Nvidia Hits $5.26 Trillion and Releases Nemotron 3 Nano Omni Plus "Ising" Open Models for Quantum

Nvidia's market capitalisation reached a record $5.26 trillion in late April 2026 as the company continued its run of headline-grabbing product announcements. Two stood out in the most recent cycle: Nemotron 3 Nano Omni, a small-but-multimodal model in Nvidia's Nemotron family, and the Ising family of open models for quantum computing.

Nemotron 3 Nano Omni

Nano Omni is Nvidia's attempt to occupy the small-model multimodal niche. Its target is developers building edge and on-device applications: vision, speech and text in a model small enough to run on consumer hardware or in cost-sensitive cloud deployments. The model was released under an open-weights licence, which both signals Nvidia's seriousness about developer adoption and underwrites enterprise integration without licence-fee friction.

The strategic logic is straightforward. Nvidia's most defensible asset is CUDA, the developer ecosystem on top of which much of modern AI is built. Releasing genuinely useful open-weight models trained, optimised and packaged on Nvidia hardware is one of the more effective ways to deepen that ecosystem against rising competition from AMD's ROCm and from cloud-internal silicon (Google TPU, AWS Trainium, Microsoft Maia).

The Ising family

Ising is a separate bet. The family is open models aimed at quantum-computing applications — variational circuits, optimisation problems, simulation of physical systems where quantum hardware is starting to show advantage. The naming references the Ising model, a foundational construct in statistical physics that maps naturally onto quantum hardware.

Nvidia's quantum strategy is hybrid by design. Most useful quantum applications today require a classical co-processor that handles the heavy compute outside of the quantum step. Nvidia builds that classical co-processor. The Ising open models are aimed at making the hybrid stack a default — and at ensuring that when quantum hardware does become commercially relevant at scale, Nvidia is in the room.

The market context

Bank of America raised its 2026 chips forecast to $1.3 trillion in revenue, up $300 billion from its prior estimate four months earlier. Nvidia, Broadcom, Marvell and AMD are named as the principal drivers. Even with AMD's Meta deal, Nvidia's data-centre momentum and forward guidance keep it dominant; the rest of the market is gaining at the margin, not at Nvidia's expense.

The Anthropic note

One quiet piece of news worth flagging: reporting that Anthropic is moving toward Trainium-based deployments for some Claude inference workloads. If that scales, it is the most consequential customer concentration shift in the cloud-AI market — not because Anthropic is the largest customer, but because it signals that the alignment-focused frontier lab views non-Nvidia inference as a strategic option. That conversation is one to watch through the back half of 2026.