Hot AI News

Nvidia posts $115.8B quarter and guides to $130B, with Kress still interim CEO

Nvidia reported $115.8B in third-quarter revenue, 92.9% of it from data centers, and guided to $130B next quarter while assuming no China data center compute sales.

Nvidia posts $115.8B quarter and guides to $130B, with Kress still interim CEO
Image: Nvidia

Nvidia reported third-quarter fiscal 2027 results on November 17: revenue of $115.8 billion, up 103% from a year earlier. It guided the next quarter to $130 billion, with no China data center compute sales assumed.

A $107.6 billion data center quarter

The quarter ended October 25, 2026. Total revenue rose 20% from the previous quarter. Data Center revenue reached $107.6 billion, up 21% sequentially and 110% on the year. A new segment called Edge Computing, which covers products such as gaming and PC chips (the release has no separate gaming line), brought in $8.2 billion, up 14% and 42% respectively.

Nvidia CEO Jensen Huang, in a black leather jacket, gestures as he speaks during a talk at Stanford University.
Photo: Anders Eidesvik / Wikimedia Commons, CC BY 4.0

Profitability stayed very high. GAAP gross margin was 74.8% and non-GAAP was 74.9%. GAAP operating income was $77.5 billion, GAAP net income $65.3 billion, and GAAP diluted earnings per share $2.70 (non-GAAP: $2.68). Free cash flow came to $43.8 billion. Nvidia notes that since the first quarter of fiscal 2027 its non-GAAP measures no longer exclude stock-based compensation, and that earlier non-GAAP figures were restated to match.

For the fourth quarter, Nvidia expects revenue of $130.0 billion, plus or minus 3%, with gross margin of 71.0% to 72.0%. That quarter has 14 weeks, one more than usual, which accounts for part of the sequential step up. Operating expenses are guided to about $11.0 billion on a GAAP basis.

Nvidia returned $28.1 billion to shareholders in the quarter. In September its board added $150 billion to the buyback authorization, and about $227.3 billion remained at quarter end. The next dividend is $0.25 per share, payable December 23 to holders of record on December 2.

Advertisements

Vera Rubin ramps while the CEO seat stays empty

Interim CEO and CFO Colette Kress framed the quarter around the new platform: “Vera Rubin is off to an explosive start, our roadmap is on track and our ambition has never been greater.” According to the release, Nvidia began volume shipments of the Vera Rubin NVL72 rack, which is running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius.

The release also lists an agreement to acquire Hugging Face, a planned expanded Meta deployment of Vera Rubin hardware, a modular reference design for AI factories called DSX POD, and NVLink Fusion support for MediaTek’s custom chips. On the Edge side it cites RTX Spark Windows laptops, DLSS 5 in several games and a 64GB DGX Spark model.

The results land five days after Nvidia said Jensen Huang had resigned as CEO and left its board, with Kress taking over on an interim basis while keeping her CFO role. As our earlier report on the resignation noted, Nvidia gave no reason, and shares fell nearly 9% early on November 9, according to AFP, after reports about a foreign intelligence effort aimed at Huang. The earnings release says nothing about either matter beyond Kress’s title.

Numbers in context: Concentration and the China gap

The scale is easier to read against its own parts. Data Center was about 93% of revenue this quarter (107.6 of 115.8), so the company’s results now track AI infrastructure spending almost one to one. Operating income of $77.9 billion non-GAAP is roughly 67% of revenue, a margin few hardware companies have approached.

The guidance implies about 12% sequential growth, slower than the 20% just delivered, and the extra week in the quarter flatters it. Gross margin is guided down to 71% to 72% from nearly 75%, which Nvidia does not explain in the release. A product transition to new rack-scale systems is one plausible reason, but that is an inference, not something Nvidia states.

The China assumption matters too. Nvidia says its outlook includes no Data Center compute revenue from China. The same day, the board created a standing committee for export controls and national security, whose members meet at least quarterly, as we reported. Export rules are set by governments, and the committee’s powers over licensing decisions are not described.

The Vera Rubin claims also deserve some care. In its MLPerf debut, Nvidia said the platform trained DeepSeek-V3 on 1,024 GPUs in about 4.14 minutes, against 8.03 minutes for GB300. That was Nvidia against Nvidia, and the Vera Rubin inference results were still pending MLCommons verification. Our coverage noted Nvidia gave no pricing or shipping dates in those posts.

Advertisements

What the release does not settle

The release gives no data on how many Vera Rubin racks shipped, what share of Data Center revenue they represent, or how the Hugging Face deal is structured, priced or timed. It also names no timetable for a permanent CEO. Those are the figures analysts will press for, and they are absent here.

For customers, the practical point is availability: Vera Rubin NVL72 capacity is now being delivered to the five cloud providers named, so access for most developers will come through those clouds rather than direct purchase. All forward-looking numbers are Nvidia’s own guidance and have not been independently verified.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Alex CarterAI News & Big Tech Correspondent

Alex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like