Compute Prices Finally Rise: Alibaba and Baidu Hike 34%, Ending AI's Price War

If you thought the AI industry was still stuck in the narrative of “models getting bigger while tokens get cheaper,” the events of April 18 will completely change your perspective.

On that day, both Alibaba Cloud and Baidu Smart Cloud announced price increases for AI compute and storage products, with hikes reaching 34% and 30% respectively.

This marks the official end of the two-year era where “compute was expensive but models were cheap.”

Why the Sudden Price Hike?

Honestly, the timing is quite interesting.

On April 17, Claude Opus 4.7, new OpenAI versions, Kunlun Tech, and Zhiyuan Robotics all released products—a super launch day in AI circles. The very next day, China’s two major cloud providers announced price increases. It reminds me of what my mom always says: “When vegetable prices go up, you gotta change the pot too.”

Joking aside, the real reason is supply and demand dynamics.

For the past two years, major cloud providers waged price wars to capture the AI market. You cut 20%, I’ll cut 30%—compute prices kept falling. But compute has real costs: GPU procurement, power consumption, data center operations—none of these are free.

A friend doing AI startup work complained to me last year: “Current compute prices are like cabbage prices. I can’t believe this can continue forever.”

Turns out, he was right. Price wars can’t continue indefinitely. Eventually, rationality returns.

What This Means for AI Startups

Short-term: bad news. Long-term: actually good.

The bad news: costs are rising.

AI applications dependent on cloud compute, especially inference services, will see noticeable cost increases. I did the math: if your business makes 100 million inference calls daily, at current price increases, you’re spending hundreds of thousands more per month.

For startups, that’s not a small amount.

The good news: industry rationalization.

The past two years of price wars created an illusion—that compute would keep getting cheaper. Some companies expanded recklessly, skipping optimization because compute was cheap anyway.

Now with higher prices, everyone’s forced to think seriously: how to improve compute utilization? How to optimize model inference? How to do more with less compute?

Long-term, this benefits AI industry health. After all, nobody wants false prosperity sustained by burning money.

My Prediction

My personal judgment: this price increase might just be the beginning.

With next-gen models like DeepSeek V4 and GPT-6 coming online, compute demand will only grow. Meanwhile, domestic GPU production (Huawei Ascend, etc.) is still ramping up—supply-demand tensions won’t ease soon.

So if your business depends on cloud compute, I suggest planning costs ahead. Don’t wait for another price surge to realize you can’t afford it.

An Interesting Observation

I noticed a detail: only domestic cloud providers raised prices.

AWS, Azure, and Google Cloud haven’t made a move yet. This might indicate tighter supply-demand tension in China’s AI compute market, or that Chinese providers have exhausted patience for price wars.

Either way, it’s worth watching.

Final Thoughts

Rising compute prices bring pressure to AI startups, but also motivation.

The pressure is cost control. The motivation is forcing technical innovation. Companies that can reduce costs by optimizing model architecture and improving inference efficiency will actually become more competitive.

It reminds me of an old saying: crisis is opportunity. It depends on how you respond.

Let’s wait and see the data. Check next quarter’s earnings to see if compute prices continue rising.