Games were once the brightest cash line in Bilibili’s listing story.

The Fate/Grand Order co-publish lit the fuse in 2016. National Business Daily later traced a curve: of nearly RMB 25 billion in 2017 revenue, games once contributed more than 80%. Ads and live streaming rose later, and games fell toward roughly a fifth of the mix—but in the “home of young users” narrative, games remained a mainline fight, not a side bet.

In August 2021, a Bilibili showcase unveiled 16 new titles, 6 of them self-developed. Chen Rui said the self-dev team already topped 1,000 people, and that within a few years half of game revenue should come from in-house products. Outlets such as Jiemian covered the ambition of that wave: spread out, run in parallel, bet on hits.

Two years later, the same Chen Rui spent an earnings call asking not how many more lines to open, but which lines to stop.

What he owned first was an era when “a good game makes money”

Greenfield green-light logic was simple.

Plenty of users were still unclaimed; a quality-enough new title could still amortize acquisition through paid UA, word of mouth, and category inertia. On the 2023 Q3 call Chen put the old rule bluntly: while new-user dividends lasted, the market was friendly to launches—if the game was good, it could make money.

Bilibili had an extra illusion: game videos on the platform kept getting longer; danmaku and fan content looked like free distribution. Feeling closest to players, greenlights easily mistook “discussion” for “willingness to pay.” Bandwagon titles felt rational—another ACG hit elsewhere, so spin one up; art a bit short, push five more points; more projects, one of them will break out.

The error was not “making games.”

It was treating greenfield success as an infinitely copyable greenlight formula.

The real bind: the rules changed; the projects stayed on the old timeline

Public data for China’s game market in 2022 showed the first decline in eight years—Economic Observer put revenue near RMB 265.9 billion, down about 10%, with users slightly down too. 2023 looked sharper: roughly 30 new ACG titles in June–July alone, near triple the prior year; after the bloodbath, public roundups often named only titles like Honkai: Star Rail and Reverse: 1999 as true standouts.

Chen’s read was colder: not merely fiercer competition—the competition rules changed. New-user dividends faded into textbook stock-market rivalry. In a stock market, “good quality” no longer equals “can monetize”—you either become the best, or become meaningfully different.

That is the core bind.

Teams did not suddenly get stupid, and players did not suddenly stop loving games. Chen had watched: player activity stayed high; time spent on game videos on Bilibili kept climbing—not that people refuse new games, but that they have no need for more of the same. Art scores of 80 vs 85 vs 87 rarely beat the switching cost for someone already long-retained in a similar title.

The crueler layer is time lag.

Most underperforming launches had been greenlit about three years earlier—when dividends still existed and the bar was still “ship well, earn.” Multi-year cycles lock a company into old rules: by launch day, the referee has changed. Bilibili’s Q3 2023 game revenue was about RMB 990 million, down 33% year on year; the company cited a high base from Time Hunter 3 plus several new titles below expectation.

Rising UA costs are the commercial side of those stock-market rules: users must be taken from someone else’s pocket; acquisition and retention costs rise together. Stack Chen’s “reasonable cost” bar on top, and mid-tier P&Ls crack first.

More projects can mean more danger

Under pressure, the easiest move is to double down.

Hire more, open another line, polish art another notch, buy another quarter of traffic—use “still in development” to postpone admitting the greenlight was already wrong. The 2021 self-dev wave carried that optimism in its bones: enough projects, and probability will smile.

In long-cycle games, more projects is not diversification.

If the greenlight standard is still stuck in the dividend era, each new line only copies one more obsolete judgment; art at 87 still fails to force a switch, while locking cost into a black hole that only reveals itself three years later. The two failure modes Chen later named—bandwagon greenlights, and differentiating via inward competition (especially art arms races)—get amplified across a portfolio, not hedged away.

Around late 2022 he personally took the game reporting line, still speaking externally of “premium self-dev, global publish.” Taking over is not the same as cutting; many CEOs’ first instinct is “I’ll help you ship the hit.” The hard admission is that some projects are not execution problems—they sit on the wrong greenlight coordinate system. Keep spreading, and you industrialize the error.

What forced the change: discovering only at launch that it cannot earn

In May 2023, public reports said Bilibili cut some game product lines; in October, Guangzhou-side studios saw further cuts. On the call Chen did not dress the moves as emotional contraction. He sounded like an accountant:

They cut projects that failed the standard at greenlight—otherwise they would miss market requirements and fail to earn after launch; better to adjust now.

Inside that sentence sits what he stopped doing:

Stop treating “finish it first, then see” as risk management.

He named three new bars; Economic Observer and National Business Daily recorded the same set:

1. Long-lifecycle operations—if you cannot run it long, do not greenlight; agency deals tighten on the same logic.
2. Vertical head or super-head—“pretty good” mid-tier has no seat in a stock market.
3. Reasonable cost—able to build but unable to earn is a prepaid loss.

Together they are not slogan upgrades; they turn “premium” from an adjective into a veto. Bandwagon greenlights and art-involution differentiation, he said flatly, no longer work.

Peers moved in the same season—ByteDance’s self-dev contraction landed alongside Bilibili’s cuts. Chen said it was not only Bilibili; the whole industry was doing it. Rule change, in his view, was not even bad news: games should return to innovation and real team edge, not an arms race of detail scores.