Adobe to build Firefly AI models on Nvidia stack, add AI agents and enterprise content tools

The partnership brings Nvidia’s AI models, libraries and infrastructure into Adobe products, while enabling custom enterprise AI and 3D product visualisation.

Adobe will build its Firefly AI models on Nvidia’s AI stack, including CUDA-X, NeMo and Cosmos models, for creative and marketing pipelines.

The partnership integrates Nvidia’s computing infrastructure, AI models and software libraries into Adobe’s creative and marketing platforms.

Adobe will also use Nvidia Nemotron models, Agent Toolkit and OpenShell to add AI agents to content creation and marketing workflows.

These agents will support tasks across content production and campaign execution.

Nvidia’s founder and CEO, Jensen Huang, said, “AI is giving every industry the ability to redefine what’s possible.”

“For more than 20 years, NVIDIA and Adobe have partnered to push the boundaries of design and creativity. Today, we are taking that partnership to a new level, uniting our research and engineering teams to accelerate Adobe’s beloved applications with NVIDIA CUDA and jointly build state-of-the-art world foundation models that reimagine creativity and transform customer experiences,” Huang added.

Adobe and Nvidia will also work on Nvidia NemoClaw, an open source stack for running always-on AI assistants.

As part of the Nvidia Agent Toolkit, it installs the OpenShell runtime, which provides a secure environment to run autonomous agents and models such as Nvidia Nemotron.

Adobe will integrate Nvidia AI technologies across products including Adobe Acrobat, Photoshop, Premiere Pro, Frame.io, Adobe Experience Platform and GenStudio.

The company will extend Firefly Foundry by integrating Nvidia technologies to build custom AI models trained on proprietary enterprise and brand data.

These models will help generate commercially safe content at scale across image, video, audio, vector and 3D formats.

Adds enterprise AI models and 3D product visualisation

Adobe is also introducing a cloud-native 3D digital twin solution for marketing, currently in public beta.

The solution creates virtual replicas of physical products for marketing and commerce use.

It is built using Nvidia Omniverse libraries and supports OpenUSD for interoperability and content workflows across tools.

This allows organisations to generate pack shots, lifestyle imagery, configurable 3D experiences and virtual try-ons.

Adobe will also apply Nvidia technologies to specific product capabilities.

In Adobe Acrobat, Nvidia Nemotron models will support document intelligence.

In Frame.io, Nvidia CUDA will be used for media processing, search and workflow automation across image, video and 3D content.

The companies will also develop a joint go-to-market approach for enterprise customers.

Nvidia will provide engineering support and access to its AI infrastructure, models and software as part of the partnership.