NVIDIA and partners redefine advertising with AI at Cannes Lions

The digital age gave speed to advertising and marketing; the AI era gives them autonomous operations. At the Cannes Lions 2025 festival (June 22-26, France), NVIDIA and its partners — Alembic, Amazon Web Services (AWS), Criteo, Higgsfield, KERV.ai, and Taboola — show how AI technologies transform creativity and enable faster, more autonomous operations at enterprise scale.

Decision intelligence at enterprise scale

Alembic, a causal AI platform, addresses one of the biggest challenges for businesses: proving which marketing initiatives truly drive growth, not just reporting what happened. Modeling true causality simultaneously across all channels, markets, and audiences requires an AI infrastructure capable of processing huge, changing datasets without reducing them to correlations. According to NVIDIA's blog, Alembic's causal AI models run on the new NVIDIA DGX Vera Rubin NVL72 systems, which deliver up to 30x performance improvement in AI workloads compared to previous generations. This allows Alembic to analyze over 10,000 variables simultaneously, run what-if scenario simulations, and quantify the true return on investment (ROI) of each marketing channel. Alembic will be the first causal AI company to use the NVIDIA DGX Vera Rubin SuperPOD supercomputers, which integrate 72 GPUs per rack and fifth-generation NVLink connectivity, for enterprise-scale causal modeling. Inference runs on private supercomputing infrastructure within Equinix data centers, where enterprise data already resides, keeping AI workloads on-premises. World Wide Technology extends this to secure and regulated environments, complying with regulations like GDPR and CCPA. This approach contrasts with traditional correlation-based attribution models, which explain only 20-30% of variance in outcomes, while causal AI can achieve up to 80% predictive accuracy, according to industry studies.

Smarter bidding at auction speed

For advertisers, serving relevant ads and recommendations among billions of daily transactions requires accurate, fast, and affordable AI. AWS is bringing cloud infrastructure, foundation models, and NVIDIA GPU acceleration to enable real-time agentic bidding. The collaboration allows advertisers to generate and execute dynamic bidding strategies without sacrificing speed, using GPU-optimized containers on AWS. According to AWS data, agentic bidding can reduce cost per acquisition (CPA) by 25% and increase conversion rate by 15% compared to traditional methods. Criteo, a partner in this initiative, already processes over 2.5 billion daily transactions and uses deep learning models to predict purchase intent. The integration with NVIDIA reduces inference latency from milliseconds to microseconds, crucial in programmatic auctions that last less than 100 milliseconds. This advance is compared to the shift from contextual to behavioral advertising in the 2000s, which multiplied campaign efficiency by 10. Now, agentic bidding promises a similar leap, but with self-learning and real-time adaptation capabilities.

AI-powered content and video generation

Higgsfield, an AI-generated video startup, presents its high-quality video generation platform for marketing. According to its press release, the platform can create 30-second videos from text or images in under 5 minutes, at 1080p resolution and 60 fps. It uses diffusion models trained on NVIDIA A100 and H100, and is now migrating to the new B200 to reduce inference costs by 40%. KERV.ai offers interactive video technology that allows viewers to click on objects within videos to learn more or make purchases. Its platform processes over 500 million monthly impressions and has been shown to increase click-through rate (CTR) by 300% in retail campaigns. Taboola, a content discovery platform, uses AI to recommend personalized articles and ads to over 500 million unique users per month. With the integration of NVIDIA Triton Inference Server, Taboola has reduced recommendation latency by 50% and increased revenue per user by 12%. These tools enable brands to create personalized content at scale, a shift similar to what programmatic advertising represented versus manual media buying.

Why it matters

The convergence of causal AI, automated bidding, and content generation marks a turning point. Companies no longer just optimize campaigns; they can predict and attribute results with precision. NVIDIA's infrastructure allows these models to run at scale without compromising speed or data privacy. According to Gartner, by 2027, 60% of marketing enterprises will use causal AI for attribution, up from 15% today. Additionally, the AI video generation market will reach $2.5 billion by 2026, according to MarketsandMarkets. However, reliance on specialized hardware (GPUs) and the cost of foundation models (up to $10 million per training) could limit access to large enterprises, creating a gap with SMEs.

Implications for the industry

Accelerated adoption of causal AI in marketing is expected, replacing correlation-based attribution models. Programmatic bidding will become more efficient, reducing ad waste, which is currently estimated at 30% of total digital ad spend (nearly $200 billion annually). AI video generation will allow brands to create personalized content at scale, but also poses risks of deepfakes and misinformation. Therefore, companies like Higgsfield implement digital watermarks and moderation systems. Competitively, early adopters of these technologies could gain a significant advantage, similar to early users of social media advertising.

What readers should know

NVIDIA doesn't just sell chips; it provides the platform for partners like Alembic and AWS to build vertical solutions. Investment in AI infrastructure is critical to stay competitive: the cost of not adopting could be a loss of up to 20% market share in three years, according to McKinsey. Data privacy remains a challenge, but solutions like local inference in Equinix data centers offer a path. Moreover, the collaboration between NVIDIA and its partners at Cannes Lions 2025 demonstrates that AI is moving from an optimization tool to an engine of strategic autonomy. For marketing professionals, the recommendation is to familiarize themselves with concepts like causal AI and agentic bidding, and evaluate vendors that offer scalable and secure infrastructure.