Bedrock Platform and Index Exchange brought the first containerized demand-side platform bidder live on April 21, 2026 - a configuration that places decisioning logic directly inside the supply-side infrastructure rather than on remote public cloud servers, compressing latency, eliminating QPS throttling, and decoupling infrastructure spend from market access in programmatic buying.

The London-headquartered Bedrock Platform today debuted what both companies describe as the world's first containerized DSP deployment inside Index Exchange's Index Cloud infrastructure. The move puts Bedrock's bidder inside Index's own data centers rather than routing bid requests to Amazon Web Services, Google Cloud Platform, or Microsoft Azure - the public cloud providers that have underpinned programmatic infrastructure for the past decade. According to Index Exchange, the entire container execution window runs in under five milliseconds.

What containerization actually changes

Traditional programmatic architecture works like this: a consumer loads a page, an ad opportunity fires, and a bid request travels across the open internet to public cloud infrastructure where a demand-side platform runs its decisioning logic, then returns a response to the exchange. Every hop adds latency. Every millisecond of latency erodes signal fidelity. And the compute costs of processing bid requests at high volume force DSPs to throttle: they evaluate only a fraction of available impressions rather than the full stream. This constraint - measured in queries per second, or QPS - has historically determined how much of the programmatic market any given DSP can reach.

By running its bidder inside Index Cloud, Bedrock removes that ceiling. According to Damian Naglak, head of engineering at Bedrock Platform, "Throttling existed because processing each bid request - egress, cross-network latency, serialization - carried a real per-request cost. Inside the cluster, that cost drops to near zero, so the bidder can evaluate every impression on a deal, not a sampled slice."

The practical implication is that auction budget shifts from network overhead to decisioning. Heavier AI models can run within the same time window. Signals that the SSP already holds become usable without copying them across the wire. Deal-level guarantees hold because the bidder sees every opportunity rather than a throttled sample.

Index Cloud: owned infrastructure, not rented

Index Cloud is not a wrapper around a hyperscale public cloud provider. According to the announcement from Index Exchange, the platform is built and operated by Index inside its own global data centers - a deliberate architectural choice that shapes what is possible. Because the exchange owns the compute environment, the signals available at the point of transaction are richer, latency is lower, and the overhead costs associated with running outside the auction do not exist.

Partners deploy their logic inside Index Cloud using containers: lightweight, virtualized environments that package and run code in isolation directly inside the auction. Each container is cryptographically signed by the partner and verified by Index Exchange, with access tightly permissioned. DSPs receive decisions, not the intelligence behind them. What runs inside each container is controlled entirely by the partner - a design that preserves competitive confidentiality while enabling co-location.

The platform supports a range of use cases beyond hosted DSP bidders. According to Index Exchange, current capabilities include AI-driven decisioning and optimization, audience and signal enrichment, deal activation, contextual and content classification, and impression valuation. Partners currently building on Index Cloud include Bedrock Platform, Chalice AI, inPowered AI, and Nano Interactive, among others. The platform remains in closed beta with select partners.

Index Exchange was also among the pioneering companies behind the Agentic Real-Time Framework (ARTF), introduced by IAB Tech Lab. IAB Tech Lab released the Agentic RTB Framework version 1.0 for public comment in November 2025, standardizing how containerized agents participate in real-time bidding infrastructure. According to Index Exchange, its live container deployments directly informed ARTF's design. The framework establishes a standardized way for containers to plug into exchange infrastructure consistently and securely across the ecosystem, and Index Cloud is built in alignment with it.

The QPS problem and why it matters

Queries per second constraints have been a structural feature of programmatic advertising for years, not a bug. They exist because processing bid requests costs money. Each inbound request requires compute, egress bandwidth, serialization, and a response back across the network. At the scale that major DSPs operate - processing hundreds of millions of requests daily - that per-request cost becomes the dominant cost driver. The natural response has been to filter: accept only the share of the bid stream that is likely to produce a winning bid, and discard the rest.

The problem is that filtering creates a performance ceiling for AI-driven optimization. If a DSP can only evaluate 20% or 30% of available impressions, its models are trained and operated on a partial view of the market. As Index Exchange CEO Andrew Casale outlined at IAB's 2026 Annual Leadership Meeting, the programmatic industry faces mounting inefficiencies driven by what he characterized as the "cloud tax" - operational costs from external infrastructure dependencies that consume resources and constrain innovation. Moving intelligence closer to the impression was the proposed remedy.

The containerized DSP model resolves this at the infrastructure level rather than the algorithm level. "Scale shouldn't be limited to the largest DSPs," said Shane Shevlin, co-founder and CEO of Bedrock Platform. "By running our bidder closer to the impression, we can participate across a broader share of available opportunities and compete on decisioning and performance rather than infrastructure."

For Bedrock specifically, the move changes the economics of operating at scale. Infrastructure spend and market access have historically moved together in programmatic: the more supply a DSP wants to see, the more compute it must buy. Index Cloud decouples that relationship.

Early testing and agency partners

A group of leading independent agencies and technology platforms are among the first to test the integration. These include InterMedia Group, Bay Street Media, and the advanced marketing platform Navigator. According to the press release from Bedrock Platform, initial testing is already showing higher win rates, more consistent delivery, and near-lossless handling of bid requests.

"This is about enabling AI to actually do its job," said David Nyurenberg, SVP of Digital at InterMedia Group. "In an AI-first world, performance comes down to evaluating and scoring every opportunity in the moment. If you can only see a fraction of the market, your models are inherently constrained. This creates an environment where each impression can be properly assessed and prioritized based on its likelihood to drive an outcome, which is a meaningful shift for performance advertisers."

These early results align with prior evidence from Index Cloud deployments. A case study published by Index Exchange in February 2026 documented a 75% reduction in cost per site visit for a major apparel retailer after inPowered AI deployed a containerized optimization model inside Index's infrastructure. The model scored each impression based on likelihood to generate site visits using patterns from historical DSP performance data, evaluating supply-side signals - time of day, page URL, browser version, operating system, carrier, and zip code - that are typically unavailable to demand-side optimization alone.

What this means for the open internet

Walled gardens - Google, Meta, Amazon - have built structural performance advantages over the past decade by keeping data, compute, and decisioning in the same place, optimizing together. The open internet has competed on transparency, flexibility, and supply diversity, but has accepted a persistent performance gap as part of that trade-off. Infrastructure fragmentation, latency costs, and QPS limitations have all served to widen that gap over time.

As PPC Land has covered extensively, the programmatic landscape has been under sustained pressure from cost and margin dynamics. Research from the Incorporated Society of British Advertisers found that publishers receive only 36 cents of every media dollar spent through demand-side platforms, highlighting systemic inefficiencies. The containerized DSP model addresses a different dimension of this efficiency problem - not which path to route a bid through, but whether the bidder can see the opportunity at all.

"We are at the outset of a structural shift in programmatic," said Andrew Casale, president and CEO of Index Exchange. "When intelligence moves closer to the impression, the constraints that have historically limited scale begin to fall away. What emerges is a more efficient, more performant, and more powerful open internet."

The architecture Bedrock and Index Exchange have implemented does not change how the auction operates or which DSPs are eligible to compete. Bedrock retains full control of its decisioning and bidding logic throughout. Index provides the environment; Bedrock controls what runs inside. This separation of infrastructure and intelligence is a deliberate design choice - one that preserves the independent roles of DSPs, SSPs, agencies, and media owners that define the open programmatic ecosystem.

Live streaming and real-time environments

According to Bedrock Platform, the containerized model is particularly impactful for high-demand, real-time environments such as live streaming, where timing, scale, and infrastructure constraints have historically limited performance. When every millisecond of latency matters and every impression is contested, having decisioning logic run inside the exchange rather than across the internet creates a meaningful operational advantage.

Zillow piloted containerized RTB with Chalice and Index Exchange as early as August 2025, embedding DSP intelligence within SSP infrastructure to access site-level data that standard bid requests do not carry. That pilot demonstrated that containerized approaches could serve buyer-side quality goals - and not just efficiency goals for the platform. The Bedrock deployment extends this logic to the bidder itself, not just the decisioning algorithm.

The industry standards infrastructure is also catching up. IAB Tech Lab formally named its umbrella agentic initiative AAMP - Agentic Advertising Management Protocols - on February 26, 2026, clarifying the relationship between the Agentic Real-Time Framework and other components of the agentic advertising stack. ARTF, which cuts latency by 90% and improves agentic real-time decisioning according to the organization's documentation, provides the technical foundation on which containerized deployments like Index Cloud are built. Amazon's April 15, 2026 donation of its Dynamic Traffic Engine to IAB Tech Lab as an open-source project extends the efficiency framework further, enabling DSPs to signal bidding priorities to SSPs to reduce QPS waste across the supply chain - a problem that the containerized DSP model addresses from the supply-side direction.

Context from the marketing community

For buyers and agencies navigating programmatic infrastructure choices, the Bedrock and Index Exchange deployment illustrates a direction that the market has been moving toward but has rarely executed at the DSP-bidder level. As thespend.net (https://thespend.net) tracks programmatic spending trends, the economics of infrastructure - who pays for compute, who captures the efficiency gains - increasingly shape which platforms can compete at scale.

PPC Land has followed the progression of Index Exchange's infrastructure strategy across multiple moves: from duration-based streaming TV metrics in September 2025, to the Gracenote contextual intelligence integration, to the sell-side decisioning case studies in early 2026. The containerized DSP announcement today represents the most direct application yet of that infrastructure strategy to the demand-side buyer relationship - taking the logic that SSPs should move intelligence closer to the impression and applying it to the bidder itself.

Independent agencies such as InterMedia Group, which does not have the infrastructure budgets of holding company trading desks, now have access through Bedrock to a materially broader share of available supply. That access was previously constrained not by the quality of their models, but by the cost of evaluating enough bid requests to make those models effective. Container-based deployment removes that constraint at the infrastructure level.

Timeline

Summary

Who - Bedrock Platform, a programmatic DSP headquartered in London, and Index Exchange, an independent supply-side platform, together with early testing partners InterMedia Group, Bay Street Media, and the marketing platform Navigator.

What - Bedrock Platform deployed the world's first containerized demand-side platform bidder inside Index Exchange's Index Cloud infrastructure, running Bedrock's full decisioning and bidding logic within Index's proprietary data centers rather than on public cloud servers. The deployment eliminates QPS constraints, reduces per-request compute costs to near zero inside the cluster, and enables Bedrock's AI models to evaluate every available impression on a deal rather than a throttled sample.

When - The announcement was made on April 21, 2026. Index Cloud is currently in closed beta with select partners.

Where - The containerized bidder runs inside Index Exchange's own global data center infrastructure, which is not built on GCP, AWS, or Azure. Bedrock Platform is headquartered in London. InterMedia Group and Bay Street Media are among the first agency testing partners.

Why - Traditional programmatic architecture forces DSPs to throttle inbound bid requests because processing each one across public cloud infrastructure carries real per-request costs in egress, latency, and serialization. This limits how much of the available market any DSP can reach, constraining the effectiveness of AI-driven optimization models that require a complete view of supply. By running the bidder inside the exchange, the cost structure changes fundamentally - near-zero per-request overhead, no QPS ceiling, and access to SSP signals without round-trip data transfer. For independent DSPs like Bedrock, the result is the ability to compete on decisioning quality rather than infrastructure budget.

Share this article
The link has been copied!