Nvidia Says Blackwell + Rubin Orders Will Top $500 Billion Through 2026

Nvidia's quarterly numbers no longer measure a chip company. They measure the rate at which the world is building AI infrastructure. At CES in January 2026, CEO Jensen Huang raised his prior estimate, telling investors that the company expects orders and revenue from the Blackwell and Rubin platforms to surpass $500 billion through the end of calendar 2026 — a number that would have been incomprehensible to investors as recently as 2023.

What Blackwell and Rubin are

Blackwell is the GPU architecture that succeeded Hopper, with H100 and H200 having been the workhorses of the 2023-2025 AI training cycle. Blackwell adds substantial improvements in performance per watt, memory bandwidth and rack-scale density. Rubin is the architecture due to follow Blackwell, expected to enter volume shipment in late 2026.

Together, the two platforms power the next two cycles of hyperscaler training and inference. Microsoft, Google, Meta, Amazon, Oracle, the major sovereign AI projects (including Luxembourg's mid-2026 MeluXina-AI), and a long tail of enterprise and start-up customers are buying or have committed to buy the systems.

The cash-flow forecast

Bank of America estimates that Nvidia will generate more than $400 billion in free cash flow across CY26-CY27 combined — roughly equivalent to Apple and Microsoft combined over the same period. That number is so large that it strains the ability of the company's existing capital-return programmes to absorb it. BofA has explicitly called it time for Nvidia to pay shareholders meaningfully more — through buybacks, a substantial dividend, or both.

The risks

Three. First, demand concentration. Hyperscalers — Microsoft, Google, Amazon, Meta and a handful of others — account for an outsized share of orders. Any slowdown in their AI capex translates almost directly into Nvidia's revenue. Second, the China question. Export controls have already pushed Nvidia to develop China-specific variants and have introduced ongoing regulatory uncertainty. Third, competitive pressure. AMD, Broadcom (custom silicon for hyperscalers), and the growing ecosystem of cloud-provider-internal accelerators (AWS Trainium, Google TPU) eat into the addressable market over time.

The market-cap question

Nvidia has been priced at a 69.5% probability of overtaking Apple as the world's most valuable public company by end-June 2026 in prediction markets. Apple's strong Q1 2026 has dampened those odds, but the underlying trajectory — orders growing faster than the broader market, cash generation outpacing returns, and a product cycle that has years to run — keeps Nvidia in the conversation.

Earnings on 20 May

Nvidia's next results are scheduled for 20 May 2026 and will be the most-watched corporate event of the quarter. The data points to track: Blackwell shipment cadence, Rubin pre-orders, China contribution, and gross-margin trajectory. Each will tell a piece of the bigger story.

For now, the headline is simple: the AI infrastructure build-out is real, it is enormous, and Nvidia is, by an order of magnitude, its largest beneficiary.