Jensen Huang Forecasts Chip Sales Doubling in 2027, Guides to $108B Quarterly Revenue

Speaking as scrutiny over AI capital-expenditure sustainability intensifies, Huang doubled his chip forecast and pointed to a step change in quarterly revenue guidance — $108 billion versus the $96.2 billion just posted. The core argument is that demand is no longer driven by speculative research budgets but by production workloads that generate revenue, meaning the buildout has a self-sustaining economic engine rather than a bubble dynamic.

Mechanically, the forecast rests on the transition of frontier labs and enterprises from training-heavy experimentation to always-on inference at scale — exactly the workload NVIDIA's Blackwell-class and successor accelerators are positioned for. Huang tied the 2027 doubling to hyperscaler commitments and the proliferation of agentic systems that run for hours without supervision, each consuming sustained compute rather than bursty training runs.

Competitively, the guidance lands amid a week where Samsung unveiled zHBM memory aimed at 10x faster AI responses and hyperscalers pushed their own silicon (Meta's MTIA 200, AWS Graviton/Trainium), all nibbling at NVIDIA's margins. Huang's framing is a direct rebuttal to investors calling a top: he is betting production economics outrun the custom-silicon threat.

Skeptics note that a CEO forecast is not a booking, and that OpenAI's own disclosure of $278B in projected cash burn through 2030 raises the question of who ultimately funds this compute. Reports also paired the forecast with Huang 'blocking a regulator,' hinting at the political friction accompanying the scale-up. Watch NVIDIA's next earnings for whether guidance converts into orders, and whether custom-silicon defection accelerates.