Jensen Huang's Nvidia: where the $68B Q4 actually comes from

Nvidia booked $68.1B in Q4 FY2026, with $62.3B from data center alone. Here's the line-by-line of where Jensen Huang's empire pulls revenue today: hyperscaler buildouts, sovereign-AI deals, and the 19,000-startup Inception loop.

May 12 at 1:00 PM5 min read

Jensen Huang, Nvidia revenue portfolio breakdown, 2026 — Jensen Huang during a 2024 visit with European Commission Vice-President Josep Borrell.· Photo by Peter Dasilva for the European Commission, via Wikimedia Commons (CC BY 4.0)

Nvidia booked $68.1 billion in revenue for the fourth quarter of fiscal 2026, of which $62.3 billion came from a single line: the data-center segment that sells AI compute to hyperscalers, sovereign-AI buyers, and enterprise model builders, Fortune reported on 25 February. Everything else under Jensen Huang, gaming, professional visualisation, automotive, embedded, and the venture / startup loop, fits inside the remaining $5.8 billion.

Data center is the entire company now

For full fiscal 2026, Nvidia’s data-center revenue was $197.3 billion, up from $115.2 billion in FY2025, according to Nvidia’s own newsroom release. That single segment grew 71% year over year and now accounts for the overwhelming share of the company’s top line. Gaming, the segment that founded the company, came in around $2.5 billion in Q4. Professional visualisation, automotive, and embedded together added roughly $1.1 billion. The dependency on data-center demand is total.

The Q4 jump is the cleanest read on Blackwell deployment economics. CNBC’s coverage of the quarter noted data center revenue rose 75% year over year, with Blackwell ramping faster than any prior generation. Free cash flow for the quarter alone was $35 billion, which is more than Nvidia’s entire FY2024 data-center revenue ($47.5 billion). The compounding is visible in the public numbers; it is not a forecast.

Nvidia Q4 FY2026 revenue split by segment — Q4 FY2026 revenue split. Sources: Nvidia Newsroom Q4 FY2026 release; Fortune.

Sovereign-AI deals are a second portfolio line, not a marketing slide

The fastest-growing component inside the data-center number is sovereign-AI: national governments and state-affiliated buyers committing multi-year purchase orders for Blackwell-class GPUs to stand up domestic compute clusters. Saudi Arabia is the largest single example. Nvidia’s announcement with Humain, the AI subsidiary of the Saudi Public Investment Fund, committed to AI factories totalling up to 500 megawatts powered by “several hundred thousand of Nvidia’s most advanced GPUs over the next five years.” Phase one is an 18,000-GPU GB300 Grace Blackwell supercomputer with InfiniBand networking.

That deal sits alongside a separate 5,000-GPU Blackwell deployment announced with the Saudi Data & AI Authority during a state visit including President Trump and the Saudi Crown Prince. The UAE has parallel ambitions through G42, which has pursued its own full-stack approach with backing from Emirati sovereign wealth funds. Fortune reported that Huang named Humain three separate times on Nvidia’s November 2025 earnings call, which is the kind of frequency company executives reserve for the deal flow they want investors to model into forward revenue.

Nvidia annual data center revenue, FY2024 to FY2026 — Data-center revenue trajectory across three fiscal years. Sources: Nvidia Q4 FY2026 release; CNBC.

The Inception program is a third channel that everyone forgets

The chip business has an attached startup loop that doesn’t appear as a separate revenue segment but does shape future demand. Nvidia’s Inception program reached more than 19,000 enrolled startups by May 2026, with 542 investments and 26 portfolio exits to date. Selected members get up to $100,000 in DGX Cloud credits and a 30% discount on standard DGX Cloud pricing once they cross a $75,000 minimum spend.

The economics of the Inception loop are easy to underrate. Every credit Nvidia hands to an Inception startup is a starter dose for a workload that, if the startup succeeds, becomes a paying DGX Cloud or H200/Blackwell on-prem account inside two years. Huang has described this consistently as a flywheel, not a marketing programme. Combined with the venture-capital alliance that lets Inception startups present to Nvidia’s VC partners, this is a market-development arm with a measurable conversion rate, separate from but downstream of the same revenue line as the hyperscaler buildout.

Saudi sovereign-AI Blackwell GPU deployment scale — Blackwell-class GPU commitments inside Nvidia’s announced Saudi deals. Sources: Nvidia Newsroom (Humain, SDAIA).

What it means

The Huang portfolio in 2026 is three loops feeding one line. Hyperscalers (Microsoft, Meta, Google, Amazon, Oracle) are the largest cluster of data-center buyers, sovereign-AI customers are the fastest-growing, and the Inception startup feeder is the long-tail conversion engine that keeps the buyer base widening. None of them is independent of the others; all three converge on Blackwell, soon Rubin, and the surrounding networking and software stack. The Q4 print is a snapshot of all three converging into one segment-level revenue print at the same time, which is why the data-center number now dwarfs everything else under Huang’s direct control.

Sources

Editorial standards: every claim is sourced. Tips: [email protected]

© 2026 StartupHub.ai. All rights reserved. Do not enter, scrape, copy, reproduce, or republish this article in whole or in part. Use as input to AI training, fine-tuning, retrieval-augmented generation, or any machine-learning system is prohibited without written license. Substantially-similar derivative works will be pursued to the fullest extent of applicable copyright, database, and computer-misuse laws. See our terms.

#Jensen Huang #Nvidia #AI Infrastructure #Data Center #Sovereign AI #Blackwell #Humain #AI Hardware #Nvidia Inception #GTC 2026