開發者的量子計算指南：哪些是真實進展、哪些是炒作，以及現在可以做甚麼

Somewhere right now, someone is storing encrypted traffic they can't read. They aren't cracking it. They're waiting for a machine that doesn't fully exist yet.

It's called "harvest now, decrypt later," and it's one reason quantum computing has moved from a physics-department curiosity to something developers need to understand.

I'll skip the "a qubit is 0 and 1 at the same time" explanation and tell you what's real, what's hype, and what you can do about it.

A short history: from a lecture to a hardware race

• 1981: Richard Feynman argues that simulating nature efficiently needs computers that obey quantum mechanics themselves.
• 1985: David Deutsch describes a universal quantum computer.
• 1994: Peter Shor publishes an algorithm that factors large numbers efficiently on a quantum computer. Suddenly RSA has a theoretical expiry date, and governments pay attention.
• 1996: Lov Grover shows a quadratic speedup for unstructured search.
• 2019: Google claims "quantum supremacy" with its Sycamore chip. Critics argue the task was contrived, but it proved the hardware could do something classical machines struggle with.
• 2021-2023: IBM pushes qubit counts up (Eagle at 127, then Condor at over 1,000). The industry learns that more qubits don't help if they're noisy.
• Dec 2024: Google's Willow chip shows error rates falling as the error-correcting code grows. Researchers had chased this "below threshold" result for about 30 years.
• 2025-2026: The focus shifts from qubit counts to logical qubits (error-corrected ones), with new systems like IBM's Nighthawk and Loon processors and Quantinuum's Helios.

How it works (the 60-second version)

A classical bit is 0 or 1. A qubit can be in a weighted blend of both, and qubits can be entangled, so their states are linked.

A quantum algorithm doesn't "try every answer at once." It arranges the qubits so wrong answers cancel out (destructive interference) and right answers reinforce (constructive interference). When you measure, you're likely to get the right one.

That's why quantum computers won't speed up your CRUD app. They're good at a narrow set of problems.

Is quantum computing a competitor to AI?

No. They solve different problems.

They're more likely to be teammates. Quantum hardware could someday help generate better training data (for example, molecular simulations), and AI is already used to decode errors on quantum chips. Nvidia is investing in tools that link GPUs and quantum processors for exactly this hybrid approach.

What's it good for? Real use cases

1. Chemistry and materials: Simulating molecules is the most natural fit. Think better batteries, catalysts, and fertilizer production.
2. Drug discovery: Modeling how molecules bind, which is brutally hard classically.
3. Cryptography: Shor's algorithm breaks RSA and elliptic-curve cryptography at scale. This is the one that affects your job today.
4. Optimization: Logistics, scheduling, and portfolio problems might benefit, but this is the most hyped and least proven area.
5. Quantum sensing and networking: More mature than computing in some cases, and already useful for precise measurement.

Not on the list: running your web app, training LLMs faster, or replacing your laptop.

Advantages vs. disadvantages

Advantages

• Exponential or large speedups for specific problems
• Can simulate nature in ways classical machines can't
• Opens new research in chemistry and physics

Disadvantages

• Qubits are fragile (heat, vibration, and stray radiation destroy them)
• Needs massive error correction: many physical qubits per logical qubit
• Expensive, power-hungry, and operationally complex
• Narrow usefulness
• Immature software and a shortage of talent

What does it take to build one?

This is the part most intros skip.

• Superconducting qubits (IBM, Google): Chips sit in a dilution refrigerator at around 10-15 millikelvin, colder than outer space. You need cryogenics, microwave control electronics, and miles of precisely engineered wiring.
• Trapped ions (Quantinuum, IonQ): Individual charged atoms held in vacuum by electromagnetic fields and manipulated with lasers. Slower, but very accurate.
• Neutral atoms (QuEra, Pasqal, Atom Computing): Atoms held by laser "tweezers," which scale to large arrays well.
• Photonic (PsiQuantum, Xanadu): Qubits made from light, which can run at warmer temperatures but need huge amounts of precision optics.
• Classical machines everywhere: Every quantum computer is surrounded by fast classical hardware doing control, calibration, and real-time error decoding. This is why "quantum + HPC" is the real architecture.

Who's building it, and what have they shown?

• IBM: Public roadmap targeting quantum advantage by the end of 2026 and fault-tolerant machines by 2029. Qiskit is the most widely used SDK.
• Google Quantum AI: Willow's below-threshold error correction in 2024, plus more verifiable-advantage experiments since.
• Quantinuum: Trapped-ion leader, launched Helios in late 2025 and went public in mid-2026. A Nature paper with Microsoft in June 2026 reported logical qubits with much lower error rates than physical ones.
• Microsoft: Pursuing topological qubits (the claims are debated in the community) and partnering on logical-qubit work.
• IonQ, Rigetti, QuEra, Pasqal, PsiQuantum, Atom Computing: Each betting on a different hardware approach.
• China: Companies like Origin Quantum and university groups are building their own superconducting systems.

The honest takeaway: in 2026 we're finally measuring progress in error-corrected logical qubits, not marketing-friendly raw qubit counts.

Challenges that still matter

1. Error correction at scale. A useful machine likely needs thousands of logical qubits.
2. Scaling hardware without scaling noise, heat, and wiring.
3. Finding real advantage. Many "quantum wins" get matched later by smarter classical algorithms.
4. Software. Turning a real business problem into a quantum circuit is still hard research.
5. Honest timelines. Credible players say fault tolerance is a late-2020s to 2030s story. Be wary of anyone promising dates with certainty.

Why developers should care now: the crypto problem

If an adversary records encrypted data today and decrypts it later, anything with a long shelf life is at risk: health records, legal files, state secrets, anything you sign or encrypt that must stay trustworthy.

NIST finalized its first post-quantum standards in August 2024:

• ML-KEM (Kyber) for key exchange
• ML-DSA (Dilithium) for signatures
• SLH-DSA (SPHINCS+) as a hash-based signature backup

Browsers, Cloudflare, and Signal already use hybrid key exchange (classical + post-quantum).

What to do this quarter

1. Inventory your crypto. Find RSA, ECDSA, and DH in TLS configs, JWTs, SSH keys, VPNs, and code signing.
2. Build crypto-agility. Algorithms should be a config change, not a rewrite.
3. Prioritize by data lifespan. Ten-year secrets first.
4. Enable hybrid key exchange where your stack supports it.
5. Try a quantum circuit. Install Qiskit and build a Bell state:

from qiskit import QuantumCircuit
from qiskit.primitives import StatevectorSampler

qc = QuantumCircuit(2, 2)
qc.h(0)          # put qubit 0 in superposition
qc.cx(0, 1)      # entangle qubit 1 with qubit 0
qc.measure([0, 1], [0, 1])

result = StatevectorSampler().run([qc], shots=1000).result()
print(result[0].data.c.get_counts())
# Expect roughly half '00' and half '11', never '01' or '10'

That 50/50 split of only 00 and 11 is entanglement in about ten lines.

Final thoughts

Quantum computing isn't going to replace your laptop or your favorite AI model. It's a specialized tool that will probably reshape chemistry, materials science, and security first.

You don't need a physics degree. You need to know where the tech is real, where it's marketing, and which of your systems depend on cryptography that has a shelf life.

What are you most curious (or skeptical) about? Tell me in the comments. I'd love to hear whether you've tried Qiskit or started a crypto inventory at work.

Disclosure: I used AI assistance to help draft and structure this post, then reviewed and edited it myself. Please verify technical claims against the primary sources linked below.

Sources and further reading

• NIST Post-Quantum Cryptography project (nist.gov)
• IBM Quantum roadmap and Qiskit docs
• Google Quantum AI, Willow announcement (2024)
• Quantinuum research and Nature publications