1 Anthropic raises $65B Series H at a $965B valuation — past OpenAI
On May 28, Anthropic closed a $65 billion Series H at a $965 billion post-money valuation — the largest private round in AI history and likely its last before an IPO. The round was co-led by Altimeter, Dragoneer, Greenoaks, and Sequoia, with Blackstone, Fidelity, Baillie Gifford, and DST participating. Anthropic's revenue run rate has reportedly crossed $47 billion, up from the $30B+ disclosed at the Series G in February, and the new valuation pushes it past OpenAI's $852B last mark.
When the largest private AI company starts pricing itself ahead of its biggest rival on revenue, not narrative, the valuation gap stops being a sentiment indicator and becomes a procurement signal. CFOs use these numbers to defend Claude line items inside the enterprise stack. Expect a wave of Anthropic-branded enterprise lock-in announcements through Q3 as the round converts into multi-year compute commitments.
2 Brussels readies a record DMA fine for Google — search self-preferencing
Per Handelsblatt and EU Commission sources reported the week of May 25, the European Commission is preparing a high triple-digit million euro fine against Google for self-preferencing its own services in search results. The investigation has been running since March 2025. The penalty — expected to be announced before the EU summer break — would surpass the €500M fine imposed on Apple last year and become the largest DMA penalty issued so far.
This isn't about a deprecated feature; it's about the precedent. Once Brussels sets a nine-figure baseline for self-preferencing, every retailer media network, marketplace, and ad platform with a "default surface" inherits board-level exposure. Plan now for the parallel universe where your own surface can't out-rank a rival's organic listing — the answer is usually independent measurement, third-party data clean rooms, and a self-audit before regulators audit you.
3 Madhive ships Maverick AI agents + MCP across 50,000 daily campaigns
On May 27, Madhive expanded its Maverick AI platform with a full suite of agentic capabilities and an MCP server that lets LLMs interact directly with the local-first DSP. The Model Context Protocol server exposes campaigns, creatives, audiences, and publisher groups via natural language. Maverick is trained on more than a decade of proprietary data from over 50,000 daily local-media campaigns, with the rollout planned over the next year.
The Big Three DSPs (TTD, DV360, Amazon) get all the agent press, but local TV is where the bulk of US political and SMB dollars actually flow. Putting MCP into a local-first DSP means small advertisers will manage CTV through Claude or ChatGPT before the holdcos build their own. Watch for white-label resellers to repackage Maverick as an "AI media buyer" service for local SMBs — the unit economics get interesting fast.
4 Pattern Group launches Pi — an autonomous ecommerce engine
On May 21, Pattern Group (NASDAQ: PTRN) launched Pattern Intelligence (Pi) at its Accelerate conference. Pi is an AI-powered autonomous ecommerce execution engine that runs active sensors across featured offers, advertising, content, pricing, and inventory; when a sensor fires, action loops act in real time. It's built on 77 trillion+ proprietary data points growing by 800B per week. Pi is available today in the ChatGPT app directory and at pi.pattern.com.
The ChatGPT app directory placement is the lead. Pattern just bypassed Amazon Seller Central and embedded itself as the recommended ecommerce intelligence inside the AI surface where brands actually research. For DTC and Amazon-first brands, the playbook is shifting: get inside the AI tools your customers and partners use to ask "is this product worth it?" — before competitors do.
5 Spotify rebuilds its ad platform around two engines — biddable now ~34%
At Spotify's May 21 Investor Day, Global Head of Ad Product Katie English detailed a rebuilt ad platform organized around two engines: high-impact sponsorships and scaled biddable channels. Biddable now tops one-third of ad revenue, with 68% advertiser growth year-over-year, and the platform now spans music, podcasts, and video for the 483M people on Spotify's free tier. Her summary: "Our ads platform should be built for Spotify, not bolted on Spotify."
Spotify spent 2025 quietly catching up to YouTube's auction architecture, and the Investor Day was the receipts. With biddable past one-third and 68% advertiser growth, the audio-only era of programmatic is over — Spotify is now selling impressions across music, podcasts, and video in a single stack. If you have podcast as a media line item, revisit pricing, bidding strategy, and creative cadence in June planning.
6 Microsoft Ads ships AI Max + Offer Highlights + Audience Generation
Microsoft Advertising piloted AI Max for Search across Bing and Copilot this month, expanding query matching and URL routing with brand controls and term exclusions. Reporting is live on day one. The platform also launched Offer Highlights, ad formats that surface free shipping and other key selling points directly inside AI conversations on Copilot, Edge, and Bing — and opened a closed US/CA pilot for Audience Generation, which translates a natural-language prompt into a complete audience targeting setup.
Microsoft's bet: if Google's AI Mode is the destination, Microsoft makes itself the AI ad surface that already works for SMBs. Audience-by-prompt is the real wedge — it removes the targeting expertise gap that's kept Microsoft Ads on agency Tier-2 plans for years. If you run paid search, test it now; the early-pilot impressions are typically undercapitalized and cheap.
7 DemandBird launches a B2B social platform built around “humans, not AI-generated content”
Portland-based DemandBird publicly launched on May 22 with a clear positioning: "the future of social media belongs to humans, not to AI-generated content." Founded in 2025, the platform combines planning, publishing, approvals, analytics, and client reporting in a single workspace designed for B2B teams and multi-client agencies.
An anti-AI-content position in 2026 is a brand bet, not a roadmap — but it's the cleanest signal yet that B2B social tooling is splitting in two. AI-first platforms for volume; human-first platforms for the orgs (consulting, legal, fintech, regulated industries) whose buyers won't tolerate generated slop. For multi-brand agencies, evaluate this as a second tool for premium client work, not a rip-and-replace.
8 Kovva ships AI agents for ad ops grunt work
Kovva, founded by former PubMatic and DSP executives, launched AI agents this week designed to automate the cross-platform operational tasks that dominate programmatic advertising. The agents handle quality assurance checks, discrepancy investigations, reporting reconciliation, budget recommendations, creative fatigue detection, and client communication drafting — a list that maps onto roughly 60-70% of an AdOps Coordinator's day.
The dirty secret of programmatic is that most ops time isn't strategy or optimization — it's spreadsheet janitor work reconciling numbers across platforms. Kovva's targeting that specifically. If it works, the AdOps Coordinator role becomes either an "agent manager" or it disappears, and holdco CFOs already have this category circled as a 2027 headcount lever.
9 California's CCPA enforcement window is officially open for targeted ads
Three new CCPA rules in effect since January 1 are now being actively enforced by the California Privacy Protection Agency. The headline requirement: businesses must complete documented risk assessments before initiating high-risk activities, including targeted advertising, selling or sharing personal information, and processing sensitive personal data. By late May, multiple compliance vendors were reporting their first enforcement-triggered audits.
The risk-assessment requirement is the part most marketers miss. Every new audience activation in California is now a documented privacy review with retention requirements. Smaller advertisers using Meta Custom Audiences or Google Customer Match can't afford the same compliance loop the holdcos run. Either consolidate vendors and centralize data flow, or hire a privacy ops contractor — there's no third path.
10 Meta’s Andrew Bosworth demoed a working AR glasses prototype on stage
On May 24, Meta CTO Andrew Bosworth showed a working augmented reality glasses prototype in a rare public demo — the company's clearest physical proof point yet that the smartphone-successor pitch is real engineering, not just R&D slides. The unit included an AI display capability. It's still a prototype tier; commercial timing remains unannounced.
Marketers don't need to plan for AR-glasses ad units in 2026. They do need to plan for what happens when smartphone replacement starts looking inevitable — Snap Specs, Meta Orion, Apple Vision iterations. Reach math changes when the screen lives on your face. File this under "test your brand IP in 3D space now, not in 2028" — brands with photo-grade 3D asset libraries (Adobe + NVIDIA digital twins, anyone?) will move fastest when ad units actually ship.
The throughline
The agent build-out kept compounding this week — Anthropic's near-trillion valuation, Madhive's MCP server, Pattern's autonomous engine, Spotify's two-engine ad platform, Microsoft's prompt-driven targeting, Kovva's ops automation. The infrastructure isn't slowing down. But this week also marked the first real counterweight: Brussels prepping a record DMA fine for Google, California actively enforcing CCPA risk-assessment rules on targeted ads, and the first venture-funded platform (DemandBird) positioned explicitly against AI-generated content.
For the operator, the implication is dual-track: keep shipping with agents, but build the audit trail in parallel. The valuations say the agentic stack is real. The regulators say it's no longer a free shot.