Hyperscalers commit $725B to AI; chip cycle accelerates

Palantir's disclosure that hyperscalers are committing $725 billion to AI infrastructure has become the most cited signal that the secular AI capex story remains intact. This figure anchors a broader narrative that counters recession fears and validates the tech earnings beat streak evident in this week's earnings season. NVIDIA, the bellwether, is approaching earnings on May 21 with analysts citing a 78% probability of rising in 60 days and a median Monte Carlo target of $248.72. AMD, which rallied 47% YTD on execution around RXT enterprise-AI cloud agreements, is similarly priced for continued outperformance (84% probability of rising in 60 days, target $628.14).

Broadcom and Arm Holdings are receiving less individual attention in social media but are flagged in the same institutional capex thesis. Goldman Sachs' decision to keep rates higher for longer does not change the fact that capex, especially when self-funded by hyperscalers' operating cash flows, remains insensitive to Fed policy. Amazon reported negative free cash flowCash generated after maintenance capex; the actual money the business throws off. of $18 billion in Q1 2026 following $44 billion in property and equipment purchases, but the stock rallied 45% over a 6-12 month window as investors accepted that growth maturation requires heavy upfront capex. The narrative is that this is not a cyclical downturn but a structural reordering of IT budgets toward AI.

The supply chain implications are significant. CleanSpark, a Bitcoin miner and data center operator, doubled MW under contract YoY and increased Bitcoin holdings by 14% while expanding ERCOT-approved capacity (585 MW). This dual mandate (both AI compute rental and Bitcoin mining) is emblematic of how data center operators are hedging demand by diversifying workloads. Innio (gas engine manufacturer), a critical supplier to data center backup power systems, is filing for a $4.8 billion IPOInitial Public Offering - a company's first public sale of stock. to tap this wave. Cerebras Systems, an AI chipmaker, also upsized its IPO target to $4.8 billion on strong demand signals.

The risk is valuation compression if capex growth decelerates faster than consensus expects, or if energy constraints (Hormuz closure, grid capacity limits) force operators to slow deployment. Additionally, the narrative assumes that pricing power in AI chips remains durable; if competition from TSMC's advanced nodes or Chinese designs intensifies, gross margins could compress despite strong unit demand. Early-stage AI startups (IONQ, QUBT) are being shorted as sentiment swings toward established players with proven revenue models.