructure buildout is based on a belief that enterprise AI adoption will accelerate. But what if the promised ROI fails to materialise? I have seen this pattern before in the crypto bull market of 2021, where projects raised billions on the promise of decentralisation, only to crash when the hype faded. Palantir’s 149% growth is impressive, but it is coming off a relatively small base. The real test is whether the growth can be sustained as the base grows. The 76% increase in revenue per customer suggests that existing clients are expanding, but that is not infinitely scalable. The customer count of 653 is small; even if it triples, the revenue per customer would likely decline as new clients are smaller.
Takeaway: The Infrastructure of Trust
The AI infrastructure triad is a bet on the future, but it is a bet that requires constant vigilance. Trust is the new token, and it must be earned through transparency, ethical design, and sustainable growth. Code has conscience, and the systems we build today will shape the digital world of tomorrow. As I look at the $150 billion WFE forecast and the $496 billion AWS backlog, I see a market that is pricing in perfection. The reality is that AI adoption will be messy, with setbacks, regulatory hurdles, and ethical dilemmas. The question is not whether these three companies will succeed, but whether their current valuations already reflect that success. Liquidity flows where belief resides, and the market’s belief in AI is strong. But belief alone is not enough. We need to see the evidence that the infrastructure is built on a foundation of trust, not just hype. Will the next downturn reveal the cracks? Only time will tell.