Nvidia's $68B Quarter Smashes Estimates, But 5% Drop Signals the 'Show Me the ROI' Phase of AI

Nvidia reported fiscal Q4 2026 revenue of $68.13 billion (+73.2% YoY), beating consensus estimates of $66.21 billion, with EPS of $1.62 (+82.0% YoY) exceeding forecasts by $0.08. Data centre revenue surged 75.1% to $62.3 billion. Forward guidance of $78.0 billion for Q1 FY27 topped expectations of $74.5 billion. The S&P 500 initially rose 0.81% to close at 6,946.13 ahead of the after-hours report.

Yet the post-earnings reaction told a different story: NVDA fell 5% on February 26 despite the record results, dragging the Nasdaq down 1.18% to 22,878.38 and the S&P to 6,908.86. Investors are increasingly concerned that hyperscaler AI capex is depleting cash flows without proportionate revenue returns — the "show me the ROI" phase of the AI cycle has arrived.

For cross-border investors, Nvidia's results crystallise the defining tension in global technology allocation. The $78 billion quarterly revenue trajectory confirms that AI infrastructure investment is accelerating globally — but access to frontier compute varies dramatically by jurisdiction. US export controls restrict the highest-performance chips from reaching certain markets, creating a bifurcated AI compute landscape. This fragmentation generates both risk and opportunity: companies positioned in export-compliant AI infrastructure, power systems, and cooling technology command premium valuations, while those dependent on restricted chip access face structurally compressed innovation timelines. The sell-the-news reaction also signals a broader rotation from AI hype toward AI infrastructure — favouring cross-border investments in power, data centres, and connectivity over pure-play software.