This Week in AI: Ads, Agents, Jobs, and the Cost of Scale

Business Insider reported that Amazon has begun buying ads on ChatGPT. The ads send users back to Amazon's own storefront instead of letting AI shopping agents aggregate Amazon's product, pricing, and inventory data.

OpenAI has been building a larger advertising business around ChatGPT, including a self-serve ad platform. Amazon's move is useful because it shows both sides of the opportunity: ChatGPT can become a discovery surface, but retailers still want to protect their own data and transaction layer.

For marketers, this is the early shape of AI commerce. Answer engines may become new ad inventory, but brands and retailers will fight to keep the customer relationship. The battle is not only for the ad click. It is for who owns the product data, the recommendation moment, and the final purchase path.

At Cannes Lions, Meta positioned its AI-powered marketing tools as a broader system for ad creation, testing, launch, collaboration, translation, brand memory, and creator-brand work. Coverage also noted Meta's claim that advertisers are seeing a $4.13 revenue return for each $1 spent, up 25% since 2022.

The exact ROI number should be read as Meta's own platform framing, not independent proof. The more important signal is product direction: Meta wants AI to manage more of the ad operating system, not only generate copy or images.

The marketer's role shifts when the platform handles more creation, testing, and optimization. The higher-value work becomes context: product data, brand rules, creative inputs, audience understanding, approval judgment, and knowing when platform optimization is helping or flattening the work.

Walmart announced it would acquire connected-TV adtech company Vibe.co. Business Insider reported that the deal became a major Cannes talking point and framed it as Walmart's push to make streaming TV buying simpler and more accountable for advertisers, especially smaller and mid-sized businesses.

Walmart already expanded its media ambitions through the Vizio acquisition. Vibe gives Walmart another piece of the stack: connected-TV buying that can connect reach, commerce data, and sales outcomes.

Retail media is moving beyond sponsored listings and marketplace placements. If Walmart can connect CTV exposure to commerce outcomes, more brands will expect TV-style media to behave like performance media.

Business Insider reported that more than 250 creators were expected at Cannes Lions and that creator strategy is now becoming a central marketing topic. The same week, Cannes conversations also circled around agentic AI, M&A, live sports, and creative effectiveness.

This is softer than a product launch, but it matters because it shows where marketing attention is moving. As AI makes generic content cheaper, creators become more valuable as trust, taste, feedback, and distribution layers.

Brands cannot respond to AI content abundance only by making more content. They need credible people, communities, and cultural context. Creator commerce points to a marketing model where creators shape product insight, audience understanding, and campaign direction, not only paid posts.

Axios reported on a new OpenAI, Columbia, Duke, and University of Pennsylvania report about Codex usage. In a sample of individual Codex users, 80.6% made at least one request estimated to represent more than 30 minutes of work by an experienced human, and 25.6% had delegated work estimated to take more than eight hours.

The thresholds are model-estimated and based on a small opted-in sample, so the numbers should not be read as universal productivity proof. But they are useful for understanding the direction of agent adoption.

The agent conversation is moving from whether a tool can perform a task to whether people are willing to delegate meaningful work units. That changes planning, review, staffing, accountability, and how teams define completed work.

Axios reported that former Commerce Secretary Gina Raimondo and former Indiana Governor Eric Holcomb launched Raise Us, a $500 million effort backed by Anthropic, OpenAI's Foundation, Amazon, IBM, Microsoft, Bank of America, Eli Lilly, state governments, educators, and philanthropies. The initiative will test ideas such as wage insurance, retraining incentives, AI-powered career coaching, and short-term credentials.

The initiative starts with state-level partnerships and a policy lab. Its impact is still unproven, but the participation of major AI and enterprise companies makes it worth watching.

AI labor disruption is no longer only a debate about future risk. The companies building and buying AI are starting to support transition systems. That raises the bar: if AI vendors argue that AI will reshape work, they will also be judged on whether they help make the transition workable.

Axios reported that Nvidia announced a next-generation AI infrastructure design at London Climate Week using liquid cooling warm enough to reduce or eliminate some additional chilling equipment. Nvidia said the coolant can run at 113 degrees Fahrenheit and argued the design could address major data-center water concerns.

The claim is important but narrow. Better cooling can reduce a visible data-center constraint, but water used in electricity generation, local grid stress, construction impact, and total AI demand remain open issues.

AI scaling is increasingly a physical-world problem. Models need chips, chips need cooling, cooling needs infrastructure, infrastructure needs land, power, water, construction, and community permission.