Semiconductors have fallen 20% in the past half month—technicals already broke down long ago.

Many people are puzzled: AI is still a must-have, and big companies’ earnings aren’t blowing up—so why did it drop this much?

The truth is, this kind of selloff is largely because the market is too crowded.

Earlier, everyone thought buying semiconductors was “easy money,” so spot, options, and leveraged ETFs all piled in.

That leads to everything going crazy together when it rallies—then triggering risk-control lines when it falls, with all algorithmic trading selling at once, and liquidity drying up instantly.

Kimi K3 is a perfect shorting reason.

In the past, people believed U.S. AI models were far ahead. Now Kimi-K3 has taken 6 first places out of 7 sub-segments in the Frontend Code field, and its overall ranking is even steadier than GPT plus 5.6 SOL.

The key is that it’s open-source!

How do those Wall Street bosses who poured heavy money into OpenAI and Anthropic feel about that?

They worry that if a price war breaks out, nobody will be able to make money—then chip demand will shrink as well.

On top of that, at the start of the month, Meta was exposed for selling off idle compute capacity, and everyone started doubting that computing power is oversupplied.

This kind of suspicion spreads like a virus. Combined with developments from Iran and the negative news from SK hynix (SK Hynix), it directly maxed out the negative feedback loop.

The most solid approach right now is to lie flat.

Based on FINRA data, retail leverage hasn’t dropped fully yet, which means the final “big bottom” hasn’t arrived.

But this is a compression-style deleveraging—falls fast and rebounds fast too.

If the fundamentals of your holdings haven’t deteriorated, cutting losses now will likely end up as pure loss when you look back.

Just wait for this round of mechanical liquidations to finish.