After a drop in the US stock market last night, today the Asian markets are also taking a hit.

On the surface, it looks like a flood of 'bad news', but if you break it down piece by piece, you'll find that there’s really only one core bearish factor—market doubts whether the AI giants can keep burning cash like this in the long run.

First, let's lay out four reasons for the dip, and then we’ll tackle them one by one:

1. Interest rate hike expectations are back, and big tech takes the first hit.

Recently, the market feels like the Fed is sounding pretty hawkish, and with the recent CPI data not looking great, the chatter about 'rate cuts being delayed even longer' is coming back.

High interest rates are most unfriendly to companies that prop up their stock prices with the story that “they can make a lot of money in the future.” Google, Microsoft, Amazon, and other big tech are now desperately burning money to do AI, betting on profits years later. When rates rise, the money they’ll make in the future gets discounted back to today—so the stock price naturally falls first, as if on cue

But is this round of inflation really that scary?

Actually, these inflation worries are largely driven by oil prices. When oil rises, transportation costs go up, and many things rise along with it, pushing the CPI numbers higher. That doesn’t mean the U.S. economy is overheating again—it’s more like energy prices are lifting the headline data

So rate-hike expectations are more like a market excuse for “selling to smash the price,” not the real underlying cause

2. Zhipu is too strong—so the market is starting to ask a life-or-death question again

This is the most worth paying attention to thing from last night

Zhipu’s new model performance is impressive. The market isn’t afraid that “domestic models are rising”—it’s afraid of a deeper issue behind this:

If latecomers can catch up by spending little, then the leaders are smashing hundreds of billions now—are they really worth it?

Remember this: user migration costs are extremely low. In many cases it’s just a matter of switching APIs. If you become slightly less capable, they turn around and leave

Google took the hardest hit this time because it didn’t really establish absolute dominance in AI models. OpenAI had a first-mover advantage, Anthropic has a strong reputation, Meta takes an open-source route, and Google’s Gemini never really created the feeling that “it’s the strongest and can suppress everyone else.” Once the market sees the story that you can catch up with low cost, the first target of questioning is it

What’s scary about this is that it makes everyone start doubting just how deep the moat of large models really is. Previously, everyone assumed: the more you spend, the bigger your advantage. Now that belief is starting to loosen

3. The real core negative: the logic of infinite capex growth has been thrown into doubt

Let me explain a term here: Capex (capital expenditures)

Put simply: companies are smashing big money upfront now, so they can make big money later. On the AI track, that means buying GPUs, building data centers, rolling out servers, and doing infrastructure

In the past AI rally, an important prerequisite for it to rise like this was that the market believed these giants would keep going crazy with spending. Spend 100 billion this year, 150 billion next year, and keep ramping up after that

As long as they keep dumping, the entire industry chain—chip sellers, memory sellers, and makers of optical modules—will all get to feast

But now, that premise has been thrown into doubt

If the lead advantage isn’t that valuable, and if later players can chase 70–80% by spending 30 bucks, then what’s the point of spending 100? Is the input-output ratio getting worse?

What the market is truly afraid of isn’t that Zhipu is very strong, but that once this “doubting chain” starts, the story that capex will grow at high, never-ending rates can no longer be told. That’s the core negative

4. Micron’s earnings: it’s not that people fear it will be bad—it's that they fear it won’t be “out of this world” enough

Micron is about to release its earnings—so the market is hiding first

Many people think Micron will most likely deliver a strong performance this time, with revenue and profit beating expectations

So why are people still nervous?

Because Micron is no longer just a memory company. It’s already become a barometer for AI hardware demand. The HBM it sells (high-bandwidth memory, the high-speed storage paired with AI chips) is an important signal of whether AI investment is heating up

What the market demands of it isn’t just “doing well on the test,” but “scoring above full marks.” If it’s merely decent, but not spectacularly so, the stock can still be smashed

So the risk on Micron’s side isn’t weak fundamentals—it’s that expectations are priced too fully

Pension reallocation, quant fund selling to smash the market, and deleveraging—these are all “finishing blows”

Finally, about liquidity conditions—don’t overthink this part. Just remember one line:

They don’t decide why the market is falling, but they determine how hard it falls

• Pension rebalancing: Stocks rose too much before, and the weighting got out of range. At month-end they have to sell stocks and buy bonds—purely a mechanical move

• CTA quant fund selling to smash the market: some systematic funds automatically sell when the price hits a certain level. It’s not that they’re bearish—it’s that the trigger signal fired

• Deleveraging in Asia: in Korea, many people previously borrowed money to trade stocks. Now when it falls, they rush to repay—selling more leads to falling more

With these actions stacked on top of each other, a market move that might have fallen only 2% gets amplified into 4% or 5%

How should we look at this selloff?

If you had to sum it up in one sentence, it’s this:

This isn’t that the market suddenly stops believing in AI. It’s that it starts seriously doubting whether AI giants can still burn money blindly like before

Rate hikes are just an external pressure. Micron is the sentiment anchor; liquidity conditions are the amplifier. The real root cause is doubt about the sustainability of capex growth

But for now, one distinction is needed: the market is only doubting whether capex will ease—not that it has truly seen anyone ease. The difference between the two is huge

If next, the earnings and guidance of big players like Google, Microsoft, Amazon, and Meta still maintain high-intensity AI spending, then this selloff looks even more like a deleveraging and emotion release after stocks have run up too far—not a complete turn in trend

My view: more like a pullback from a high level, not the end of the main trend

At this stage, I’m more inclined to see this move as: down first, deleveraging next, and then waiting for verification

What really needs vigilance isn’t the selloff itself, but whether there will be hard evidence afterward. Watch three things:

1. Have the capex guidance from the big players gone down? This is the most critical point

2. Has Micron and the upstream supply chain given any signals of demand slowing

3. Is there any real slowdown in cloud business and inference demand

As long as that chain of evidence doesn’t show up, then for this selloff, I’d rather treat it as an opportunity to set up positions on dips—not a signal that the whole trend is over