PubMatic outlines new Architecture of Advertising Intelligence for evolving ad tech

New Delhi: PubMatic has introduced a new artificial intelligence system called the Architecture of Advertising Intelligence, a three-layer framework designed to strengthen the technical foundation underpinning AI-led decision-making in digital advertising. 

The system, unveiled by co-founder and chief executive Rajeev Goel, reflects the broader shift in the digital economy where architecture, not individual algorithms, is becoming central to the sector’s evolution.

According to the company, the acceleration of AI across industries is exposing long-standing inefficiencies in data quality, speed, and infrastructure. Businesses are increasingly moving from operational expenditure to capital expenditure, reflecting a belief that intelligence cannot reliably develop on rented systems. 

Industry commentary suggests that organisations built on durable, well-designed architecture are more likely to benefit as AI becomes embedded across operations.
Within advertising technology, this shift has become more immediate.

The sector expanded quickly over the past decade but often without equivalent progress in insight, automation, or workflow design. As AI sharpens decision-making across the supply chain, companies are facing renewed pressure to upgrade the systems supporting data processing, optimisation, and campaign delivery.

PubMatic’s three-layer model begins with the infrastructure layer, which covers data-centre systems supporting AI workloads. This includes GPUs, memory, storage and network capacity configured to handle high-volume advertising signals. 

The company reports that next-generation GPU inferencing has delivered up to five-times faster bid responses, reducing auction timeouts by more than 85% and recovering substantial levels of previously lost advertising opportunities. 

GPU-accelerated data pipelines are also shifting systems from batch to stream processing, improving latency and relevance. High-bandwidth GPU clusters are said to process more signals per impression, enabling more precise traffic-shaping decisions.

The second layer focuses on applications, where AI is integrated directly into advertising workflows. When embedded across campaign planning, forecasting, setup, pacing and troubleshooting, the technology reshapes how teams operate. 

PubMatic reports reductions of 87% in campaign setup times and 70% in troubleshooting due to automation, with continuous optimisation becoming feasible at a scale that human teams cannot sustain independently. The company maintains that the intention is not to displace human judgement but to support strategic and creative work by reducing routine operational tasks.

The third layer, the transaction layer, covers the point at which agentic AI systems begin interacting directly with the market. This includes systems capable of verifying, adjusting and optimising advertising outcomes autonomously. 

Industry-wide efforts are underway to support this transition, with emerging protocols such as the Model Context Protocol (MCP) and the Ad Context Protocol (AdCP) establishing frameworks for interoperability, metadata exchange and trust between machine-led systems. 

In such environments, each transaction contributes fresh data back into the system, creating what is described as a self-improving market structure.

Analysts tracking the sector note that prototype demonstrations may generate early interest in AI, but long-term change depends on the strength of the architecture supporting these technologies. 

The companies expected to influence the next phase of ad-tech development are likely to be those investing in scalable, resilient systems rather than seeking short-term novelty. 

According to the commentary supporting the launch, the industry is moving towards a machine-led advertising economy in which strategy remains human-led but execution becomes increasingly autonomous and instantaneous, shifting AI “from hype to habit” and “from potential to proof”.