Are you ready to reach holiday shoppers on CTV? For many streamers, the season is already underway.

Last year, according to Roku internal data, holiday content searches on Roku saw an 86% increase during October. And in November, those searches rise an average of 39% weekly. This surging holiday spirit is mirrored by consumer activity, as about half of Roku streamers shopped Black Friday and Cyber Monday deals in 2025.

Meeting those shoppers is easier than you might think. Begin with the seasonal outcome you hope to drive: awareness, online sales, leads, store visits, or another measurable goal. From there, define your audience, upload your creative, set the budget and campaign dates, and establish measurement before launching. You can even create a shoppable ad unit that lets streamers buy with a click of their remote.

Our holiday playbook offers a month-by-month guide to reaching shoppers during this all-important time of year.

First-time spenders get a $5,000 match on the first $5,000 spent with code BEEHIIV5K. Terms apply.

Launch your campaign today.

Seventy-six percent of voters polled by NBC News on Election Day said they opposed data centers in their own communities. Meanwhile, Big Tech has been reporting another quarter of record spend on CapEx. AI now faces a thorny political/public relations road ahead, along with competitive and engineering ones.

what the Dems’ win means for the tech industry.

The blowout results of the Democratic Party in the midterm elections has created a new operating landscape for the tech industry.  NBC News found that 76% of Election Day respondents opposed having data centers built locally. Voters also elected candidates who favored moratoriums on new facilities, and they backed ballot measures aimed at restricting them, according to The Verge's post-election reporting. Few issues in American politics have drawn agreement on that scale. Whatever else divides these voters, a large majority of them do not want a warehouse full of GPUs down the road. AI has a PR issue.

The Information reports that Scott Wiener, Democrat of California, who won his bid to succeed Nancy Pelosi, is headed to Washington.  One of Wiener’s main issues is AI safety; Gov. Gavin Newsom signed his SB 53, requiring that AI companies write and comply with their own safety policies.  Further, Wiener is an advocate for ensuring that workers displaced by AI are compensated.  He also wants to expand protections for workers inside those companies who choose to serve as whistleblowers on what they see “behind the curtain.” (You may remember that late September/early October, OpenAI fired three AI researchers for alleged misconduct.  One of those AI researchers had served as an internal OpenAI contact with safety organizations like Redwood Research and METR during the OpenAI – Hugging Face incident.   Stronger whistleblower protections may discourage such behavior by frontier AI players.)

This matters because compute is already a constrained resource. Every prompt you send ends up in a building that needs land, electricity, cooling, and the goodwill of the people who live nearby. So far, the AI industry has treated that goodwill as (more or less) a given, something to be handled with tax incentives and nondisclosure agreements. And the Trump administration has been “all-in” on AI, as we race to keep our lead against China. The midterms suggest that the picks and shovels operators of AI will be held more actively accountable for any externalities that the build-out incurs.

Moratoriums and ballot measures work through zoning boards, county commissions, and utility proceedings, not through federal legislation. The effects will therefore show up as delays, higher siting costs, and projects quietly moved to friendlier jurisdictions, not as a single dramatic shutdown. The practical risk for users of AI is a tighter supply of capacity at the moment demand is climbing. Further, this could bring more rate usage limits and higher prices for the heaviest users of AI.

The AI industry will also need to improve its communication approach with the populace.  So far, tech has been very good at explaining what models can do, but less adept at explaining to communities what they get in exchange for brown water (and other externalities). A data center employs relatively few people once it is built, and it consumes a great deal of electricity. Voters can do that arithmetic without help.

Therefore, we may get more usage on smaller models, on-device inference, and software that spends compute only where it earns its keep.

If you build workflows on AI, this is a good time to audit and review them. Route routine drafting, sorting, and summarizing to lighter models, and reserve the heavyweight frontier AI  systems only to the problems that require them. The professionals best positioned for a more constrained compute market will be those who are creative about how they distribute their workflows (which may mean experimenting more with Chinese models and open American models, e.g. those offered by Nvidia).

also on the radar:

Big Tech keeps spending, and the returns have not arrived.

After Amazon, Microsoft, Alphabet, and Meta reported Q3 earnings, The Verge found all four still spending record sums on data centers without yet showing the returns they have promised investors. The money is moving on conviction, which means the tools you use today are subsidized by a bet that has not yet been proven. The ‘AI Bubble” thesis continues…

Uber counted its agent-booked rides, and the number was four.

Uber disclosed that third-party AI agents booked just four rides in the third quarter, plus around 500 Uber Eats orders during Halloween week. Agents may well become a major channel, but this is a useful reminder that capability demos and actual consumer behavior are separated by a wide gap.

Meta buys Conductor to bring agent teams to coding.

Meta acquired the AI coding startup Conductor, whose tool lets engineers deploy a team of agents that work on different parts of a project concurrently. The deal signals reinforces the notion that enterprise is adopting agentic AI at a furious pace, and that coordination among agents is still a challenge that has not been solved.  Notably, back in September, Noam Brown of OpenAI told Dwarkesh Patel that ten thousand humans can still coordinate better than ten thousand agents (we can put that one down as a score for the humans…for now).  Meta’s acquisition indicates that they see AI agent coordination as a frontier challenge, and one that will help determine who wins from our current phase of the AI buildout.

worth your time.

You might consider checking out the Lawrence Berkeley National Laboratory's 2024 United States Data Center Energy Usage Report. It estimated that data centers consumed about 4.4 percent of U.S. electricity in 2023 and projected that share could reach somewhere between 6.7 and 12 percent by 2028. It is a dense government report, but it is the clearest baseline for understanding the numbers behind the local debates over AI. The executive summary alone will sharpen how you read the next headline about a proposed facility.

The human mind is the original generative machine.

The human mind is the original generative machine.