Chinese Developers Are Spending Thousands Out of Pocket on AI Tools Their Employers Won't Pay For

A few days ago while browsing a tech community, I came across a rather interesting post.

A guy in tech said he spends at least over a thousand yuan a month just on AI subscriptions.

Claude Pro $20, Cursor $20, ChatGPT Plus $20, plus Gemini, Perplexity, v0, and all that other stuff.

He added the line, "Going to work not only doesn't make me money, I'm actually paying out of pocket."

The comments section immediately blew up. A bunch of people followed up venting their frustrations. Some said they also bought Midjourney and Notion AI on the side. Some said their company wouldn't even provide ChatGPT. Others asked if anyone wanted to split a subscription.

As I was scrolling through, I started to laugh. Not the kind of schadenfreude laugh, but the kind of "so I'm not the only one" laugh.

Because I myself am one of these people.

As AI Becomes a Productivity Tool, the Era of Paying to Work Is Quietly Arriving

Think about it carefully. ChatGPT Plus, Claude Pro, Cursor Pro, GitHub Copilot, Gemini — tens of dollars or over a hundred yuan a month doesn't seem expensive.

But the question is, who should be paying this money?

In theory, these stopped being entertainment expenses long ago. They are production tools, the same category as Photoshop, IntelliJ IDEA, Office, and Figma.

You need Excel for work, the company provides it. You need Figma to draw prototypes, the company buys it. But you need Claude to do your job? Sorry, pay for it yourself.

This is where it gets absurd.

I did a rough calculation. Apart from a few annual subscriptions, my own monthly spending on AI is probably around seven or eight hundred yuan.

And this isn't an isolated case. The developers, testers, product managers, operations staff, designers, and even HR people around me — anyone who has been seriously using AI to get work done over the past two years is most likely paying out of their own pocket.

The deeper someone goes, the more tools they subscribe to. A heavy user spending ten or twenty thousand yuan a year is completely normal.

So, as ChatGPT, Claude, and Gemini become productivity tools, something is quietly happening!

The era of paying to work has truly arrived!

A Programmer's Work Gear

A developer's toolkit used to be simple.

One computer, one IDE, one browser. Done.

And today? Many people's toolchains have ballooned into this:

• Claude Pro, responsible for architecture design.
• Cursor Pro, responsible for writing code.
• ChatGPT Plus, responsible for analyzing problems.
• GitHub Copilot, responsible for code completion.
• Perplexity, responsible for looking up information.
• Midjourney, responsible for generating images.
• Bolt.new, responsible for deployment and release.
• ...

Each tool handles its own area, and strung together they form a complete AI pipeline.

If you also count subscribing to various domestic Coding Plans, like GLM5.2, KIM 3, DeepSeek Pro, and so on, a developer has to spend at least several thousand yuan a year.

But here's the problem.

Most of these tools are not for personal life, but for work.

AI Is Becoming the New "Office Software"; Those Who Don't Use It Are Losing Competitiveness

Think back. Every generation of professionals has a threshold they must cross.

Twenty years ago, if you didn't know how to use a computer, it was hard to find a job. You couldn't even print a resume; who would hire you?

Ten years ago, if you couldn't program or write code, you were severely limited when job hunting. Someone else could write a script to finish a task in three seconds while you were still grinding away manually for half a day.

Today, it's AI's turn.

Using AI and not using AI have started to become two different levels of production efficiency.

A task that used to take you half a day to a full day manually can now be handled in less than ten minutes with AI.

Within this, the real change isn't whether AI will replace someone.

It's that people who can use AI are replacing people who can't.

This sentence is no longer just a slogan; it's happening around you day by day.

Not using AI doesn't mean you lack ability, but you are indeed becoming slower.

And in the workplace, what often matters in competition isn't just ability, but how much value you can produce per unit of time.

You and someone else have similar technical skills, but he uses AI to finish in one day what takes you three. Who do you think the boss will keep in the end?

It's not AI eliminating you.

It's that the person using AI simply crushes you in efficiency.

Companies Only Pay Lip Service to Encouragement, But Don't Reimburse AI Expenses

I've pondered this issue before. There are actually a few practical reasons.

First, the financial budgeting system hasn't caught up.

It's not that the company won't reimburse you; it's that they literally can't. You take a Claude bill to the finance department, and the finance person immediately gets a headache. Which accounting category does this go under? Office supplies? Doesn't look like it. Software procurement? The process doesn't match. R&D expenses? You'd need a project code for that.

Many companies' procurement processes are still designed for traditional software. Fill in the supplier, fill in the number of licenses, sign a contract, issue an invoice — the whole nine yards. What is an AI subscription? Monthly deductions, no contract, sometimes not even a standard invoice, and overseas payments are an even bigger headache.

The finance department itself can't figure it out, so the simplest solution is: employees buy it themselves.

It's not about being stingy over a couple hundred yuan; it's that the financial system simply never created a line item for AI.

Second, the boss only sees the money going out, not the costs being saved.

The boss sees $20 per person per month, several hundred employees, hundreds of thousands a year, and the first reaction is one word: expensive.

But he doesn't do the reverse calculation.

Using AI, suppose a person saves an average of two or three hours a day.

Convert that into salary. A developer with a monthly salary of 15,000 yuan has an hourly rate of about seventy or eighty yuan. The time saved in a month is worth four or five thousand.

Spending a hundred-something yuan on a subscription in exchange for four or five thousand yuan worth of time.

But the problem is, the money spent is clearly written on the books, while the time saved is invisible. The boss stares at the bill and naturally thinks it's expensive. He doesn't see the other side.

Third, security and compliance concerns.

Source code leaks, commercial data outflow, customer privacy — these are all real risks. Many companies would rather hold still than rashly open things up.

So they tend to lean toward enterprise editions, private deployments, local models, building their own AI infrastructure. But that takes time.

Until then, employees either pay out of pocket or go without.

That said, not all companies are this stingy.

There are some companies on the market that reimburse employees for AI subscription fees. Some route it through R&D funds, some specifically approved a budget for AI tools, and some team leads sign off on reimbursements for their team members themselves. The most decisive one I've heard of directly opened Claude Team accounts for all staff, covering everyone from product to development to testing, with the company footing the entire bill. That boss could do the math clearly: spend a little to save a lot.

But honestly, this is still the minority.

The state of most companies is: they verbally encourage everyone to use AI to boost efficiency, but when it's time to actually pay up, they play deaf and dumb. In meetings, it's "we strongly support AI innovation," and then the reimbursement form gets rejected — you pay for it yourself.

Rather Than Waiting for the Company to Provide It, Better to Start Running on Your Own First

To be fair, AI becoming a standard company provision is just a matter of time.

In the future, companies will likely provide large models, AI Agents, and intelligent assistants to employees just like they provide internet, water, and electricity. Job requirements will slowly shift from "proficient in Java, MySQL" to "proficient in using Claude, familiar with Agent workflows."

Because what truly efficient teams compete on isn't just excellent talent, but whether excellent talent possesses equally excellent production tools.

Employees shouldn't have to pay out of pocket to do their jobs — that principle is certainly correct.

But that day hasn't arrived yet.

Budget systems need to change, compliance needs to be established, infrastructure needs to be built. Fast case, a few months; slow case, a year or two.

This window period is the truly critical part.

Can you afford to wait?

Just figure out one thing. The computer, desk, chair, and internet the company provides you — which of those can you take with you the day you leave? Not a single one.

But the Claude, Cursor, and ChatGPT you bought with your own money, the AI intuition you've trained, the prompt library and Agent workflows you've accumulated — all of these things go with you. Switch to another company, and your efficiency doesn't reset to zero just because you changed desks.

On the surface, this money looks like it was spent for the company, but the capability that grows out of it is rooted in you.

I sometimes feel that AI isn't like water, electricity, gas, or office furniture. It's more like a muscle.

You have to train it, use it, touch it every day. If you don't use it, it atrophies; if you use it, it strengthens. If the company gives you a gym membership but you don't go, the muscle won't grow. Conversely, even if the company doesn't provide it, if you spend your own money to train, what you build is yours.

Every time a wave of technology rises, there's always a group of people who, while others are still agonizing over "should this money be spent or not," pay up first themselves.

It's not that they have more money than sense. It's that they vaguely sense that this thing is going to change the game, and if they don't enter now, they won't even be able to eat the dust later.

This window period of one, two, three, or five years is the watershed.

By the time the company finally provisions AI, that group who paid first will already be the most knowledgeable on the team. The people who kept waiting, when they finally get their company account, will probably just use it to look up information and polish weekly reports.

The gap is widened just like that.

Final Thoughts

The phrase "paying to work" sounds like a joke.

But think about it from another angle. This might be the cheapest entry ticket an ordinary professional can buy in recent years.

A few hundred yuan a month buys you a surplus of two or three hours of freedom every day, buys you the intuition of mastering new tools one step ahead of others, buys you a capability that you can take with you no matter which company you go to.

So you ask me, if the company doesn't provide AI, is it worth paying out of your own pocket?

I think it is.

Not because of the time saved. It's because what you buy back is something others can't catch up to.

What you're paying isn't money.

It's time that others can't catch up to.