《AI Chip Plunge: Is It a Bubble Bursting—or a New Round of Wealth Redistribution?》

AI chips are plunging— is the compute shortage real demand, or just the next capital-market story?

Recently, the AI chip sector has shown clear volatility.

NVIDIA, Broadcom, AMD, and the entire semiconductor industry chain have all gone through a round of capital readjustment.

Many people are starting to ask:

Has AI been overhyped?

As for the so-called compute shortage, is it truly demand—or a new story fabricated by the capital markets?

My view is very clear:

Compute demand is real, but the market has already partially priced in the future.

These two things must be viewed separately.

What is the biggest change in AI over the past year?

Not that ChatGPT went viral.

It’s that enterprises have started to invest in AI infrastructure for real.

For giants like Microsoft, Google, Amazon, and Meta, one of this year’s biggest capital expenditure directions is data centers.

Why?

Because in the model competition, at its core, it has already become a compute competition.

Training needs GPUs.

Inference needs GPUs.

When AI Agents roll out widely in the future, they will also require continuous computing resources.

So from an industry perspective, the compute shortage is not fake.

The real question now is:

If demand is real, how much of the future has the stock price already reflected in advance?

That’s the core of what the market trades.

Why have AI chips been falling recently?

Because the capital markets always trade expectations.

When everyone knows AI is the future, capital starts looking for:

How much growth hasn’t been priced in yet?

NVIDIA’s rise over the past few years, in essence, is the market redefining it.

From a chip company, it became the key supplier of AI infrastructure.

But as valuations climb higher, the market’s requirements rise too.

Previously:

If revenue grew 50%, the market felt surprised.

Now:

If revenue grew 50%, the market may feel it’s not enough.

That’s the harshness of growth stocks.

But I believe it’s still too early to say an AI bubble has burst.

A real bubble usually has a few characteristics:

First, demand starts to decline.

Second, companies stop investing.

Third, inventories across the industry chain begin to pile up.

At present, there hasn’t been an obvious reversal in AI infrastructure.

Some segments are even still in short supply.

Especially:

HBM high-bandwidth memory.

That’s also why I’ve been paying attention to Micron (MU) for a long time.

AI doesn’t just need GPUs.

The stronger the GPUs, the higher the demand for high-speed storage.

In the future AI industry chain, it’s impossible for only NVIDIA to profit.

Storage, network equipment, and even power infrastructure could all be beneficiary directions.

So can we still buy AI stocks now?

My view:

You can’t panic-sell just because they’ve fallen.

And you also shouldn’t blindly chase just because the long-term logic is right.

Now the market is entering the second stage.

First stage:

Capital grabs certainty.

Buy NVIDIA.

Buy GPUs.

Buy the most core assets.

Second stage:

Capital starts searching for a balance between valuation and growth.

In the future, real opportunities may not be the companies with the biggest stock gains.

But rather those that:

are delivering on fundamentals,

have not fully priced in valuation,

and are positioned at critical points in the industry chain.

So my judgment is:

AI is not over.

But the first round of疯狂疯狂 upside mania is already in the past.

Next, the market will be more discerning.

The future profit effect may gradually spread from:

GPU → storage → data centers → power infrastructure

The truly big opportunities don’t necessarily come from where everyone already knows to look.

They come from when the market starts to reprice.

A line from the trading desk:

The biggest risk for AI isn’t that there’s no demand.

It’s that the market has already bought the demand several years in advance into the stock price.

The real opportunity always belongs to those who can spot industry trends—and who also know how to wait for the right price.
#夏日创作营 #AI算力