Nvidia posts record $81.6bn quarter on AI demand

On 20 May, Nvidia reported record Q1 FY2027 revenue of $81.6 billion (quarter ended 26 April 2026), up 85% year on year and 20% sequentially. Data-centre revenue hit $75.2 billion (up 92% YoY) on the ramp of the Blackwell 300 generation — now 92% of total revenue. Gaming, automotive, and professional visualisation chips account for less than one-tenth of sales. CEO Jensen Huang called the AI-factory buildout 'the largest infrastructure expansion in human history' — and said it is still accelerating.

The board added $80 billion to its buyback authorisation and raised the dividend to $0.25 per share — signalling confidence that the demand is durable, not a spike. At 92% revenue concentration in data centre, Nvidia is effectively a pure-play AI infrastructure company. That is also the catch: there is no diversification buffer if AI capital expenditure slows.

The next-generation Rubin platform is in production, with partner systems targeting availability in H2 2026 — Nvidia's stated next leg after Blackwell, built around its NVLink networking fabric for large GPU clusters. If it lands on schedule, the growth run has a product reason to extend into fiscal 2028. If hyperscalers — Microsoft, Google, Amazon, Meta — begin to pull back capex before Rubin ships, that is the earliest warning sign to watch. The risk: with data centre at roughly 92% of revenue and virtually all frontier AI running on Nvidia silicon, a supply stumble or capex digestion phase would reprice the whole sector, not just this one stock.