NVIDIA earned $216 billion this year — and the world’s largest funds want more

The argument for NVIDIA is simple enough to state on a trading floor and complicated enough to take a decade to unwind. The company controls roughly 85% of the global market for AI training accelerators — the chips that teach large language models, generate images, and power the inference layer inside every major product released this year. When that number compresses to 75%, or 65%, the investment thesis changes. So far, it has not.

Revenue grew from $27 billion in fiscal 2022 to $216 billion in fiscal 2026 — a run that produced free cash flow of $49 billion in the latest quarter alone. The company plans to return roughly half of that to shareholders this year. In a sign that its capital appetite is expanding rather than plateauing, it recently launched a bond sale of at least $20 billion, its first such issuance in five years.

What drives the billionaire accumulation is not the earnings trajectory alone. It is the structural position. David Tepper’s Appaloosa Management has reorganized its portfolio so that four of its five largest holdings tie directly to AI infrastructure. Chase Coleman’s Tiger Global holds nearly half of its $22.8 billion portfolio in five names, all AI-related. These investors are not trading on quarterly beats. They are reading the chip stack the way an engineer reads a load-bearing wall.

The bear case is real. Custom silicon from Amazon, Google, and Microsoft is projected to grow from approximately 21% of the AI accelerator market in 2025 to nearly 28% in 2026 — a meaningful share shift at scale. AMD’s MI300 series has closed the performance gap on specific workloads. And a company priced at roughly $4 trillion faces a compound problem: it must keep growing fast enough to justify a multiple that already prices in several years of dominance.

For companies outside the shareholder class — AI labs, universities, and enterprise customers building on NVIDIA’s hardware — the concentration is less a financial story than a supply-chain one. A vendor controlling 85% of the market sets pricing on its own terms. The companies paying those prices have limited alternatives and no leverage to demand otherwise.

NVIDIA’s next quarterly results are expected in August 2026. The proceeds from the recent bond sale will be deployed for general corporate purposes, including potential refinancing of existing debt.

Tags: nvidia, AI infrastructure, semiconductor, AI economy, NVDA