1 NVIDIA published the GPU stack under the agentic ad stack
Ahead of Cannes Lions, NVIDIA detailed how its partners are running advertising on GPUs — with Alembic powering causal marketing measurement on DGX Vera Rubin NVL72 SuperPODs, AWS using NVIDIA Triton Inference Server for real-time bidding, Criteo on cuEmbed/Blackwell GPUs claiming a 2x training speedup and roughly 17,000 GPU hours freed per year, HiggsField using the NVIDIA Agent Toolkit (NemoClaw, OpenShell) for full-lifecycle marketing automation, and KERV.ai using Nemotron 3 Nano Omni for video content intelligence at 10x+ speed.
NVIDIA is doing what NVIDIA always does in a new vertical: publish the reference architecture before the vertical knows it needed one. For marketers, the headline isn't the silicon — it is that the agentic ad stack now has a documented, GPU-backed performance ceiling. The cost of running an autonomous bid agent or a measurement model just dropped by hours of compute, and the brands and agencies whose vendors are not on that stack are buying older economics. Ask your DSP and your measurement vendor where they sit on this; the answer is now a cost-of-goods question, not a tech question.
2 TikTok teased Symphony Agent — cultural intelligence as a buying signal
TikTok previewed Symphony Agent, an AI tool built with what it calls "cultural intelligence" that combines advertiser goals with platform insights, performance signals, and trends to create TikTok-first campaigns, source creator content, and identify relevant creators. The pitch goes well beyond auto-generated copy: Symphony is positioned to take a brief and produce a culturally fluent set of creator-led concepts on the platform with the most velatile trend cycle in marketing.
The most underrated phrase in this announcement is "cultural intelligence." TikTok is signaling that the agent is trained on the platform-specific signal layer — sounds, formats, micro-trends, regional usage — not just brand briefs. Generic AI creative loses on TikTok specifically because the platform punishes generic creative; an agent that learns the platform's grammar is a different product. The corollary for brands: a Meta-shaped asset run through TikTok Symphony is not a TikTok asset. Treat the platforms as different operating systems, not interchangeable surfaces.
3 OpenAI's ChatGPT ad terms went live — the rulebook arrived before the rollout
On the eve of its first Cannes appearance, OpenAI's Ad Tools Terms defining Audience Tools (first-party data upload) and Creative Tools (AI-generated ads) inside ChatGPT were in force, though neither tool was confirmed live broadly. Sponsored links sit live in seven countries — US, UK, Canada, Australia, New Zealand, Japan, South Korea — with Brazil and Mexico queued and the strongest verticals being travel, retail, health, beauty, and financial services. ChatGPT crossed 900M weekly users, with roughly 20% of queries showing direct commercial intent.
The rulebook before the rollout is the tell. Publishing audience-matching and AI-creative policy is how a platform tells the market it intends to compete on the same primitives as Meta and Google, and the cleanest first-mover signal is for brands to start preparing first-party data for upload and modular creative for generative pipelines. Waiting until the buttons turn on means waiting through someone else's case study. The hard work is unglamorous: data hygiene and creative ops. Start now.
4 Smartly built Synapse — an AI memory layer across $7B in spend
Smartly previewed Smartly Synapse, an AI orchestration and memory layer behind every agent on the platform, connecting campaign data, business context, and performance outcomes across its 800-brand client base and $7B+ in managed ad spend. The pitch: agents that remember what worked for your brand last month, not agents that start every prompt fresh.
Memory is the unsexy ingredient that decides whether agentic marketing actually works. Without it, every agent restarts from zero on every campaign — which is fine for one-off generation and terrible for ongoing optimization. Synapse is Smartly admitting that the next year of differentiation in agentic ad tech is not "more agents" but "agents that learn." Watch this concept replicate: any platform that does not have a memory story by Q4 is going to look generic in pitches against ones that do.
5 Cannes put creators on the main stage — 250+ at the Palais Beach
The Cannes Lions program for the week ahead integrated LIONS Creators into the main festival footprint at the Palais Beach, with more than 250 creators walking the Croisette — the most ever and a structural change from prior years where creator programming sat in side rooms. Meta announced a unified Creator Marketing Hub (consolidating Creator Marketplace and Partnership Ads Hub into one destination), and TikTok's pre-Cannes briefings centered on creator-AI tooling.
Creators stopped being a side conversation at Cannes because they stopped being a side line item in client budgets. The serious question for brands going into H2 is no longer "should we work with creators" but "do we have the infrastructure to manage hundreds of partnerships, briefs, rights, and FTC compliance at the speed AI now allows us to scale to?" Most brands do not. The brands that build the ops layer this year will outrun the brands that just keep buying more individual creators.
6 Adobe set up its CX Enterprise partnership rollout for Monday
Adobe spent the weekend setting the stage for its June 22 announcement of CX Enterprise partnerships with Accenture, Omnicom, Stagwell, WPP, Anthropic, and Microsoft — including general availability of Adobe CX skills and MCP servers inside Anthropic's Claude Enterprise and Microsoft 365 Copilot. The pitch positions agency holding companies, not just IT departments, as the operators of Adobe's agentic stack.
Reading the partners list is reading the playbook: Adobe is shipping its CX capabilities into the AI environments enterprise marketers already use. That means the agency that wins your business in H2 will be measured partly on how well they operate someone else's agent stack on top of your data. If you are an enterprise CMO, the question to ask agencies in Q3 RFPs: which CX vendor's MCP do you operate, and how are you measured on outcomes inside it?
7 Reddit's "Max Campaigns" lined up the AI media-buying play
Reddit confirmed it would unveil its Community Intelligence-powered ad products at Cannes — including Max Campaigns, its automated, AI-driven media-buying offering aimed directly at Meta Advantage+ and Google Performance Max. The lineup also includes a free-form ad generator (Beta), tailored creative for Max Campaigns (Beta), Redditor Highlights (GA), and Shopping Listing Ads (Alpha — Reddit's first multi-advertiser ad format).
The "Max" naming is not subtle. Reddit is telling buyers: the AI media-buying loop you have on Meta and Google now has a Reddit equivalent, with audience signal sourced from 23+ billion posts and comments rather than behavioral graphs. For DTC brands and community-led categories (fitness, finance, gaming, indie tech), this is a real shopping list. For others, it is an experimentation budget. Either way, the third "automated buying" surface is now in the mix.
8 The creator economy hit the AI tension head-on
Creator-economy operators heading into Cannes flagged the central tension explicitly: AI-generated content is cheaper, faster, and increasingly competent — but creators remain the most reliable source of cultural authenticity. The platforms reinforcing this: TikTok with Symphony for creator-led campaigns, Meta with Creator Marketing Hub and Brand Memory for brand-consistent generation, and 27 creator-economy voices surveyed by NetInfluencer naming "AI tension" as the festival's defining throughline.
The honest reading is that AI does not replace creators; it changes the unit economics. Brands can now afford to work with more creators at smaller average spend because AI handles the production, distribution, and measurement scaffolding. The creators who win will be the ones whose taste, voice, and audience trust cannot be auto-generated. The brands that win will be the ones that pair AI's scale with human selection. As we will see across the Cannes week, the platforms that figure out how to broker this trade will own more of next year's budgets than the ones that just push more AI tooling.