Quantum Computing — The Next Capital Cycle After AI, or Another Bubble?

In February 2025, Microsoft revealed a bet it had been making quietly for 25 years.

A new kind of quantum computer — Majorana 1.

Around the same time, IonQ stock was up 250% in twelve months, and the quantum computing ETF was up 60%.

As the AI capital cycle starts to feel exhausted, quantum is being named as the next hype candidate.

But what is a quantum computer, and what changes when it's actually built?

This piece works backwards from the destination — what becomes possible — to how the technology works and who's positioned to win. Zero quantum physics background assumed. Every term is introduced with an analogy.

Let's look at what happens when quantum computers reach what the industry calls 'the threshold'.

We'll define 'threshold' in detail later. For now, think of it as 'a quantum computer that can run stable calculations without errors'.

Let's say you want to develop a new drug today.

Average 10-15 years, $2.6 billion.

The reason it takes so long: molecular simulation is impossible. You can't accurately predict how a drug candidate will bind to a human protein on a computer, so you end up synthesizing tens of thousands of variants and running physical experiments.

When quantum hits the threshold, this changes fundamentally.

Molecules are quantum systems. So a quantum computer can simulate a molecule as a molecule.

If you screen 1 million Alzheimer's candidates in quantum simulation before any lab synthesis, and only the top 100 get actually made

10 years becomes 1 year.

Roche, Merck, and Pfizer already run a quantum drug discovery consortium with IBM Quantum.

JPMorgan joined too.

This is the use case the industry agrees on most: quantum's clearest commercial path.

Think about EV batteries.

Lithium-ion is near its chemical limit.

To make batteries that are lighter, longer-lasting, and faster-charging, you need entirely new chemistry.

The hard part: the number of possible molecular combinations exceeds the number of stars in the universe.

Classical computers can't simulate them all, so we lean on intuition and trial-and-error in the lab.

Quantum computers can exhaustively search this chemical space.

New materials that today get discovered once a year by accident — quarterly, 5-10 at a time.

The holy grails — like a room-temperature superconductor (Korea's LK-99 moment in 2023) — become answerable by simulation.

Samsung, LG Energy Solution, BMW are already invested in quantum battery simulation.

Bosch signed a 5-year contract with IonQ in 2024.

This is probably the most provocative possibility.

All digital security today rests on RSA-2048 (banking), ECDSA (Bitcoin and Ethereum), and TLS (the lock icon on every website).

These cryptographic systems all sit on one mathematical assumption — 'factoring large numbers takes supercomputers billions of years'.

In 1994, MIT mathematician Peter Shor published an algorithm that lets a quantum computer factor large numbers in seconds.

The industry calls it Shor's Algorithm.

Which means: the moment a sufficiently large quantum computer exists, internet security collapses.

The industry calls this day 'Q-Day'.

Expected timing: between 2030 and 2035.

Before that, every system must migrate to Post-Quantum Cryptography (PQC).

Estimated global migration cost: over $500 billion.

Training a frontier LLM like ChatGPT takes tens of thousands of NVIDIA GPUs running for months.

Cost per training run: hundreds of millions of dollars.

Quantum computers can exponentially accelerate certain ML primitives.

The industry calls these HHL algorithms, Quantum Neural Networks, and so on.

Quantum won't do all of AI training. But if a critical step gets 1000x faster, total training time collapses from days to minutes.

If that turns real, some of NVIDIA's value migrates to the quantum names.

That's why quantum stocks suddenly matter as the AI cycle starts to feel mature.

JPMorgan, Goldman Sachs, HSBC, BBVA.

Every global megabank now runs a quantum research team.

Why?

- Portfolio optimization: picking the best combination from 10,000 assets → exponential quantum speedup
- Options pricing: Monte Carlo simulation → 1000x quantum acceleration
- Fraud detection: graph search → √N speedup via Grover's algorithm
- Credit risk: enormous correlation matrices → natural fit for quantum

JPMorgan runs the world's largest quantum consortium (Q-Initiative).

Goldman has been working with IBM Quantum since 2020.

If the quantum era arrives, part of banking's competitive edge gets re-decided at the quantum infrastructure layer.

Current IPCC climate models are heavily simplified.

The real climate system is molecular chemistry × ocean currents × atmosphere × ice, all quantum-coupled.

Classical computers force enormous simplifications just to make the model run.

Quantum makes much more accurate climate models possible.

Which means: 'what does this policy do in 50 years' becomes a scientifically answerable question.

Carbon capture materials, more efficient solar cells — quantum can design them directly.

This is actually the most important item.

In 1995, when the internet first went commercial, nobody predicted Uber, Instagram, ChatGPT.

Quantum will be the same.

We currently see 6 clear use cases. The real impact will come from places we can't imagine.

That's the pattern of every genuinely foundational technology.

The reason classical computers can't do the 6 things above isn't simply that they're 'too slow'.

It's mathematically impossible.

Understanding this is the key to seeing why quantum isn't just 'a faster computer'.

Computer science divides problems into two categories.

First: polynomial problems.

For input size N, runtime grows as N², N³ — manageable. Sorting 1,000 names is this kind. Faster computers help. Humans solve these well.

Second: exponential problems.

For input size N, runtime grows as 2^N.

Example: visit 1,000 cities in shortest order (the traveling salesman problem). Make computers a trillion times faster — barely matters once N grows.

The numbers are shocking:

- 10 cities → 1 second
- 20 cities → 17 minutes (1,000x slower)
- 30 cities → 12 days
- 40 cities → 35 years
- 50 cities → 35,000 years
- 100 cities → age of the universe × 10^10

A millionfold speedup buys you a handful more cities before you hit the age of the universe again.

This is the exponential wall.

Let's look at exactly why drug discovery is so hard.

Molecules have electrons.

Caffeine has 102 electrons. A simple protein has 10,000+. A protein inside a human cell has hundreds of thousands.

Every electron is quantum-entangled with every other electron. The industry calls this entanglement, which we'll define properly in a moment.

What matters now: to compute a molecule's exact state, a classical computer has to track 2^(number of electrons) states simultaneously.

- 30-electron molecule → about 1 billion states. Supercomputer can handle this
- 50-electron molecule → ~1.1 quadrillion states. Supercomputer takes months
- 70-electron molecule → more states than stars in the universe. Forever impossible

Which is why, in 2025, even caffeine — a small molecule — still can't be exactly simulated by classical computers.

That's why drug discovery has to lean on physical experiments. It has no choice.

Now we walk through how a quantum computer actually works. Seven concepts, built up in order.

Classical computers run on bits.

A bit is either 0 or 1. One or the other.

Like a piece of chalk lying flat (0) or standing upright (1) on a desk.

The quantum computer's basic unit is a qubit (quantum bit).

The essence of a qubit in one sentence: "a bit where 0 and 1 can exist simultaneously".

The chalk spinning at 45° between flat and upright. The instant you measure, it resolves into flat or upright.

This sounds nonsensical — that's normal. It isn't part of everyday experience.

But in quantum mechanics, this has been experimentally verified for 100 years.

Quantum particles exist in multiple states simultaneously until measured.

This is called superposition.

1 qubit = 2 states held simultaneously (0 and 1).

2 qubits = 4 states (00, 01, 10, 11).

3 qubits = 8 states.

N qubits = 2^N states.

The numbers grow terrifyingly.

- 10 qubits = 1,024 states processed at once
- 50 qubits = ~1.1 quadrillion states
- 100 qubits = ~10^30 states (more than stars in the universe)
- 300 qubits = more states than atoms in the universe

This is quantum's real power.

Not a faster classical computer — a fundamentally different dimension of computation.

With a catch: you can't directly access all the states a qubit holds. Measuring collapses it to just one.

So quantum algorithm design is the art of engineering the system so that the correct answer is the most likely one to survive measurement. Genuinely hard.

Which is why only a handful of meaningful quantum algorithms have been invented in 50 years.

The qubit's superposition is extremely fragile.

Analogy: you're trying to balance a coin upright on your palm in a library. Someone coughs, the AC turns on, a book drops — any disturbance topples the coin.

The qubit is that coin.

Ambient temperature, electromagnetic noise, cosmic rays, a magnet in the next room — anything that touches the qubit collapses the superposition.

This is called decoherence. It's the biggest reason quantum computing remained theoretical for 50 years.

Quantum computers therefore operate near absolute zero (-273.15°C — colder than outer space).

Those photos of IBM quantum systems — the giant gold chandeliers — that's almost entirely cooling and shielding hardware.

The chip actually doing computation is the fingernail-sized piece at the very bottom. 90% infrastructure, 10% quantum chip.

Headlines say things like 'IBM unveiled a 1,121-qubit chip'.

What does that mean?

On the surface: more qubits = stronger quantum computer. 100 qubits handle exponentially more states than 50 qubits.

That's why IBM, Google, and Atom Computing race on qubit count.

But there's a trap: qubits are fragile. 1,000 qubits where 999 of them error out is useless.

Just as important as how many qubits — how long and how accurately they operate (gate fidelity).

Analogy: 1 TB of RAM is meaningless if data corrupts every second.

What the industry watches alongside qubit count:

- Gate fidelity — closer to 1.0 is better. Current best: 0.999
- Coherence time — how long a qubit holds state before collapsing. Microseconds.
- Connectivity — can every qubit interact with every other?

Everyone knows qubits are fragile.

So for 50 years, the real goal of quantum computing has been: how do you correct errors?

The core idea: bundle many fragile physical qubits into one stable logical qubit.

In theory: 100-1,000 physical qubits = 1 logical qubit.

Logical qubits are vastly more stable than physical qubits.

For meaningful quantum algorithms, you need hundreds of logical qubits.

Meaning: hundreds of thousands of physical qubits.

Where are we today? Zero logical qubits.

In December 2024, Google's Willow chip was the first to experimentally prove that bundling more physical qubits reduces error exponentially.

Meaning: the path to logical qubits is demonstrably possible in principle.

It's considered one of the most important experiments in quantum computing history.

Quantum computing splits into two levels.

NISQ (Noisy Intermediate-Scale Quantum)

- 50–1,000 qubits
- No error correction (all physical qubits)
- Noise prevents useful problems from being solved
- Where we are today
- Term coined by Caltech's John Preskill (2018)

Fault-Tolerant Quantum Computing (FTQC)

- Hundreds–thousands of logical qubits
- Errors are automatically corrected
- Can solve genuinely valuable problems — drug discovery, materials, cryptanalysis, optimization
- Every quantum company's ultimate goal
- IBM targets achievement by 2029 with its Starling system

Plain English: NISQ is demo-grade, FTQC is production-grade.

In 2025, we're in the final phase of NISQ and stepping into the threshold of FTQC.

Willow showed us that doorway for the first time.

These two terms get confused in headlines.

Quantum Supremacy: a quantum computer solves one problem that classical computers practically can't. The problem doesn't need to be useful.

In 2019, Google's Sycamore chip (53 qubits) claimed to solve a 53-qubit circuit sampling problem in 200 seconds. They argued it would take supercomputers 10,000 years.

IBM countered with '2.5 days, actually.' But the moment was symbolically pivotal.

The problem solved had zero everyday utility — it was a contrived math problem.

Quantum Advantage: a quantum computer solves a genuinely valuable problem faster or cheaper than classical.

Nobody has achieved this as of 2025.

The moment it happens, quantum's commercialization begins.

The sequence: Quantum Supremacy (2019) → larger NISQ systems (2020-2025) → Fault-Tolerance (2025-2030) → Quantum Advantage (2030+) → full commercialization.

Now that we have the vocabulary, the recent milestones in the industry can be read with their actual meaning intact.

First claim of Quantum Supremacy.

Low qubit count — but historically symbolic.

Google published in Nature: 'solved in 200 seconds a calculation that would take supercomputers 10,000 years.'

IBM countered with '2.5 days on our supercomputer.' But the moment was the first time quantum showed measurable advantage of any kind.

The problem solved had zero utility — random circuit sampling. But as a proof of principle, it was huge.

Largest qubit count in human history (at the time).

IBM packed 1,121 qubits onto a single chip.

But gate fidelity was low — meaning it was big but inaccurate.

This announcement made the industry start to see the limits of the qubit-count race.

IBM themselves pivoted to quality-first with Heron (156 qubits, 0.997 fidelity).

One of the most important single experiments in quantum history.

First experimental proof that bundling more physical qubits exponentially reduces error.

Meaning: it demonstrated the principle viability of error correction.

It opened the doorway to fault-tolerance.

IonQ stock jumped 40% in the week of the announcement. The QTUM quantum ETF spiked 25%.

The moment the market first repriced quantum as 'real'.

A completely different type of qubit: 'topological'.

The first verifiable result of Microsoft's 25-year bet.

Uses Majorana fermions — quantum states theoretically immune to noise.

If real, it's a game-changer — error correction overhead drops exponentially.

Academia is skeptical. In 2018, Microsoft published a similar result that was later retracted.

The 2025 announcement comes with more robust data, but verification takes time.

If it holds, the entire quantum cycle could be pulled forward by 5 years.

There isn't one way to build a quantum computer.

Like VHS vs Betamax, nobody knows which approach will dominate.

Six approaches are currently competing. Here are their strengths, weaknesses, and key players.

① Superconducting — best known. IBM went from 5 qubits to 1,121 in six years. A Moore's-Law-like curve. Google Willow uses this too.

② Trapped Ion — IonQ's path. Fighting the market with the message 'quality matters more than count.' Coined algorithmic qubit: '1 IonQ qubit = 50 superconducting qubits.'

③ Neutral Atom — the fastest-rising approach of 2024. Atom Computing's 1,180-qubit system leads the world by qubit count. QuEra, co-founded by Harvard's Mikhail Lukin, is the other key player.

④ Photonic — PsiQuantum's bet. Room temperature. 'Adding qubits one at a time is pointless. We'll build a 1-million-qubit fault-tolerant system in one shot.' Risky but game-over if it works. $617M Australian government contract in 2024.

⑤ Topological — Microsoft's 25-year bet. Mathematically noise-immune. But the Majorana particle itself remains unverified. Microsoft has been on this path for 25 years and previously retracted a paper.

⑥ Silicon Spin — Intel's dark horse. Quietest, possibly smartest. If silicon spin works, the infrastructure to produce tens of billions of qubits is already in place.

So far we've looked at what becomes possible when quantum reaches the threshold.

But until then, what are these companies actually selling to stay alive?

Five business models.

QaaS = Quantum-as-a-Service.

In plain English: rent a quantum computer from the cloud.

Go to IBM's website right now — anyone can use a quantum chip for free for an hour.

Enterprises pay $1,000-$10,000/hour to rent quantum resources.

Three platforms dominate:

- IBM Quantum — oldest, 600,000+ registered users globally
- Microsoft Azure Quantum — aggregates IonQ, Quantinuum, Rigetti, etc.
- AWS Braket — aggregates IonQ, Rigetti, D-Wave

The quantum chip suppliers to these three platforms are companies like IonQ, Quantinuum, Rigetti, D-Wave.

Which means: most quantum company revenue arrives via AWS or Azure — not direct.

US NSA, DARPA, UK MOD, Australian government, Korea's KQI.

These entities place quantum orders in the billions of dollars.

Two reasons.

First: defense preparation for Q-Day (the bitcoin-breaking day mentioned above).

Second: military R&D, especially new materials.

PsiQuantum signed $617M with the Australian government in 2024 — the largest single quantum contract in history.

IonQ signed $54M with the US Air Force in 2024, and additional contracts with US Cyber Command in 2025.

30-50% of public quantum company revenue comes from government contracts.

Roche, Pfizer, Merck, JPMorgan, Goldman Sachs, BMW, Bosch, Samsung.

These companies sign 5- and 10-year R&D partnerships with IBM Quantum or IonQ.

Annual fees: $1M-$10M.

The goal: be the first user when quantum hits the threshold.

There's no ROI today, but they start the race ahead of competitors.

JPMorgan's Q-Initiative alone has 25+ global megabanks as members.

Making quantum chips is insanely hard.

But the software running on top can be sold like normal cloud SaaS.

Players:

- Classiq (Israel) — quantum algorithm auto-generation platform. $33M Series C
- Zapata Computing — quantum ML middleware. Public via SPAC
- QC Ware — consulting + algorithms. Goldman Sachs is an investor

They make money in the quantum ecosystem without building chips.

Like NVIDIA built GPUs and OpenAI built ChatGPT on top.

Accenture, Deloitte, BCG, McKinsey.

They don't make chips. They don't write algorithms. And they're earning quantum revenue the fastest.

Their job: advise enterprises on quantum-readiness strategy.

Accenture's quantum consulting practice did an estimated $200M+ in 2024.

Looks bubbly — but it's real revenue.

Remember why NVIDIA grew so large in the AI cycle.

Every AI lab — OpenAI, Anthropic, Google — had to buy GPUs from one company.

NVIDIA sat at the bottleneck of the value chain.

Quantum is highly likely to repeat this pattern.

Let's map the quantum value chain.

Bottleneck #1: Bluefors (Dilution Refrigerators) — Quantum's ASML

This is the hidden gem of the quantum industry.

Dilution refrigerators are the massive machines that create temperatures near absolute zero (-273°C).

Every superconducting quantum computer (IBM, Google, Rigetti) must use one.

Globally, only two companies make them in volume.

Bluefors (Finland) and Oxford Instruments (UK).

Bluefors holds about 70% of the market.

Price per system: $500K-$3M.

Demand is exploding. Every new quantum system means revenue for Bluefors.

Bluefors is private. Estimated revenue $200M+, profitable.

If they IPO — this could be the ASML of the quantum era.

Bottleneck #2: Optical Components (Coherent, Lumentum)

Ion trap (IonQ, Quantinuum) and photonic (PsiQuantum) approaches all need precision lasers.

Coherent ($COHR, ~$13B market cap) and Lumentum ($LITE, ~$5B) together hold 80% of this market.

These companies also earn from 5G telecom, datacenter optical interconnect, and medical devices.

Which means: even if quantum collapses, the core business survives. Safer bet.

Bottleneck #3: Quantum Machines (Israel) — Quantum Control Systems

Manipulating qubits requires precision microwave pulses.

The systems that generate these are called Quantum Control Systems.

Israeli company Quantum Machines holds 60-70% of this market.

IBM, Google, AWS, MIT, Caltech, Harvard — virtually every quantum lab uses them.

Private. 2024 revenue $50M+. Series B raised $170M at $1B+ valuation.

Industry nickname: 'Quantum's NVIDIA Toolkit' (analogous to NVIDIA's CUDA).

So how much are these companies actually earning today?

Estimated 2024 revenue and losses for public quantum companies.

EV/Sales multiples of 200-4,000x.

For context: NVIDIA EV/Sales ≈ 30x.

Which means quantum stocks are priced on 2035 possibility, not 2024 revenue.

That's the risk signal — if the threshold isn't reached, 90% drawdowns are on the table.

The flip side — if the threshold becomes visible, they could 10x again.

BigTech quantum divisions (not separately disclosed; industry estimates):

- IBM Quantum — ~$200-400M revenue (government + consortia)
- Google Quantum AI — near zero revenue, 100% R&D investment
- Microsoft Quantum — near zero revenue, some Azure Quantum

Private quantum companies (most active):

- Quantinuum (Honeywell + Cambridge Quantum) — ~$20-50M revenue, $5B valuation
- PsiQuantum — near zero revenue (R&D phase), $4B+ valuation
- Atom Computing — near zero revenue, Series B $60M

Private pick-and-shovel plays (most interesting):

- Bluefors — ~$200M+ revenue, profitable, IPO candidate
- Quantum Machines — ~$50M+ revenue, $1B+ valuation

Multiple consulting firms have estimated the quantum market.

The range across estimates is huge — which itself tells you how uncertain this is.

For reference:

- 2024 AI market: $200B
- 2024 cloud market: $700B
- 2024 global semiconductor market: $600B

So BCG's conservative case = 2040 quantum = 25% of 2024 semis.

McKinsey's bull case = 2035 quantum = 2x the 2024 cloud market.

The truth lies somewhere in between.

Separately — the PQC migration market is much more certain.

NIST estimate: $500B+ global migration cost.

Beneficiaries: Cloudflare, Palo Alto Networks, IBM (PQC standard adoption).

This cost will be incurred whether quantum succeeds or fails.

Scenario A: BigTech Wins (IBM/Google/Microsoft) — 40% probability

Most likely.

BigTech reaches the threshold first, thanks to capital depth.

Pure plays are technically ahead but capital-starved.

In this case, IonQ/Rigetti/D-Wave get acquired or fade.

Scenario B: Pure Play Wins (IonQ/PsiQuantum) — 20% probability

A pure play reaches fault-tolerance first.

BigTech acquires or licenses.

IonQ/PsiQuantum stocks go 100x.

Scenario C: Pick-and-Shovel Wins (Bluefors/Coherent/TSMC) — 30% probability

Historically the most frequent outcome.

Whoever reaches the threshold — they still have to buy the components.

Bluefors IPO becomes quantum's ASML.

Coherent/Lumentum are quantum + 5G + AI multi-bets.

Scenario D: Permanent Bubble (Cisco's Lost Decade) — 10% probability

Quantum is technically possible but commercially failed.

NISQ plateau persists indefinitely.

In this scenario, quantum stocks drop 90%.

Five signals to watch over the next 1-2 years:

1. Logical qubit count — going from 1 to 10 is the threshold
2. BigTech acquisition announcements — cycle-start signal
3. Bluefors IPO — biggest pick-and-shovel opportunity
4. NIST PQC standards mainstream adoption — separate $500B market kicking off
5. First case of quantum vs classical cost advantage — Quantum Advantage begins

Quantum computers are built by physicists, commercialized by engineers.

So you need to know the players on both the academic and industrial sides.

Peter Shor (MIT) — Discovered Shor's Algorithm in 1994. The quantum algorithm that breaks RSA. He predicted the first thing that would happen when quantum computers exist, 25 years before it could.

John Preskill (Caltech) — Coined NISQ. The scholar who defined quantum computing's current era. His annual talks set the state-of-play compass for the industry.

Mikhail Lukin (Harvard) — Pioneer of neutral atom approach. Co-founded QuEra. His 2023 Science paper validated neutral atom's potential.

Pan Jianwei (潘建偉, USTC) — China's quantum godfather. First in the world to demonstrate satellite-to-ground quantum communication. Symbolic figure of China's quantum ambitions.

Alain Aspect + Anton Zeilinger — 2022 Nobel Prize in Physics. Experimentally proved that quantum entanglement is real. Theoretical foundation of quantum communication.

Hartmut Neven (Google Quantum AI) — Leads Google's quantum division. Drove both 2019 Sycamore and 2024 Willow. Symbol of technical leadership in the industry.

Jay Gambetta (IBM Quantum) — IBM Quantum VP. Publishes IBM's quantum roadmap annually. His announcements function as the industry's official timetable.

Krysta Svore (Microsoft Quantum) — Leads Microsoft Quantum. Responsible for the 25-year topological qubit bet. Led the Majorana 1 announcement.

Chad Rigetti (Rigetti) — Founder of Rigetti. Berkeley PhD, then IBM, then founded. First SPAC IPO for pure-play quantum.

Christopher Monroe (IonQ) — IonQ co-founder, University of Maryland professor. The academia-industry bridge for trapped ion.

Jeremy O'Brien (PsiQuantum) — PsiQuantum CEO, Bristol-trained. Leading the most ambitious photonic quantum bet.

Peter Chapman (former IonQ CEO) — From Amazon. Led IonQ through IPO. Stepped down in 2024; Niccolo de Masi is the current CEO.

Quantum is also a nation-state competition.

Where AI is US-China bipolar, quantum is US-China-EU-Japan four-pole.

Government investment by country + key hubs:

US-China quantum decoupling is underway.

Starting in 2024, the US restricted exports of quantum-related technology to China.

China is building its own supply chain (HBM was SK Hynix-dependent, but quantum is going closed-loop domestically).

How this reshapes the industry is still unclear.

What's clear: quantum tech is becoming a geopolitical asset faster than AI did.

We've repeated 'the real alpha is the supply chain'.

But we haven't explained how Lumentum, Coherent, TSMC, Bluefors are actually exposed to quantum, what they earn from today, and what bottlenecks they'll hold in the cycle.

This section unpacks that. For each company — business model / quantum linkage / current bottleneck / future quantum bottleneck / financials / Wall Street consensus.

## ① Lumentum Holdings ($LITE) — Precision Laser Supplier

One-liner: Exposed on both sides — AI datacenter optical transceivers + precision lasers for trapped-ion quantum computers.

Spun off from JDSU in 2015. Two segments:

- Cloud & Networking (~60%+ of FY26 revenue): 400G / 800G / 1.6T optical transceivers. Customers: NVIDIA, Google, Meta, Microsoft, Amazon — all AI datacenter.
- Industrial Tech (~40%): Industrial lasers, EUV light sources, Apple iPhone Face ID 3D sensing.

Acquired Cloud Light for $750M in November 2023 — securing direct 800G/1.6T transceiver supply chain.

Lumentum's components are essential in 3 of 6 quantum approaches:

- Trapped Ion (IonQ, Quantinuum) — ions need extremely precise lasers (729nm Yb, 422nm Sr, etc.).
- Neutral Atom (Atom Computing, QuEra) — optical tweezers using precision lasers to hold atoms.
- Photonic (PsiQuantum) — InP wafer tech is core for photonic quantum.

Coherent + Lumentum together hold ~80% share in quantum-grade precision lasers.

Wall Street view: Wells Fargo, Citi, Morgan Stanley mostly Buy / Overweight.

Key debates — 1.6T transition pace / China exposure / quantum revenue recognition timing.

Thesis one-liner: The 'pickaxe' of AI DC + the 'bonus' of quantum cycle. Core business survives if quantum fails.

## ② Coherent Corp ($COHR) — Vertically Integrated Optics + Materials

One-liner: Lumentum's rival, and larger. Similar quantum exposure but stronger on the materials science side (SiC, InP wafers).

Formed when II-VI Inc acquired Coherent for $7B in 2022.

Three segments:

- Networking (~50%): Optical transceivers, InP devices — direct Lumentum competitor
- Materials (~25%): Engineered materials, SiC substrates (for EV power semiconductors)
- Lasers (~25%): Industrial lasers, semiconductor laser systems

CEO Change (June 2024): Jim Anderson (former Lattice Semi CEO) joined. Aggressive restructuring — divesting non-core, accelerating 1.6T production.

Similar path to Lumentum, plus one differentiator:

- InP wafer vertical integration — Coherent owns the full stack from InP wafer to transceiver. PsiQuantum's photonic systems need InP.
- Precision lasers: Supplied to trapped-ion systems (IonQ, Quantinuum)
- SiC substrates: Some usage in cryogenic CMOS for quantum control electronics

Wall Street view: Mostly Buy / Overweight post-CEO change (Jim Anderson, June 2024).

Key debates — SiC market slowdown (EV) vs 1.6T share / Lumentum competition.

Net Debt is high (II-VI merger leverage) — rapidly deleveraging.

Thesis one-liner: More diversified than Lumentum. SiC may drag from EV slowdown, but AI + quantum exposure is equal.

## ③ TSMC ($TSM) — The Final Destination of All Leading-Edge Silicon

One-liner: Absolute winner of AI cycle. Quantum exposure is indirect — but unbreakable in any scenario.

World's largest pure-play foundry. Manufactures 60%+ of all leading-edge semiconductors.

Key customers: Apple (largest), NVIDIA, AMD, Qualcomm, Broadcom, Marvell, Intel (yes, even Intel uses TSMC for some chips).

Process node roadmap:
- 3nm: In production (NVIDIA H200, GB200, Apple M4)
- 2nm: Production starting 2026 (Taiwan + Arizona fabs)
- 1.6nm: 2027-2028 roadmap

Revenue mix (2025):
- HPC (incl. AI): ~50%+
- Smartphones: ~30%
- IoT, auto, others: ~20%

TSMC's role in quantum is direct but not central:

- Quantum chip foundry: IBM uses some in-house + some TSMC. Rigetti uses TSMC. Most new quantum startups use TSMC.
- CoWoS advanced packaging: Same packaging used in AI chips. Also used in quantum control systems. CoWoS is TSMC-exclusive.
- Cryogenic CMOS: Quantum control chips operating at ultra-low temperatures — TSMC's specialty process.
- Hybrid quantum-classical chips: Quantum chip + classical interface — both need leading-edge fab.

But quantum is <1% of TSMC's revenue. TSMC's main game is AI.

Wall Street view: Near-universal Buy / Overweight — no controversy.

Key debates — China geopolitical risk / 2nm yield ramp / US fab progress.

2nm production starts H2 2026 (Apple A20 & NVIDIA Rubin).

Thesis one-liner: Whether AI or quantum wins, TSMC makes the chips either way. Single common winner. Only risk = Taiwan geopolitics.

## ④ Bluefors (Private, Finland) — Quantum's ASML

One-liner: 70% share in the essential component for every superconducting quantum computer. The single biggest bet of the quantum era if it IPOs.

Founded 2008 in Helsinki, Finland. Nearly single-product revenue.

Product: Dilution Refrigerator

- Creates near-absolute-zero (-273°C) environment
- The giant gold chandeliers in IBM/Google quantum system photos — that's this company
- Price per unit: $500K - $3M
- Manufacturing lead time: 12-18 months

Every superconducting quantum computer must use one.

- IBM Quantum — every system
- Google Quantum AI — Sycamore through Willow
- Rigetti, Quantinuum (gate model portion) — Bluefors
- Academic labs (MIT, Caltech, Oxford, USTC, KIST, etc.) — All Bluefors
- Quantum sensing, dark matter detection also use the same equipment

Global market share ~70%. Competitors:
- Oxford Instruments (UK, ~25%)
- Others (~5%)

Why "Quantum's ASML":

Just as ASML monopolizes leading-edge EUV lithography — forcing TSMC, Samsung, and Intel to all buy from ASML —

Bluefors holds 70% share of the essential component for every superconducting quantum computer.

Whether IBM ships 1,000-qubit chips or Google scales to 100,000 — Bluefors sells that many refrigerators.

Thesis one-liner: If it IPOs, this is the ASML of the quantum era. Direct exposure impossible until then. Monitor for IPO announcement.

## Summary — Which Company for Which Investor

Too many terms appeared in this article. You don't need to memorize them all on first read. This is the place to look up anything that tripped you up.

5 categories — Quantum Basics / Quantum Hardware / Quantum Stages / Security / Semiconductors + Optics.

Before closing, three honest takes on quantum computing.

The tech works. Willow proved that.

But commercialization is 5-10 years out.

The internet worked in 1995, but it took Amazon 8 years to turn a profit.

Quantum will likely follow a similar curve.

EV/Sales of 200-4000x is very dangerous.

A delayed threshold announcement could trigger 50-90% drawdowns.

But the moment the threshold becomes real and visible, the same stocks can 10-100x.

So — short-term trading is dangerous; 5+ year holding makes it an asymmetric bet.

History keeps saying so.

A Bluefors IPO in 2026-2027 is the most likely single catalyst.

Beyond that, Coherent ($COHR), Lumentum ($LITE), and TSMC remain safer bets — their core businesses survive even if quantum fails.