Watching Stephen Colbert record a speech from his bathtub at home was actually consoling during the hardest months of the outbreak. In a disorganized living room, Jimmy Fallon played the guitar. James Corden used whatever area he had to do anything. For those who had very little normal left, the shows’ continuity—familiar faces, familiar formats, and the tacit promise that the evening would conclude with some sort of normal—meant something. The shows were scrappy and weird, but they persisted.
That flexibility is genuine. A crisis that shut down live entertainment, interfered with supply chains, and drastically changed how people spend their evenings did not affect late-night television. For a while, the genre’s ability to bend without breaking seemed to be evidence of its durability. That tale is not conveyed by the data.
The Late Show with Stephen Colbert, which leads all major network shows in late-night ratings, has been costing CBS tens of millions of dollars per year. The shows are based on a legacy cost model that was created for a time when thirty-second linear advertising slots sold for enough money to justify everything. This model includes large writing staffs, multiple executive producers, guest booking infrastructure, studio overhead, and full orchestra budgets. There will be no return to that time.
It is easy to explain what transpired with the audience. Late at night, people stopped watching linear television. The following morning, if they watch anything at all, they watch YouTube videos. Instead of turning on the TV at 11 PM, they listen to podcasts while cooking or on a commute that no longer exists.
The network shows have responded to this by focusing on creating clips; a memorable celebrity interview or a strong monologue segment is packaged for social media and occasionally receives millions of views. However, compared to broadcast, the ad prices on those platforms are significantly lower. In order to produce content that makes a fraction of what the broadcast model needed to break even, the shows are essentially supporting large-budget projects.
This was made more difficult by the political change. After 2016, there was a noticeable shift in late-night television toward political commentary that was mostly on one extreme of the spectrum. Although the genre had always had a political edge (David Letterman was not impartial, Johnny Carson took jabs), the post-2016 shows leaned into it differently and more consistently, leaving less space for the kind of broad appeal that allowed those earlier hosts to hold an audience that spanned demographics and viewpoints.
Those who had already consented were amused. Those who didn’t began looking for alternative content to watch. The audience became increasingly like to a constituency, which is both commercially constrained and politically cohesive.
The era of podcasts has also made a format question inevitable. For three hours, Joe Rogan sits across from a guest and records a conversation that millions of people can listen to whenever they want. Production value is not what makes it appealing. It’s the lack of restrictions that squeeze each interview into a five-minute promotional slot, such as time limits, format rules, and commercial interruptions. In contrast, late-night interviews are organized more for the format’s requirements than for the conversation itself. Visitors are aware that they have seven minutes. The hosts are aware that they must answer a particular query. The entire production is effective in ways that viewers have begun to see as somewhat phony.
All of this does not imply that the format vanishes smoothly. Next quarter, CBS, NBC, and ABC will continue to air their late-night shows. The programs continue to create brand associations, attract viewers that some advertisers value, and continue to be considered prestige programming in the traditional television sense. However, the economic rationale that gave rise to the genre—expensive production, premium ad revenues, a large audience, and a nightly habit—has been under constant strain for long enough that the question is no more if it will change, but rather how much collapse is required before it does.
