To The Moon

When investors think about artificial intelligence infrastructure, the conversation usually starts with one object:

The GPU.

NVIDIA.

AMD.

Custom accelerators.

More compute.

More chips.

More data centres.

But there’s a problem.

You can build the most powerful AI accelerator on Earth and it becomes dramatically less useful if data cannot move between thousands of accelerators quickly enough.

That is why I think one of the next major AI infrastructure battles will not simply be about computing data.

It will be about:

MOVING IT.

And suddenly Corning becomes much more interesting.

🧠 AI’S BOTTLENECK KEEPS MOVING

Every technology boom creates bottlenecks.

First, AI needed GPUs.

Then it needed enough electricity to power them.

Then investors started focusing on networking, memory and cooling.

Now hyperscalers are attempting to connect increasingly gigantic computing clusters.

That creates enormous demand for:

Fiber.

Optical cable.

Connectors.

Switches.

Transceivers.

Networking equipment.

And the physical infrastructure required to move absurd quantities of data around AI factories.

That physical layer gets far less attention than the shiny processors.

But it is becoming increasingly difficult to ignore.

📈 LOOK AT WHAT CORNING IS ACTUALLY REPORTING

Corning’s Q2 core sales grew 17% to US$4.74B.

But inside that number, Optical Communications sales jumped 32% to US$2.07B.

Enterprise Networks grew 65%.

And management specifically said Gen AI-related product sales were growing significantly faster.

That does not look like AI connectivity demand disappearing.

It looks like it is accelerating.

Corning also announced major agreements with some very familiar names.

Amazon entered a multiyear, multibillion-dollar agreement for Corning optical fiber, cable and connectivity products supporting its expanding U.S. data-center footprint.

NVIDIA and Corning announced a long-term partnership under which Corning plans to increase U.S. optical connectivity manufacturing capacity 10-fold and expand domestic fiber production capacity by more than 50%.

Think about what that implies.

NVIDIA’s AI opportunity is no longer simply:

Sell more GPUs.

The infrastructure surrounding those GPUs is becoming an industrial buildout of its own.

🔌 THE GPU DOES NOT WORK ALONE

Imagine spending tens of billions of dollars filling a data centre with accelerators.

Those accelerators constantly exchange data.

They communicate with storage.

They communicate with networking equipment.

They communicate with other accelerators.

The faster and larger AI clusters become, the more important the connections between them become.

If communications cannot keep pace with computation, those expensive chips spend more time waiting.

And a US$30,000 or US$40,000 accelerator sitting idle because infrastructure cannot feed it efficiently is a terrible use of capital.

That’s why I think AI infrastructure investing needs to evolve beyond:

Who makes the best chip?

The better question could become:

WHO REMOVES THE NEXT BOTTLENECK?

🐂 THE CORNING BULL CASE

Corning doesn’t need to beat NVIDIA.

It benefits because NVIDIA wins.

It doesn’t need Amazon to stop building custom silicon.

It benefits because Amazon builds more AI infrastructure.

It doesn’t need every AI workload to use the same architecture.

It needs the total amount of connected computing infrastructure to keep expanding.

That’s a very different investment thesis.

Corning’s internal Springboard plan targets an annualized sales run rate of:

US$20B by the end of 2026

US$30B by the end of 2028

US$40B by the end of 2030

Management also expects earnings to grow faster than revenue, alongside stronger returns on capital and substantially more free cash flow.

If AI data-centre connectivity becomes a structural growth market, Corning could be one of the less obvious ways to participate.

🐻 BUT THIS STOCK HAS ALREADY TAUGHT INVESTORS A LESSON

Great industry exposure does not automatically make a great stock at every valuation.

Corning had an enormous run earlier this year.

Then it collapsed.

One 13% single-day decline came after a huge run-up, and the shares later suffered another brutal selloff when Q3 sales guidance disappointed expectations despite strong underlying AI-related demand.

Sound familiar?

It is the same lesson we just saw across other AI infrastructure names.

The market increasingly distinguishes between:

Strong demand

and

Expectations already pricing in extraordinary demand.

Corning can simultaneously have a brilliant long-term AI opportunity and still experience violent valuation compression.

Both things can be true.

🎯 MY TAKE

I don’t think the most important AI infrastructure question anymore is simply:

How many GPUs will hyperscalers buy?

We already know the answer is:

A fucking lot. 😂

The next question is what those GPUs require around them.

Power.

Cooling.

Memory.

Networking.

Fiber.

Optics.

Data-centre construction.

AI is slowly becoming less like a semiconductor story and more like an enormous industrial infrastructure cycle.

That potentially expands the investment opportunity far beyond NVIDIA.

And that’s why I think Corning’s volatility is worth watching.

The AI boom may not be cracking.

The market may simply be discovering that:

THE NEXT AI BOTTLENECK IS NOT ALWAYS THE CHIP.

Sometimes the biggest opportunity is the infrastructure nobody notices until there isn’t enough of it.

GPUs created the AI revolution.

Now somebody has to connect the fucking things.

👇 COMMUNITY QUESTION

Where do you think the next AI bottleneck appears?

A) Compute chips 🧠

B) Power and cooling ⚡

C) Memory and networking 💾

D) Fiber and optical connectivity 🔌

And which second-order AI infrastructure company are you watching that the market may still be underestimating?

$GLW $NVDA $AVGO $MRVL $AMZN

Personal market analysis only. Not financial advice. Always do your own research.