AI moves to the supply side: PubMatic and hyundai redefine programmatic with containerized decisioning

PubMatic's launch of Decision Fabric on AgenticOS marks a pivotal innovation by embedding partner AI decisioning models natively within the programmatic supply path, enabling real-time qualification on richer, full-signal data that is typically compressed or delayed before reaching DSPs. This real-time intelligence layer allows DSPs and algorithm companies to deploy proprietary AI models as containerized agents directly inside the supply-side auction infrastructure, significantly enhancing targeting efficiency and audience qualification by leveraging extensive, privacy-safe datasets drawn from PubMatic's exchange-wide bidstream.

This architectural shift addresses a longstanding flaw in programmatic advertising where buy-side decisioning operates on incomplete and filtered data, resulting in wasted spend and missed opportunities. By enabling AI models to run within the SSP environment, Decision Fabric removes constraints on queries per second and grants access to additional contextual signals—such as page ad counts and video ad sound status—not typically passed in OpenRTB packets, thereby allowing more nuanced and accurate impression evaluation.

Beyond technological enhancements, Decision Fabric exemplifies a broader industry trend toward containerized AI decisioning within programmatic auctions, with early pilots by Zillow and standardization efforts like the IAB Tech Lab’s Agentic RTB Framework paving the way. This native integration fosters collaborative intelligence between buy-side and sell-side agents, enabling both to contribute predictive insights that improve bid pricing accuracy and outcome prediction, a capability previously hindered by fragmented visibility into supply- and demand-side data.

Owning proprietary cloud infrastructure gives PubMatic a compounding competitive advantage by enabling faster AI feedback loops and maintaining control over model performance and advertising activity. As CEO Rajeev Goel emphasizes, this infrastructure ownership is critical for sustaining innovation, allowing PubMatic’s SSP to participate meaningfully in outcome prediction through enriched supply-side data such as SDK-level behavioral signals and session context—data that elevates SSPs from mere transaction executors to active, predictive agents within the programmatic ecosystem.

Containerization is fundamentally reshaping programmatic advertising by relocating AI decisioning closer to the supply path, as demonstrated by PubMatic’s Decision Fabric and Index Exchange’s Index Cloud. This architectural shift reduces latency and infrastructure costs while enhancing signal fidelity, enabling advertisers and agencies to integrate more directly with SSPs and maintain greater control over data flows and execution without surrendering decisioning logic. As an industry executive put it, containerization is "a way to get closer to the metal," blurring the traditional SSP-DSP divide and offering a technical pathway to operate nearer to supply.

The competitive landscape is increasingly defined by how ad tech firms approach cloud infrastructure ownership, with companies like PubMatic opting to run their own data center servers to secure a "compounding advantage," as CEO Rajeev Goel explains. This strategy accelerates AI feedback loops and attracts more advertising activity by hosting native partner AI models within SSP infrastructure, reducing reliance on costly public cloud bandwidth. However, public cloud remains indispensable for handling unpredictable demand spikes during events like the cricket World Cup, underscoring a hybrid model where firms balance cost control and operational flexibility.

Beyond infrastructure, the industry is transitioning from a transactional marketplace to an integrated, outcome-driven ecosystem where platforms owning direct supply and SDKs gain a decisive edge. By eliminating the 'Middleman Margin' and leveraging predictive, native AI models enriched with supply-side data, these platforms behave less like brokers and more like operating systems optimized for business results rather than mere media metrics. This evolution, highlighted in recent analyses, signals a competitive dynamic where control over end-to-end environments accelerates learning and execution efficiency, fundamentally redefining programmatic advertising’s value proposition.

Hyundai's pilot with Chalice AI and SSP OpenX showcases the tangible benefits of containerized AI decisioning by enabling SSPs to inject rich, real-time contextual signals—such as in-app purchase events and sustained engagement with product comparison content—directly into auctions. This enriched data layer, bolstered by proprietary SDK integrations and survey-validated intent signals, empowers SSPs to meaningfully contribute to outcome prediction and audience qualification, moving beyond traditional bidstream limitations to enhance targeting precision.

A key advantage of Hyundai’s approach lies in the ownership and portability of its custom AI bidding models through Chalice’s containerized technology, which allows these models to be seamlessly deployed across multiple SSPs or custom bidders. As Hyundai CMO Sean Gilipin emphasized, this capability not only provides a competitive edge but also shifts the focus from chasing the lowest CPM to identifying the most valuable pools of potential car buyers, underscoring a strategic pivot toward precision targeting over cost minimization.

The containerized AI framework fosters unprecedented collaboration between buy-side and sell-side agents by hosting advertiser AI within SSP platforms, encouraging publishers and SSPs to share richer, more granular data. This agentic partnership model enables both sides to jointly predict auction outcomes and customize inventory valuation, as Gilipin noted, aiming to empower brands with bespoke control over their media investments and to unlock new efficiencies through shared intelligence.