Nvidia announces third-quarter results and transition to Vera Rubin chips

2–4 minutes

Nvidia will release its third-quarter financial results on Tuesday, November 17, at around 6:20 p.m. Brasília time, followed by a teleconference at 7 p.m. The quarter ended on October 25, just before Jensen Huang resigned as CEO and left the board on November 12. The results will reveal how the company performed before the leadership change and will mark the first test for interim CEO Colette Kress.

Revenue forecast and the importance of the Vera Rubin system

Analyst consensus compiled by LSEG predicts revenue of around $109 billion, slightly higher than the $108 billion forecast by the company in August. In the same quarter last year, Nvidia generated $57 billion in revenue. Beyond the total figure, investors are closely watching the performance of the new Vera Rubin system, which Kress estimated in August would account for about 20% of data center revenue during the quarter.

The Vera Rubin system is already being delivered to customers, though its scale remains uncertain. CoreWeave announced in September that its subsidiary Cognition is already using the system, and Oracle launched the NVL72 commercially in October. The NVL72 combines 72 Rubin GPUs and 36 Vera CPUs in a single rack, requiring high-bandwidth memory, fast connections, and advanced cooling.

Technical specifications and performance adjustments

Nvidia’s original specification mentioned up to 22 terabytes per second of memory bandwidth per GPU and 3.6 terabytes per second for the NVLink interconnect. The current commercial specification has reduced these figures to 19.2 terabytes and 3.0 terabytes per second, respectively. This reduction does not necessarily mean lower speed, as the impact depends on the workload and existing bottlenecks.

Production challenges and competition

Issues with HBM4 memory could delay production expansion until 2027, according to TrendForce. The consultancy predicts a gradual adoption of Rubin throughout the year, without Nvidia declaring a general delay. Meanwhile, competitors like AMD already offer their Helios racks to Meta, and Google is expanding the use of its own chips for model training and application execution.

Margin outlook and market valuation

Kress forecast a gross margin of 74% for the current quarter, with a tolerance of half a percentage point, and 71% to 72% for the next quarter, reflecting higher memory costs. The margin is a critical issue, especially since Nvidia also finances part of the infrastructure where its products will be used—a strategy that caused disagreements between Kress and Huang.

Market valuations reflect this uncertainty. Baird downgraded the stock to neutral, citing a lack of visibility into supply and investments by major cloud providers in 2027. Morgan Stanley removed Nvidia from its favorites list but maintained an above-average recommendation. Analyst Stacy Rasgon of Bernstein cut the target price in half to $200, without indicating an immediate deterioration in operations.

What’s next for Nvidia

Exceeding the revenue forecast will only be part of the answer. If Rubin gains traction without further pressuring margins, Kress can argue that the transition is on track. If volume deliveries fall short of expectations, Blackwell chips could sustain the quarter, but they won’t resolve the question Huang left behind: when will the next generation take over most of the demand for artificial intelligence?

Frequently Asked Questions

  • When will Nvidia release its earnings report? On November 17, at around 6:20 p.m., with a teleconference at 7 p.m. Brasília time.
  • What is Vera Rubin? It is the next generation of AI systems following Blackwell. In the NVL72 configuration, a rack includes 72 Rubin GPUs and 36 Vera CPUs.
  • Has Vera Rubin been delayed? Initial deliveries have already been announced. Analysts are questioning the pace of volume delivery, not the product’s existence.