Palantir's US commercial revenue jumped 149% year-over-year. That's not a typo. The data is clear: enterprise AI spending is no longer experimental. Three Wall Street analysts—from BofA, JPMorgan, and Oppenheimer—each named a single stock as their top pick. Their choices: Palantir, Amazon, and Lam Research. On the surface, these are three unrelated companies. But the on-chain data—if we treat corporate financials as a ledger—tells a coherent story about AI infrastructure buildout.
Context
The analysis dissects three stocks at different layers of the AI stack. Palantir represents the application layer—deploying AI into enterprise workflows. Amazon (AWS) is the cloud platform layer, providing compute and storage. Lam Research is the physical infrastructure layer, supplying semiconductor equipment for chip fabrication. Each analyst gave a buy rating with significant upside: BofA's $255 for Palantir (+48%), JPMorgan's $365 for Amazon (+33%), Oppenheimer's $400 for Lam (+29%). The report I reviewed, dated August 2026, cross-referenced 31 data points from earnings calls, SEC filings, and analyst commentary.
Core Evidence Chain
First, Palantir's US commercial revenue grew 149%, and the company raised its guidance to 134% for the next quarter. The math works: 35% more customers and 76% higher revenue per customer multiply to ~138% growth, close to the reported 149%. This is high-quality expansion—not just adding small accounts. The ledger remembers everything: A 350% increase in average revenue per customer means Palantir is embedding itself deeper into Fortune 500 decision-making.
Second, Amazon's AWS backlog hit $496 billion, nearly 2.5x year-over-year, with 37% revenue growth. Smart contracts have no mercy, but cloud contracts do—and this backlog provides 2+ years of visibility. AWS's self-designed AI chips (Trainium, Inferentia) are listed as a growth driver, signaling that ASIC-based inference is gaining on NVIDIA GPUs.
Third, Lam Research reported NAND equipment revenue doubling and raised its 2026 WFE (wafer fab equipment) spending forecast to ~$1500 billion—a record. The CEO expects 2027 to be "exceptionally strong." This is a lagging indicator: chipmakers expand capacity only after seeing sustained demand from cloud and enterprise.
Contrarian Angle
Correlation does not equal causation. The three stocks form a supply chain, but the chain has weak links. Palantir's $172 stock price implies a P/S ratio of 80-95x based on 2026 revenue estimates. On-chain data doesn't lie: At 653 US commercial customers, the addressable market is limited. Even if they double the customer base, revenue per customer may compress. Amazon's backlog growth is impressive, but conversion to revenue depends on actual AI workload consumption—some pilot projects may shrink. Lam Research's NAND boom may be partly cyclical storage recovery, not pure AI demand. The analysts ignore ethical and regulatory risks: Palantir's government contracts face EU AI Act scrutiny, and Lam's China revenue is vulnerable to export controls.
Takeaway
Follow the TVL, not the tweets. In this case, TVL is the total value locked in order backlogs and revenue streams. The strongest signal is AWS's $496 billion backlog—it's real, audited, and growing. The weakest link is Palantir's valuation, which requires perfect execution for years. The next week's signal: watch AWS's Q3 2026 earnings for AI workload revenue share. If that number surprises to the upside, the entire infrastructure stack gets re-rated. If it disappoints, the house of cards trembles. Data doesn't have feelings. Neither do smart contracts. Neither should your portfolio.
