NVIDIA Extends 20% Rally in Seven Days as H200 Capex Cycle Widens Beyond Training to Inference

NVIDIA's 20% rally over the past seven days, pushing the company to near $5.5 trillion in market capitalization, reflects a critical shift in how traders and analysts are framing the AI infrastructure cycle. The narrative has moved beyond simple "more chips for LLM training" to a longer-durationBond price sensitivity to interest rate changes., higher-TAM story centered on inference, edge deployment, and sustained capex across multiple customer cohorts. This broadening of the addressable opportunity is what's driving the stock's near-vertical move and the sustained breadth of enthusiasm across semiconductors and adjacent industries.

CoreWeave's newly announced $21 billion partnership with Meta represents a pivotal data point: instead of a one-time training capex burst, the market is now pricing in multi-year inference infrastructure buildout. The deal spans not just GPU clusters but power, cooling, and networking, the full stack that enables AI-as-a-service. Simultaneously, Cisco's strong results and forward commentary on AI networking demand (switches, optics, scale-across routing) confirm that capex is widening into adjacent layers. NVIDIA's denial of acquisition rumors around PC makers suggests confidence that organic chip demand, not M&A, will drive growth. Cisco soared on commentary that AI infrastructure demand is scaling beyond model training into long-term inference capacity, Meta's CoreWeave agreement underscores this thesis.

For the broader market, the NVDA momentumThe empirical fact that winners keep winning over the medium term. has compressed breadth and concentrated gains among mega-cap tech: the Magnificent Seven are now outpacing the Nasdaq, and the Russell 2000 is struggling. Sector rotation from financials, energy, and small-cap value back into mega-cap growth is evident. Market internals show the VIXThe 30-day implied volatility of S&P 500 options. The 'fear gauge.' near 14, supportive momentum, but some analysts warn that tight breadth and stretched valuations in the tech complex create tail risk if AI capex forecasts miss or rate expectations shift. The Trump-Xi Beijing summit, which included Huang and Musk, further buoyed sentiment by reducing perceived geopolitical risk to semiconductor supply chains.

Critics point out that the narrative rests heavily on sustained capex from a small number of mega-cap customers (Meta, Google, Microsoft, Amazon). If one of these firms signals capex moderation or efficiency gains, the entire cohort risks repricing lower. Additionally, the rise in US Treasury yields (driven by inflationThe rate at which prices rise across an economy. from the Iran war) could pressure technology valuations if real rates rise further.