Nvidia's Vera Rubin chips enter production as Huang refutes 12-month delay

Jensen Huang confirmed Vera Rubin is already in production, pushing back against a report that claimed the platform faced more than a year of delays.

Nvidia Corp. Chief Executive Officer Jensen Huang said the Vera Rubin AI chips are already in volume production, countering a research report that claimed the platform faced delays of more than 12 months.

"Vera Rubin is already in production. Giant amounts of production incoming," Huang told reporters at a developer event, according to a transcript of his remarks.

The confirmation follows a report from semiconductor research firm SemiAnalysis that Nvidia's Kyber NVL144 rack-scale solution — designed to house the Rubin Ultra architecture — had been delayed to 2028. Huang said the reports were "not true," though he did not address the Kyber rack timeline specifically. Nvidia had previously confirmed Vera Rubin production in January.

The reassurance matters because Nvidia has forecast $1 trillion in combined sales from Blackwell and Vera Rubin through 2027. Any disruption to its annual chip cadence could open the door for rivals such as Advanced Micro Devices Inc. to capture share in the $200 billion-plus AI chip market.

Nvidia's revenue has grown for 14 consecutive quarters, most recently rising 85% year over year to a record $81.6 billion in Q1 fiscal 2027, which ended April 26. The company's practice of refreshing its AI accelerator architecture every year has kept hyperscaler customers — including Microsoft Corp., Amazon.com Inc. and Alphabet Inc.'s Google — locked into its ecosystem. Those three companies alone are expected to spend more than $200 billion combined on AI infrastructure in 2026, according to industry estimates.

Vera Rubin is built on a next-generation architecture that succeeds the Blackwell platform. While Nvidia has not disclosed full specifications, the chip is expected to use TSMC's advanced process node and CoWoS (chip-on-wafer-on-substrate) packaging technology, which has been a bottleneck for previous generations. The Blackwell chip uses TSMC's 4nm-class process with 208 billion transistors and delivers 20 petaflops of AI performance per GPU. Vera Rubin is expected to push those metrics significantly higher, though Nvidia has not released official figures.

The supply chain for advanced packaging remains a key constraint across the semiconductor industry. TSMC, which manufactures Nvidia's chips and handles CoWoS packaging, has been expanding capacity at its facilities in Taiwan. Any disruption to TSMC's output could ripple through Nvidia's production timeline, making Huang's confirmation of "giant amounts" of production a critical signal for investors monitoring supply chain health.

Even the research firm that flagged the potential delay remains bullish on Nvidia. SemiAnalysis forecasts that Nvidia's data center revenue will exceed analyst estimates by 20% in the second half of fiscal 2027, according to its report. The firm's analysis suggests that demand for Nvidia's AI accelerators continues to outstrip supply, even as competitors like AMD and startups such as Cerebras Systems Inc. attempt to gain traction.

The next concrete test for Nvidia's roadmap will come Aug. 26, when the company reports earnings for its fiscal second quarter. If guidance continues to top expectations, it will signal that production is on track and that demand from hyperscalers remains strong. Nvidia has beaten consensus revenue estimates in each of the past six quarters.

Nvidia shares have gained more than 140% over the past 12 months, trading at roughly 35 times forward earnings. The stock's valuation reflects investor confidence that the company can maintain its dominant position in AI hardware, where it commands an estimated 80% market share for training chips. Any confirmed delay would have threatened that premium, making Huang's public denial a meaningful event for shareholders.

This article is for informational purposes only and does not constitute investment advice.