Talk of 996—the 9am-to-9pm, six-day schedule—usually runs through humanity, emotion, interests, law, game theory, and surplus value. I have yet to see an argument that lands the point. The test is simple: if appeals to humanity, emotion, interests, or law could change hiring rules in the internet sector, they would have changed them already. The same complaints have been recycled for at least a decade. A rule that survives that much moral criticism is probably being produced by something more structural.
I think 996 is best explained by a competition rule: fast fish eats slow fish, and the winner takes all. Why has overwork become so severe in China’s internet industry? Four forces stand out.
The marginal cost of serving another user is close to zero
An internet product is not a traditional product. Once it has been developed, the cost of adding another user is almost nil. Serving 100 users and serving 10,000 users costs very little more; at a larger scale, serving a million users and serving the whole country are not that far apart either. That pushes internet products toward user scale first, and makes first-mover advantage unusually strong.
The industry has grown mainly at the application layer, not in basic theory
China’s internet sector has concentrated on applications because the path to profit is shorter. But the application layer has low competitive barriers. A good idea is easy to copy. A company therefore has to move fast enough to enclose territory before rivals react.
The domestic market is large enough to justify high-risk investment
China has one-fifth of the world’s population and one-quarter of its internet users. The market is large enough to justify high-risk investment. The risk comes from two directions. One is the extra cost of beating competitors on time; if you run slower, you may have run for nothing. The other is the full cost of running in the wrong direction; before anyone reaches the finish line, no one knows which way is right, and a wrong direction also means running for nothing.
A large market and high payoff make big risks worth taking. But investors are not fools. Worth taking and avoiding risk are different things. No one can be sure of the direction, so the only way to avoid risk is to run fast.
Given these three forces, an internet company in China has to run fast to survive and profit, and to run fast it has to run without stopping. That is the fundamental driver of 996.
Programmer work is not divisible
Programmer work looks divisible, but for an internet company it is not, certainly not in the way assembly-line work is. In a highly developed industry, each programmer’s work has already been cut into the smallest block. Within the block assigned to each person, it cannot be cut again or split among collaborators. To raise project efficiency, the only lever left is longer hours.
Two common misunderstandings
The work can still be divided
From a purely technical angle, programmer work may still be divisible. From an industrial-efficiency angle, it is not. The reason is simple: under current conditions, slicing work that requires coherent thinking sends management costs through the roof. Four people working one day after the split may produce less than one person working two days, or even less than one person working one and a half days.
Even if production tools and management methods improve enough to allow finer division, 996 would not change. Tool improvements and management upgrades are shared across the internet industry. As long as the fast-fish-eats-slow-fish, winner-takes-all rule remains, sector-wide productivity gains do nothing for programmers’ working hours. The only exception would be a specific company that innovates its own tools and management and secures a lead large enough to offset the efficiency lost outside eight hours. That could improve its employees’ condition. I do not see it happening in the near future. No visible company has that kind of innovation capacity. And even if one did, it would probably use the advantage to run faster and widen the lead, not to improve employee welfare.
996 is exploitation
The claim that 996 is exploitation and extraction is popular online. I think it is nonsense. Given the indivisibility of programmer work, one employee on 996 costs a company more than 1.5 employees on an eight-hour schedule. Extending labor time costs the company more, not less. Exploitation does not enter the picture.
Why is one 996 employee more expensive than 1.5 eight-hour employees? First, a 996 employee cannot really work 12 hours; 11 is already the ceiling. That is less than 1.5 normal workdays. Second, fatigue lowers efficiency. By output, a 996 day is well below 1.5 eight-hour days. Third, 996 pushes up the unit cost of labor per unit of time.
Programmers are not scarce. The salary range in IT—from tens of thousands of yuan a month down to a few thousand—shows that. If they were truly scarce, such a spread would not exist. 996 reduces the number of actual jobs but raises requirements, because there is no room for error or wasted time. Companies sort programmers by whether they can handle the intensity, whether they are experienced, whether they know multiple development languages, how fast they learn, whether they work well in a team, and even whether family obligations will slow them down. Only a small share meet all conditions. That pushes their price up. The requirements exist only so they will not waste time and can finish projects as fast as possible. Without time pressure, those conditions could be dropped, and the vast majority of IT workers could handle the vast majority of IT work.
So 996 does not save the company money, and it does not make employees create more value relative to their pay. The system itself does not create exploitation. Its only effect on an internet company is to shorten project completion time and increase competitive advantage. Once 996 becomes universal, it turns into an extra burden on the internet sector. No company benefits.
Related questions
Is overtime voluntary or forced?
Many people take the moral high ground and say programmers are forced to work overtime. I do not accept that. Programming is a highly mobile occupation. Companies do not restrict personal freedom, and one rarely hears of contracts that demand huge compensation for resigning. How can completely free people be forced to work overtime? If they do not want overtime, they can change to a job without it.
Forced means limiting the right to choose, or raising the cost of choosing. Neither applies to programmers.
Why do some internet jobs avoid overtime?
There are roughly two reasons. One is that the company does not want to run too fast and prefers natural growth; its pay will not be especially high. The other is that the company has a separate track, with no one else playing on it. Whether it runs or walks, it is first.
Why do no-overtime companies pay about as well as big 996 companies?
As noted, the internet is a winner-takes-all industry. A wrong direction means running for nothing; running slowly means running for nothing. The eventual winner may be one in a hundred. For the industry as a whole, the cost of doing that project is 100 times one company’s cost, not one company’s cost. For a company on its own track, the cost is just what it spends. Even if it pays the same personnel cost and has only one-tenth the efficiency of a competitor on the public track, its overall cost is one-tenth that of a fully competitive field. Why would it work overtime?
Is 996 a failure to enforce labor law?
This view is popular because Article 36 of the Labor Law says the state adopts a working-hour system of no more than eight hours a day and 44 hours a week on average. Programmers clearly exceed it. But curing 996 has nothing to do with the Labor Law, or rather the Labor Law cannot cure 996. It is like the campaign to reduce the burden on primary-school students. Before 2000, school ended after 17:30; now it ends at 15:30. Yet children still have three or four hours of homework a day, and their sleep may be shorter. Programmers are the same. Force down the hours spent in the office, and a large amount of homework will move to the home.
I think the crux is not weak enforcement of the Labor Law. It is weak enforcement of the Intellectual Property Law and other laws that shape the competitive environment.
Is 996 because Chinese people are hardworking?
Many people apply a double standard to 996: they call it exploitation of employees, and they also call it a sign of Chinese diligence. The first assumes coercion; the second assumes voluntariness. The two are contradictory. Having given my explanation above, I do not think 996 exists because Chinese people are hardworking. It exists because of the competitive environment.
People who hold the diligence view may be reaching for a national label when they see a special phenomenon. I also regard diligence as a fine tradition of the Chinese nation, but a national label should explain general phenomena, not a special one. If diligence were the cause, every industry would be on 996.
Courier, media, and other industries are now doing 996. Did IT set a bad example?
Again, this is about the competitive environment. In short, any industry with the following features may end up on 996. The precondition is a fully competitive field:
- time value matters in competition;
- normal competition can produce a scale monopoly;
- there is no high technical barrier;
- existing competition norms impose weak constraints or protection.
It is foreseeable that as internet technology spreads through production, every industry speeds up. If an industry also meets those four conditions, 996 will follow.
A final clarification
This is not an attempt to defend internet companies. It looks at the deeper cause of 996 from the angle of the industrial competitive environment. Nor is it a denial of existing intellectual-property protection. China’s progress in that field is visible to the world, and I recognize it. But from an industrial perspective, there is still plenty of room to improve constraints and protection in many sectors, and structural adjustment still needs guidance and rules.