# No Editor Should Ever Have to Ask 'Can We Afford This': The NYT CEO on Running a Media Business in the Age of AI
> Odd Lots hosts New York Times CEO Meredith Kopit Levien. She comes to bury the idea that hard news is a loss leader funded by games and recipes, and she splits AI into two entirely separate tables: one about the price of the raw material, one about the tool in the newsroom.
Published: 2026-08-10
Locale: en
Tags: odd-lots, podcast-notes, media, subscriptions, ai-licensing, moats, pricing-power
TL;DR: The idea she wants dead is that hard news is a loss leader funded by games and recipes — at the Times, high-quality independent news journalism is the single most economically value-creating thing they do. And AI gets split into two entirely different tables: one about the price of the raw material, one about the tool in the newsroom.
> *The Master said: "When the multitude detests a man, you must examine it yourself. When the multitude admires a man, you must examine it yourself."*
> —— Confucius, *Analects*, Book XV (5th c. BC)
>
> *The Master said: "To hear something on the road and repeat it on the road is to throw virtue away."*
> —— Confucius, *Analects*, Book XVII
Confucius had no algorithm to complain about, but he covered both halves of it. Whatever the crowd hates and whatever the crowd loves, go and check for yourself. And what you heard on the road and repeated on the road is virtue discarded. This episode is about a company that turned "go and check for yourself" into a profitable business — while the thing you heard on the road has acquired a new name: slop.
## What this episode is about
Odd Lots hosts Meredith Kopit Levien, president and CEO of The New York Times, recorded the day after the company reported earnings.
It opens with Joe Weisenthal's own story. In 2009 he was at Business Insider, his first real journalism job, when someone called and asked whether he'd like to come in and interview at DealBook. His reaction: why would I want to board the Titanic? The stock was six dollars, this was old print media, and he considered himself part of the internet. He passes the story to his guest later in the conversation, and she gives it one line back: what a shame you didn't come, that team is killing it.
Seventeen years on, beyond the obvious — the Times was not the Titanic — one thing has genuinely inverted. Back then the ambition of independent media was to get inside a big brand. Today people inside the big platforms are running the numbers on walking out and starting their own.
**Original episode**: Bloomberg's *Odd Lots*, "NYT CEO Meredith Kopit Levien on Running a Media Brand in the Age of AI," 10 August 2026, with Meredith Kopit Levien, president and CEO of The New York Times.
## The notes I took
**She came specifically to kill the "hard news is a loss leader" story.** The version in circulation: serious journalism is something you do because you have to, for the mission, while the money comes from games and sports. She wants that "put to bed" — the first dollar in the place goes to high-quality independent news journalism, hard stop. And she goes further: writ large, the most economically value-creating thing the company does *is* that journalism. The multi-year civil wars in Myanmar and Sudan, how and why Ebola is spreading differently in the DRC this time — those aren't done for clicks or video views, and they get covered whether or not they'll draw the attention they deserve.
**More persuasive than the declaration is the mechanism: she has removed "can we afford it" from the editorial decision.** Who decides how many people are on the ground in Ukraine — roughly seventy in and out last year, five years into a grinding war? The editors do, and often below the level of executive editor: the international editor, and the people even closer to what is happening. She defines her own job as making the company successful enough that no journalistic leader ever has to say "can we afford to send Declan Walsh into the Ebola zone." That is a concrete piece of organisational design — not a promise to value journalism, but an arrangement in which the question that erodes journalism has no institutional place to be asked.
**Retention isn't a salary argument; it's "you can't make this thing anywhere else."** Several stars left this year — Ross Douthat to 60 Minutes, both Hard Fork co-hosts out — and standing up your own newsroom now takes a few clicks. Her answer counts the whole apparatus: the best editors, lawyers, security people, graphics and photography, plus the thing every outside hire tells her afterwards — publishing the same story at the Times reaches a different order of audience. Jonathan Swan came to work alongside the best White House reporters there were; Ezra Klein hands credit to his editors and the support system around him. Her metaphor is a good one: the people they want are obsessed with both the name on the front of the jersey and the one on the back. Then she adds the honest part — that isn't for everybody, and I accept that it isn't.
**The core of the subscription strategy is that she refused to answer "free or paid."** The reason the Times has the subscriber base it does, she says, is that they never accepted a choice between having a lot of work widely available to sample and keeping enough value and gravity behind the wall. The free side has to stay good enough to plant the seeds for the next million subscribers; the paid side has to keep its weight. On price she reaches for the jargon deliberately: the strategy is meant to get everybody under the demand curve — from an international prospect buying the games product on sale, up to someone on a family plan sharing it with the whole household. The hero product, news plus everything, starts at a dollar a week; you engage, and eventually you hit a step-up to a much higher price — a ladder she thinks holds because the products are inherently getting more valuable every day. She also mentions her niece asking to share her password. Do you know what I do for a living?
**AI's first table is price, and her arithmetic is blunt.** Last year they spent close to two billion dollars producing half a million works of content, and they have been doing some version of that for 175 years, which adds up to an enormous corpus. They have done a deal with Amazon; they are also nearly three years into suing OpenAI and Microsoft, and sued Perplexity last December. Her three conditions for a good deal: it has to be continuous with a strategy where the publisher's business stays sustainable and growing; it has to grant explicit permission with control over all the ways the work gets used; and it has to be a sustainable, fair value exchange. The pressure point is well chosen — these companies spend tens and sometimes hundreds of billions on talent, compute and power, so they should be spending commensurately on what they call data and what is in fact high-quality independent journalism.
**AI's second table is the tool, and not one of her examples is "help me write."** The Epstein files landed on a Friday night, something like three million pages; the newsroom's AI initiatives team built tooling to comb the trove for patterns and clusters, and by the next week they were telling different kinds of stories, faster. The same use covers sweeping a cabinet nominee's full public record efficiently. The most interesting one is the Sydney Sweeney jeans furore: the social-media version was that the left was outraged, and they used AI to show the outrage was first a construction of the right — which is the episode's cold open. Her favourite is smaller: to convey a trend in sport, they used AI to sort through existing aerial imagery of tennis courts and count how many had been converted to pickleball courts. She pauses mid-sentence to ask the hosts if they can guess what's coming.
**On what AI costs and who it replaces, she declines to give a clean answer.** Asked whether token spend has reached the level of a 2027 planning item, she says every CEO is asking, and that anyone who gives you a confident answer right now probably doesn't know yet. The boundary she does draw is sharp: what work should be human-led — high-quality independent journalism, creative work, human-led. On the old newspaper split between reporters and writers, and whether an AI could take the reporter's dictation and generate the piece, she doesn't say she's offended; she defines reporting first. Going out into the world, following a professional process to fairly unearth new information, talking to people, going to the scene — then, with sensitivity and judgment, translating that into what a reader receives. School shootings, she says, need that translation to be human. If technology means they get to do more reporting, she isn't offended at all.
## Where my mind went
**One: this episode points the bottleneck at a layer that's easy to skip — first-hand collection.**
The question I run on any industry is: when demand doubles, which layer breaks first? Applied to the AI content chain, the instinct stops at compute and models. This episode locates it further upstream — the thing the models consume is something someone else paid to go out and dig up.
The intermediate step sits inside her definition of reporting: its cost structure is **acquiring** information, not processing it. What generative AI has collapsed is the processing side — rewriting, summarising, translating, pattern-finding, all of it cheaper. The acquisition side has not moved; the plane ticket into the Ebola zone costs exactly what it did. When some links in a chain see costs collapse and others don't, value concentrates at the links that didn't move. That also explains why the second of her three deal conditions — control over all uses — reads as more important than the money. What's actually for sale isn't the archive; it's the ongoing capacity to produce.
The falsification condition is statable: if AI starts completing the acquisition side on its own — sensors at scale, automated cross-checking of public records, autonomous first-hand verification — this layer loosens. So the thing to watch isn't how well the models write. It's whether the models start showing up at the scene.
**Two: when is "we're investing for the long term" credible? Check the governance first, then check whether anything can be marked to reality.**
Joe puts it directly: the stock sold off after each of the last two quarters. She doesn't dodge the short term; she opens up the two reasons the family control structure exists — to keep the journalism independent and in service to nobody's interest, not even the company's own commercial interest, and to allow the company to play a long game.
That gives a usable filter. When a company claims to be investing for the long term, the credibility precondition is whether its governance **permits it to look bad for a while**. Without that, "long term" is usually just what a bad quarter sounds like. But the honest next step: dual-class structures protect the people who are right and the people who are wrong equally well. So it's a precondition, not a verdict.
The verdict lives a layer down: has the long bet left anything that can be marked against reality? This one has, with direction and a time window — whether video grows into daily engagement at the millions-of-people scale audio did, whether the dollar-a-week entry point keeps converting up the price ladder, what The Athletic's growth curve does. Against which "I am confident to my core that we have great potential to do economic value creation" cannot be wrong, and therefore carries no information.
**Three: where the market is winner-take-most, "the model works" is not "this company works."**
The most valuable counterweight comes in the outro, from the hosts themselves: what the New York Times has proven is the New York Times. Joe pushes to the uncomfortable part — its success is itself part of why everyone else is struggling. It has become something close to the monopsony buyer of great talent in news. The White House reporter at the Dallas Morning News, if they're good, ends up at the Times; and most local outlets no longer staff a White House reporter, because at internet scale readers say they'll just read the Times' White House coverage. The Athletic is that mechanism in concentrate — go into every big sports city and scoop up the best beat reporters.
The investing translation is a valuation discipline: in a market with returns to scale steep enough that it only has room for one, the peer average is meaningless, because the number two isn't running the same business. Using a peer multiple to find a bargain in the number two is reading "the model works" as "this company works."
Tracy's addition is worth keeping too: newspapers always had crossword puzzles. The Kansas City Star had one. So games and recipes and news under one roof isn't an invention — it's the most normal thing in the world. The real question isn't whether the model is new; it's how many one-stop shops the English-speaking world has room for, and how large and how durable that "curious, educated, open-minded" cohort really is — or whether it eventually splinters off toward whatever is more satisfying to each of them.
## Worth a look
- Original episode: Bloomberg's *Odd Lots*, "NYT CEO Meredith Kopit Levien on Running a Media Brand in the Age of AI" (10 August 2026)
- The New York Times Company is publicly listed; subscriber counts, ARPU, the split between subscription and advertising revenue, and quarterly net subscriber adds are all in the public filings and investor decks — the "price ladder is working" claim can be checked directly
- The dual-class share structure and family control arrangement are described in the company's public proxy filings; her two stated reasons are the company's own account, but the structure itself is verifiable
- The suits against OpenAI and Microsoft, and against Perplexity, are public court matters — complaints and responses can be read in full
- Media has no direct equivalent of an external sustainability certification, but several public annual surveys track news trust and willingness to pay (the Reuters Institute Digital News Report among them) for cross-reference
- The two *Analects* passages at the top are a footnote I brought to the episode, not part of it
---
**Disclaimer**: This is a personal set of listening notes and study material, offered as educational content. It is **not investment advice, an offer, or a solicitation**. Companies, people, policies, figures and litigation mentioned come from the public episode and other public sources. Nothing here recommends any security, provides price targets or timing, or forecasts share prices, industry structure or legal outcomes. Investing carries risk; make your own judgment based on your circumstances and risk tolerance, and consult a qualified professional where appropriate. Copyright in the quoted material remains with the original programme — please listen to the original and support the creators.