Crooked Media's Jon Favreau: Billionaires Bought the Media and Still Lost
Jon Favreau tells Decoder that billionaires' capture of X, Meta and CBS could not beat voter anger over prices. Crooked Media now counts 65,000 paid subscribers, 16 shows and 100 staff.

Why it matters
- Crooked Media has 65,000 paid subscribers, 16 active podcasts and roughly 100 employees.
- 41 of 43 data center moratoriums passed nationwide in the 2026 cycle.
- Vote Save America recruited 2,000 listeners to run for office this cycle.
- Favreau says AI regulation must arrive before 2028 or 'the horse will have well left the barn.'
- Crooked Con drew over 3,000 attendees after the election.
By Rebecca Stone
Democrats swept the 2026 midterms weeks after tech billionaires spent years and billions acquiring the country's major information platforms, and Crooked Media co-founder Jon Favreau says the election proved a simple rule: money cannot outrun material conditions.
Favreau made the argument on The Verge's Decoder podcast with host Nilay Patel. Favreau co-hosts Pod Save America, hosts Offline, and co-founded Crooked Media, the company he started with fellow Obama staffers Tommy Vietor and Jon Lovett in 2017. Crooked now runs 16 podcasts, employs about 100 people, and has crossed 65,000 paid subscribers.
Why did the midterms break against the platform owners?
Patel opened with the central paradox of the cycle: Elon Musk bought Twitter and turned it into X, Mark Zuckerberg tore down Meta's content moderation infrastructure, the Ellison family took stakes in TikTok and CBS News, and Skydance integrated Warner — yet the electorate swung left anyway.
Favreau rejected the framing that media capture decided anything. "I think it's a repudiation of a government that came into power with the chief promise of bringing inflation down and making life more affordable, and it didn't do that," he said.
He traced the pattern back decades. Voters punished Obama's party in 2010 because the recession outlasted the recovery legislation, punished Biden in 2024 because people did not feel the statistical recovery, and punished Trump in 2026 because gas, groceries and housing stayed expensive. "People's lived experiences end up being more salient than the messages that they get from politics," Favreau said.
Patel conceded he had gotten the story wrong himself. He had covered the Trump administration's consolidation of cultural institutions — the Oracle-Ellison TikTok deal, the CBS takeover, FCC Chairman Brendan Carr's pressure on comedians — as a mechanism for permanent political advantage. Bannon's doctrine that "politics is downstream of culture" appeared to be operating.
Favreau's answer: "Politics is downstream from culture, but material conditions rule out."
What does AI have to do with the realignment?
Two issues, Patel argued, show Republicans and Democrats converging against the tech industry: surveillance and data centers. He cited the Flock Safety CEO's hard-right pivot during the cycle — warning that "the radical leftists want a crime world" — which failed with voters.
Favreau linked surveillance backlash to Trump's immigration politics. The same voters who wanted a more secure border, he said, recoiled at "masked, armed federal agents coming into my community and kidnapping people." Privacy, like affordability, is an issue people feel personally.
On data centers, Patel noted that 41 of 43 data center moratoriums passed nationwide this cycle. Favreau called it classic NIMBYism plus something larger: communities discovering projects wrapped in NDAs, negotiated in secret, and imposed over local objections. "My electricity bills are going up. It's fucking noisy," Favreau said of residents near the facilities. He criticized Meta's Louisiana data center and Kevin O'Leary's Utah project for arriving without community negotiation — Patel reported O'Leary's lawyers threatened The Verge with legal action over its reporting.
The deeper problem, Favreau said, is exhausted trust. Elites promised social media and globalization would benefit everyone. On AI, the public hears the same pitch. "No, I don't trust you anymore," is the response, he said.
What does Crooked Media actually look like as a business?
Favreau laid out the structure. The company started in 2016 as a single podcast and launched in 2017. In 2022 it hired Lucinda Treat as its first CEO, backed by investment reported by Variety at the time including George Soros's fund. Under Treat sit a head of news and politics and a head of culture and entertainment, plus standard finance, marketing and people functions.
Vote Save America, the advocacy arm headquartered in Washington, raised a couple million dollars this cycle and recruited 2,000 listeners to run for office; a number of them won. Crooked recently launched Still Counting with Nate Silver, Clare Malone and Galen Druke — a reboot of FiveThirtyEight, which ABC News shut down — giving the company a New York presence. Alex Wagner's Runaway Country joined the roster. Crooked Con drew over 3,000 attendees after the election.
Podcast advertising remains the revenue backbone, Favreau said, because it is "reliable and sustainable." YouTube monetization is catching up. Subscriptions hedge against ad-market fickleness. Every show is now a video podcast, and the company is betting heavily on YouTube growth even while acknowledging the dependency. "The fact that we're all dependent on a YouTube algorithm is not great," Favreau said. "The only solution for that is just to be on as many platforms as possible."
He ruled out a Kalshi or betting-market integration for Pod Save America and Offline, arguing political prediction markets mostly reflect conventional wisdom and caught up with the modelers after Election Day rather than out-predicting them.
The political mission constrains revenue. "There are a lot of advertisers that we won't work with. There are a lot of advertisers who won't work with us because of our political affiliation," Favreau said. The founders review potentially controversial advertisers personally.
Can Congress regulate AI before 2028?
Favreau, who has turned Offline into something close to an AI policy show this year, closed with his sharpest warning. He wants oversight that "finally puts some guardrails or restrictions on the development of super intelligence, AGI, recursive self-improvement," not "show trials where we all get to dunk on Mark Zuckerberg."
"By the time we get to 2028, it feels like the horse will have well left the barn," he said. "It's already left the barn."
He faulted both parties for shallow understanding — "safeguards" could mean chatbot risks to children or loss of human control of AI — and identified cybersecurity, biology and recursive self-improvement as the areas demanding bipartisan work. He accused Trump of answering every AI-risk concern with "China" while his administration lifted export controls on chips. Even without legislation, Favreau argued, the federal government should use its convening power to force coordination among the AI giants, paired with detailed diplomacy with China rather than photo-op summits.
He also pressed AI safety researchers to do more than post alarm on X. "If you're just going to scare people without providing a path to how to fix the problem or how to rein it in, then eventually, it's just going to be noise," he said.
Favreau pointed to a post-election All In podcast in which Chamath Palihapitiya admitted the wealthy did well and the public did not follow — a realization, Favreau said, that arrived years late. Silicon Valley, he argued, still treats politics as an engineering problem, the theory behind DOGE, when politics is "problems of the human condition."
With the 2028 primary already effectively underway — Favreau named Cruz, DeSantis, Rubio and Vance on the Republican side and "a million people" on the Democratic side — he predicted Congress will matter less than the campaigns, and that government shutdowns will become the Democrats' main lever until then.
Source: theverge.com
More from Rebecca Stone
Show full bio
Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
802 articles