At Cannes Lions 2026, the global advertising industry conference running June 22–26 in Cannes, France, NVIDIA and its ad tech partners are unveiling enterprise AI tools for advertising and marketing workflows, including Criteo’s 2x faster model training on Blackwell GPUs and Alembic’s causal AI decision stack. The solutions address a core industry pain point: legacy ad infrastructure still relies on manual rules that cannot scale to the volume of real-time ad auctions or the complexity of cross-channel marketing measurement. The full partner suite eliminates the need for ad tech vendors and enterprise marketing teams to build custom AI infrastructure from scratch.
NVIDIA AI Advertising Marketing Tools Power Causal Decision Intelligence at Cannes Lions
Alembic, a causal AI platform for marketing measurement, will be the first company in its category to deploy NVIDIA DGX Vera Rubin SuperPODs for enterprise-scale causal modeling, per the partners’ joint announcement released for Cannes Lions 1. Unlike correlation-based attribution tools that only report historical past performance, Alembic’s stack models true causation simultaneously across every channel, market, and audience to quantify which specific marketing initiatives drive business growth and where capital is being wasted. All inference runs on private supercomputing infrastructure housed in Equinix data centers where enterprise customer data already resides, eliminating the need for data to leave private enterprise environments for processing, a key requirement for regulated industry customers.
World Wide Technology is extending deployment of the stack to regulated, secure environments. The tool is purpose-built for C-suite executives and data leaders accountable for marketing capital allocation.
Real-Time Bidding and Creative Optimization Gains GPU Speedups
For real-time ad bidding, AWS has released a production-ready reference implementation for adtech vendors to run AI-powered bid optimization directly inside live ad auctions, powered by NVIDIA Triton Inference Server 1. The inference server is built to meet the sub-second latency requirements of real-time bidding windows, which often close in less than 100 milliseconds. This lets demand-side platforms, supply-side platforms, and independent ad tech vendors replace rigid rules-based decisioning with AI models for bid price optimization, audience activation, and deal scoring without retraining existing models, reducing time-to-deploy for AI-powered bidding stacks by eliminating custom infrastructure buildouts.
Criteo, which operates one of digital advertising’s largest product recommendation networks, reported cutting its model training time by approximately 50% when running workloads on NVIDIA Blackwell GPUs, a performance gain enabled by the NVIDIA cuEmbed open library 1. The efficiency improvement frees an estimated 17,000 GPU hours annually for the company, with additional scaling work currently in progress.
Taboola is using NVIDIA GPU infrastructure to power its DeeperDive AI answer engine, and is extending that same infrastructure to third-party AI platforms and chatbots to support native advertising revenue streams.
Agentic AI Automates End-to-End Marketing Workflows
For full-funnel marketing automation, Higgsfield AI’s Supercomputer platform uses NVIDIA Agent Toolkit software, including NVIDIA Nemotron open models, to power specialized subagents that manage the entire marketing lifecycle in a single interface 1. The agents run continuously inside every active campaign with no manual intervention, handling end-to-end tasks from campaign ideation and planning, to creative production via integrated image, audio, and video generation models, to automated posting and ongoing performance optimization. The platform orchestrates more than 35 image, audio, and video generation models alongside leading large language models, including Higgsfield’s custom-built Soul and Soul 2.0 large language models that run on NVIDIA Blackwell GPU architecture, and integrates NVIDIA NemoClaw blueprints and the OpenShell secure runtime to add enterprise-grade trust controls including safety guardrails, full auditability, and role-based permissioning required for regulated brand use cases. The platform is already used to create marketing campaigns for nearly 400 Fortune 500 companies.
Contextual Content Intelligence Scales for Ad Targeting
For contextual ad targeting, KERV.ai’s Moment Match Engine uses NVIDIA’s multimodal AI stack to evaluate signals across every frame of video and individual media assets to identify scenes, objects, and embedded products 1. The engine matches ad creative to relevant content based on the visual and textual elements of both the ad and the surrounding media, driving improved user engagement for targeted ad placements.
Frequently Asked Questions
What AI ad tech is NVIDIA showcasing at Cannes Lions?
NVIDIA and its ad tech partners are demonstrating enterprise-grade AI tools for full-funnel advertising and marketing workflows at Cannes Lions, the global ad industry conference running June 22–26 in Cannes, France. The suite covers causal ROI measurement, real-time bidding, agentic campaign automation, and contextual ad targeting, all built to eliminate the need for custom AI infrastructure builds for ad tech vendors and enterprise marketing teams.
How much faster is Criteo’s model training on NVIDIA Blackwell GPUs?
Criteo, operator of one of digital advertising’s largest product recommendation networks, reported cutting its model training time by approximately 50% when running workloads on NVIDIA Blackwell GPUs, a gain enabled by the NVIDIA cuEmbed open library. The efficiency improvement frees an estimated 17,000 GPU hours annually for the company, with additional scaling work in progress.
What is Alembic’s causal AI stack for marketing measurement?
Alembic, a causal AI marketing measurement platform, is the first in its category to deploy NVIDIA DGX Vera Rubin SuperPODs for enterprise-scale causal modeling. Unlike correlation-based attribution tools that only report past performance, its stack models true causation across every channel, market, and audience to quantify which marketing initiatives drive business growth and where capital is being wasted, with all inference run on private Equinix data center infrastructure where enterprise customer data already resides.
How does NVIDIA’s AI stack support real-time bidding for ad tech?
AWS has released a production-ready reference implementation for adtech vendors to run AI-powered bid optimization directly inside live ad auctions, powered by NVIDIA Triton Inference Server. The server meets sub-second latency requirements for real-time bidding windows that often close in less than 100 milliseconds, letting demand-side platforms, supply-side platforms, and independent ad tech vendors replace rigid rules-based decisioning with AI models for bid price optimization, audience activation, and deal scoring without retraining existing models.
Bottom line: Ad tech operators and enterprise marketing teams can access production-ready, GPU-accelerated AI tools for the full campaign lifecycle via NVIDIA’s partner ecosystem, with verified performance gains including 2x faster model training for Criteo’s recommendation systems, sub-100ms inference latency for real-time bidding via AWS’s reference implementation, and enterprise-scale causal ROI measurement via Alembic’s DGX Vera Rubin SuperPOD deployment, with no requirement to build or maintain custom AI infrastructure from scratch.
