Matt Swanson

Krishna sized the open-source protection TAM in the multiple billions of dollars, describing an AI-driven patching factory and a client-count flywheel with 7,500 packages remediated in three weeks.

Arvind emphasized model-neutral strategy: working with all frontier providers, open-weight models, and producing smaller domain-specific models. IBM helps clients deploy AI to capture value using Confluent for data, watsonx for orchestration, across any infrastructure.

Arvind predicted 50/50 split between public and private/on-premise AI in 3-5 years, with 80-90% of inferencing potentially on-premise given smaller model efficiency. Announced sovereign core offering.

Arvind highlighted watsonx Code Assistant for Z, watsonx Assistant for Z launching in Q3, watsonx.data with DataStax integration. Orchestrate differentiates through heterogeneous agent support across 150+ prebuilt agents.

Arvind said AI conversations are shifting from infrastructure to applications and ROI, citing a mid-sized client expecting 30% savings in back-office finance processes.

Green argued AI does not disrupt the DSP model, it is the model: a DSP decides which of roughly 20 million ad opportunities per second to buy, a job enhanced by the AI TTD has invested in for years. He stressed objectivity and protecting advertisers' data as the moat while competitors prioritize owned-and-operated inventory, and called agentic AI one of the biggest opportunities advertising will ever see.

Green separated cyclical from structural, argued structural drivers extremely strong. When conditions stabilize, provides natural tailwind. Highlighted Audience Unlimited, measurement advances, CTV adoption, retail data partnerships, and agentic AI as execution opportunities that can drive improved growth.

Green rejected brand vs performance distinction. Everything is performance, just at different funnel positions. Problem is last-touch attribution giving all credit to final touchpoint. Working with brands like Hershey's to create better measurement frameworks that reward top-of-funnel brand building alongside performance.

Green explained TTD's deliberate top-down strategy starting with largest advertisers. SMB is in ambitions but not current focus since large advertisers represent nearly half of TAM. Serving tough large clients first makes it easier to simplify for smaller ones later by removing dials/reports.

Green argued Amazon's DSP is primarily a buying tool for Prime Video, not an objective open internet platform. Content owners prefer TTD partnerships because TTD doesn't compete with them.

Anil framed agentic software as Adobe's next platform-leadership opportunity across creativity, productivity, and CX; directional, not quantified.

Anil explained SEMrush's outside-in search/prompt data combines with Adobe content intelligence to build a brand-visibility solution.

Shantanu explained ad platforms value Adobe's ability to deliver campaigns faster and close the attribution loop. Agencies partnering for automation technology. On models, predicted only 2-3 large language models will survive long-term while smaller model companies face viability challenges.

Shantanu emphasized Adobe's unique ability to close the loop from campaign creation through delivery and commerce analytics. Anil added that 10-20% of total marketing budgets are spent on content, and GenStudio helps reduce costs while enabling more agile multichannel content production.

Adam stressed smart, high-talent-density A/B testing accelerated by LLMs over sheer iteration count; more research scientists still needed. Direct.