#Lumentum Jumps 14% — Is CPO the Next Memory Trade or Just Another AI Hype Cycle?

#LITE #COHR #MRVL #NVDA #AI #CPO #OpticalNetworking #DataCenters #Semiconductors #TigerTrade

Lumentum’s 13.72% surge makes sense, but I would not treat it as a simple “buy everything optical” moment.

My view is:

AI optical interconnects are a real structural opportunity, but optical stocks are now trading with hype-cycle behaviour. I like the theme, but I would not blindly chase $LITE after a vertical move.

The reason investors are suddenly paying attention is simple. AI infrastructure is no longer only about GPUs and HBM memory. Once hyperscalers build massive rack-scale systems like GB200, Vera Rubin and future AI factories, the bottleneck increasingly shifts from pure compute to data movement.

A GPU cluster is only as powerful as its ability to move data between chips, racks, storage and networks.

That is where optical interconnects become important.

Why Optics Matter in AI Data Centres

AI workloads require enormous movement of data:

• GPU-to-GPU communication

• Rack-to-rack networking

• Switch-to-switch bandwidth

• Storage-to-compute movement

• Training cluster synchronisation

• Inference traffic at scale

• Data-centre-to-data-centre connectivity

Copper works well over short distances, but as bandwidth rises and clusters spread across racks, copper becomes more power-hungry, more lossy and harder to scale.

Optics solve this by using light instead of electrical signals.

That is why investors are starting to view optical components as the next AI “picks-and-shovels” trade after GPUs, HBM and advanced packaging.

The simplified AI infrastructure chain looks like this:

AI COMPUTE

$NVDA GPUs / Custom ASICs / XPUs

↓

MEMORY

HBM / DRAM / Storage

↓

NETWORKING SILICON

Switches / DSPs / Retimers

↓

OPTICAL INTERCONNECTS

Lasers / Transceivers / CPO / Photonics

↓

AI FACTORY SCALE

More GPUs working together efficiently

The market has already rewarded compute.

Then it rewarded memory.

Now investors are trying to find the next bottleneck.

Optics are a logical candidate.

Why Lumentum Benefits

Lumentum is not just a random hardware supplier. It sits in areas directly connected to AI data-centre scaling, including high-speed optical components, lasers, photonics and cloud networking.

The bull case is that as AI clusters become larger, hyperscalers will need significantly more optical connectivity per accelerator, per rack and per data centre.

If that happens, Lumentum’s addressable market expands meaningfully.

The company’s opportunity is not only tied to more data centres being built. It is tied to each data centre becoming more optically dense.

That distinction matters.

OLD DATA CENTRE:

More servers = more networking

AI DATA CENTRE:

More accelerators = exponentially more data movement

If AI clusters scale from thousands of GPUs to tens of thousands or hundreds of thousands of accelerators, the amount of high-speed connectivity required could grow faster than compute units themselves.

That is the structural opportunity.

Why CPO Is the Big Narrative

CPO, or co-packaged optics, is important because traditional pluggable optical modules may become less efficient at extremely high bandwidths.

In a traditional setup, the signal travels electrically across a board before being converted into light at the pluggable transceiver.

At very high speeds, that electrical path wastes power and creates signal-integrity problems.

CPO moves optics closer to the switching chip.

The goal is:

• Lower power consumption

• Higher bandwidth density

• Lower latency

• Better signal integrity

• Improved rack-scale networking

• More efficient AI clusters

This is why CPO is being treated like a major architectural transition rather than just a component upgrade.

For AI factories, power efficiency is not a small issue. Every watt spent moving data is a watt not spent on useful compute.

If optics reduce networking power and improve utilisation, hyperscalers will pay for it.

Is This the Next Memory Trade?

Partly, yes.

The similarity to memory is that both HBM and optics become more valuable when AI infrastructure scales.

HBM feeds the GPU.

Optics connect the GPUs.

Both sit near critical bottlenecks.

But there is also a major difference.

HBM has a clearer near-term pricing and supply-demand story. The market can see tight capacity, customer commitments, wafer constraints and pricing power.

Optical components are more fragmented.

Different companies supply different parts of the stack:

• Lumentum: optical components, lasers, photonics

• Coherent: optical transceivers, lasers, materials, components

• Marvell: DSPs, switching, custom silicon, optical interconnect technology

• Ciena: optical networking systems

• Corning: fibre and glass infrastructure

• Broadcom: networking silicon and custom AI infrastructure

• NVIDIA: full AI platform and networking architecture

This makes the opportunity real but harder to underwrite.

Not every company in the optical chain will capture equal economics.

Some suppliers may see strong volume growth but limited margin expansion if competition increases or hyperscalers negotiate aggressively.

That is why I would not simply call optics “the next HBM” and buy without discrimination.

The Valuation Problem

The rally has been violent.

$LITE recently traded around the mid-$800s with a market capitalisation above $80B and a P/E ratio above 150x.

$COHR also trades at a very elevated valuation, with a market capitalisation above $70B and a P/E ratio above 170x.

Those numbers matter.

Even if the AI optics thesis is correct, the stocks are no longer priced like undiscovered cyclicals.

They are being priced like strategic AI infrastructure winners.

That creates two risks:

1️⃣ The business keeps improving but the stock still corrects because expectations were too high.

2️⃣ The theme is real, but investors overpay for the wrong part of the value chain.

The market is already capitalising a major AI optical cycle.

That does not mean the stocks cannot go higher, but it does mean the margin of safety has collapsed.

The Biggest Risk: Optical Stocks Are Historically Cyclical

This sector has a long history of boom-bust behaviour.

Optical stocks can experience explosive rallies when demand accelerates, capacity tightens and investors extrapolate growth.

But they can also suffer brutal drawdowns when:

• Hyperscaler orders pause

• Inventory builds

• Component pricing weakens

• Customers dual-source supply

• New capacity comes online

• Margins disappoint

• A product transition gets delayed

• Valuation multiples compress

This is not software.

These are hardware-heavy, supply-chain-sensitive, capital-cycle businesses.

The danger is that investors start treating optical components like monopoly AI platforms when they may still behave like high-beta semiconductor cyclicals.

That is the key tension.

STRUCTURAL DEMAND: Bullish

CYCLICAL EXECUTION RISK: High

VALUATION RISK: Very high

THEME QUALITY: Strong

ENTRY QUALITY: Weak after a vertical rally

Why the Theme Still Has Legs

Despite the risks, I do think AI optical interconnects are structural rather than pure hype.

The reason is that the AI infrastructure problem is physical.

You cannot scale AI purely with more chips if the chips cannot communicate efficiently.

Large AI systems require:

• Faster networking

• Lower power per bit

• Higher bandwidth density

• Better rack-scale interconnect

• More efficient data movement

• Reduced latency across clusters

These requirements do not disappear if stock prices fall.

They are dictated by physics, architecture and economics.

If anything, as AI models become larger and inference demand explodes, data movement becomes more important.

This makes optics a genuine long-term opportunity.

The question is not whether optics matter.

The question is how much future success is already priced into $LITE, $COHR and the rest of the optical complex.

What Would Make Me Bullish Enough to Buy?

I would want one of three things before aggressively buying $LITE:

1️⃣ A meaningful pullback after the hype cools

2️⃣ A multi-week consolidation where price stops making parabolic moves

3️⃣ Another earnings report proving that AI optical demand is translating into revenue growth, margin expansion and strong forward guidance

I would also watch whether management commentary confirms durable demand from AI customers rather than a short-term order surge.

Key things to monitor:

• Data-centre revenue growth

• Gross margin improvement

• Backlog quality

• Hyperscaler concentration

• 800G and 1.6T demand

• CPO product traction

• Customer commitments

• Inventory levels

• Capex requirements

• Free cash flow conversion

If revenue rises but margins do not, the market may eventually punish the stock.

The best AI infrastructure plays do not just grow sales. They convert growth into durable earnings power.

My Scenario Analysis

🟢 Bull Case — 30%

AI data centres rapidly transition toward more optical density, CPO adoption accelerates, and hyperscalers treat Lumentum as a critical supplier.

Revenue grows faster than expected, margins expand, and investors continue re-rating optics as a strategic AI infrastructure category.

In this scenario, $LITE can continue outperforming despite looking expensive.

🟡 Base Case — 50%

AI optical demand remains strong, but the stock has already pulled forward several years of optimism.

Lumentum grows well, but shares become volatile, moving sideways or correcting while earnings catch up to valuation.

This is my main expectation.

Great theme, difficult entry.

🔴 Bear Case — 20%

The sector becomes overcrowded, valuations overshoot, CPO adoption takes longer than expected, or hyperscaler ordering becomes lumpy.

Revenue still grows, but not enough to justify the multiple.

The stock suffers a sharp drawdown even though the long-term thesis remains intact.

This is exactly how high-quality cyclical AI hardware names can hurt late buyers.

Would I Chase Lumentum Here?

Personally, no.

I would not chase $LITE immediately after a 13.72% rally driven by broad optical enthusiasm.

That does not mean I am bearish on the company.

It means I respect the difference between a good business theme and a good entry point.

For existing holders, I would consider holding a core position while trimming some strength if the position has become oversized.

For new buyers, I would wait for volatility.

Optical stocks rarely move in a straight line. If this is truly a multi-year AI infrastructure theme, there will likely be better entries than buying during the most euphoric part of the move.

My Ranking of the AI Optical Trade

For risk-adjusted exposure, I would think about it like this:

🥇 $NVDA — Best overall AI platform exposure

🥈 $MRVL — Strong custom silicon + networking + optical DSP exposure

🥉 $COHR — Broad optical and materials exposure, but high volatility

🏅 $LITE — High upside to optical interconnect demand, but very valuation-sensitive

🏅 $CIEN — More systems-level networking exposure

🏅 $GLW — Fibre infrastructure exposure, less pure-play AI torque

$LITE may have stronger upside torque than some larger names, but it also has higher downside risk if the optical rally unwinds.

Final Verdict

AI optical interconnects are not fake hype.

They are a real structural bottleneck in the next phase of AI infrastructure.

As AI clusters scale, the market will increasingly care about how data moves between GPUs, racks and data centres. That makes optical networking, lasers, photonics and CPO strategically important.

However, the stocks are behaving like a hype cycle.

When a historically cyclical optical name jumps nearly 14% in one day and trades at triple-digit earnings multiples, investors need discipline.

My view:

AI OPTICS THEME: STRUCTURAL

CPO OPPORTUNITY: REAL

LUMENTUM BUSINESS CASE: IMPROVING

CURRENT VALUATION: DEMANDING

SHORT-TERM PRICE ACTION: SPECULATIVE

BEST ACTION: WATCH, WAIT, ACCUMULATE ON PULLBACKS

OVERALL VERDICT: BULLISH THEME, CAUTIOUS ENTRY

I would rather buy Lumentum when the market temporarily forgets the story than when everyone suddenly agrees optics are the next AI gold rush.

The best AI infrastructure trades are usually bought during doubt, not during peak excitement.

@Lumentum $LITE

@CoherentCorp $COHR

@MarvellTechnology $MRVL

@NVIDIA $NVDA

@Ciena $CIEN

@Corning $GLW

@Broadcom $AVGO

#Lumentum #Coherent #NVIDIA #Marvell #Broadcom #Ciena #Corning #ArtificialIntelligence #AIStocks #OpticalInterconnects #CoPackagedOptics #SiliconPhotonics #Photonics #DataCenterStocks #AIInfrastructure #NetworkingStocks #SemiconductorStocks #GrowthStocks #MomentumStocks #StockMarket #MarketAnalysis #LongTermInvesting #InvestingOpinion #T