Santa Clara, CA — NVIDIA pulled back the curtain on its next-generation GPU architecture, codenamed “Vera,” during a private briefing with select analysts and cloud partners on Wednesday.

The Vera architecture promises a 4x improvement in inference throughput over the current Blackwell platform, while cutting power consumption by 30%. Production chips are expected to begin shipping in Q1 2027, with hyperscalers Microsoft, Google, and Amazon already securing advance orders.

The Numbers

NVIDIA’s data center revenue hit $38.2 billion last quarter — up 74% year-over-year — and the company now commands an estimated 92% of the AI training chip market. CEO Jensen Huang framed Vera as the architecture that will power “the next wave of agentic AI and physical intelligence.”

Shares of NVIDIA rose 2.3% in after-hours trading on the news. Competitor AMD, which has been gaining ground with its MI400 series, fell 1.1%.

Sources: NVIDIA Investor Relations, Bloomberg, Tom’s Hardware