One hundred thousand qubits can crack RSA-2048. That's the bombshell from Iceberg Quantum — an Australian startup claiming their "Pinnacle" architecture makes the cryptographic apocalypse ten times easier than anyone expected. Which means in 2026, our encryption is far more vulnerable than we imagined.
The idea isn't new. Google managed to slash requirements from 170 million to 20 million quantum bits back in 2019. Then in 2025, Craig Gidney dropped the number below one million. Now Paul Webster's team in Sydney claims they can do it with 100,000. If they're right, "Q-Day" — the moment quantum computers obliterate digital security — just got a lot closer.
📖 Read more: China Breaks Fusion's 36-Year Density Barrier with EAST
🔬 The Pinnacle Architecture and QLDPC Codes
The trick lies in Quantum Low-Density Parity Check (QLDPC) codes. Until now, most predictions relied on surface codes — an error correction method that needs hundreds or thousands of physical qubits for every logical qubit.
Think of the difference this way: surface codes protect data by connecting every element in a dense network — reliable, but heavy and hardware-intensive. QLDPC codes achieve the same protection with far fewer connections per qubit. Like a sparse network that still catches errors, but uses much less hardware.
Iceberg Quantum structures Pinnacle from modules they call "processing units," "magic engines," and optional "memory blocks." Each processing unit contains QLDPC code blocks protecting logical qubits, plus measurement hardware enabling random logical Pauli measurements during each correction cycle.
Magic States and Production Pipeline
Here's the clever part. The "magic engine" produces one encoded magic state per logical cycle while simultaneously consuming one prepared in the previous cycle. This pipeline maintains steady performance without dramatically increasing hardware overhead.
The result? A completely different approach to fault-tolerant quantum computers. Instead of building massive, densely connected qubit networks, Pinnacle uses sparser but more intelligently organized connections.
📖 Read more: 15-Minute Cities: Everything Within Walking Distance
📊 Benchmarks and Real Performance
To test the architecture, Webster's team analyzed two benchmark problems. First, they estimated ground state energy of the Fermi-Hubbard model — a widely studied problem in condensed matter physics.
For a 16×16 lattice with 256 sites simulating interacting electrons, the study reports roughly 62,000 physical qubits would suffice at a 10⁻³ physical error rate. Compare that to previous surface-code predictions of 940,000 qubits. More than a ten-fold difference.
RSA-2048 Numbers: With 10⁻³ physical error rates and 1-microsecond code cycle time, Pinnacle needs fewer than 100,000 physical qubits to break RSA-2048. On slower hardware with millisecond-scale cycles, it would require about 3.1 million qubits for the same task — but then the process takes a month.
Clifford Frame Cleaning
One innovation boosting this flexibility is what the team calls "Clifford frame cleaning" — a technique letting processing units merge and later separate with limited additional logical measurements. This supports selective parallelism without fully entangling the entire machine's logical state.
Sounds technical — and it is. But the essence is simple: instead of needing one massive, monolithic quantum computer, you can build smaller modules that work together when needed.
📖 Read more: 3D Organ Printing: Transplants in Hours
⚡ Cryptographic Impact and Q-Day
RSA-2048 forms the backbone of much of today's public-key cryptography. While most experts expect post-quantum cryptographic schemes to replace RSA before large-scale quantum computers arrive, hardware timelines remain uncertain.
If fewer than 100,000 physical qubits truly suffice to break RSA-2048 under realistic error models, the threshold for cryptographic risk could arrive sooner than surface-code estimates suggest.
"Implementing utility-scale quantum computing necessarily depends on designing practical, low-overhead fault-tolerant architectures."
— Iceberg Quantum Research Team
Of course, the result comes with conditions. You need sustained physical error rates at one mistake per thousand operations, efficient QLDPC decoding in real time, and modular hardware integration at scale.
What This Means for Business
Iceberg Quantum's predictions change the cybersecurity landscape in 2026. Companies relying on RSA to protect their data need to reassess their timelines.
The transition to post-quantum cryptography isn't a "someday in the future" issue anymore. It's happening now. And if Pinnacle works as promised, 100,000 qubits becomes a much more achievable target than the million we expected.
📖 Read more: 3D-Printed Houses: Housing in 24 Hours
🧬 Limitations and Next Steps
The analysis relies entirely on numerical simulations and theoretical compilation — no experimental demonstration backs these claims. Logical error rates come from fitted models, and the decoding method used in simulation may not reflect practical, low-latency decoders required for real machines.
Magic state distillation, while optimized in the architecture, remains the dominant cost. Rejection rates of distilled states increase at higher physical error rates, requiring idle cycles and slightly increasing logical depth.
Hardware Challenge
While qubit count dropped, building and stabilizing 100,000 high-fidelity qubits with microsecond-scale cycles remains far beyond current hardware.
Real-Time Decoding
The analysis doesn't cover decoder speed for real-time hardware control — a critical element for practical applications.
There's also a fundamental uncertainty: the study was published on arXiv without peer review. That doesn't mean it's wrong, but the scientific community hasn't thoroughly examined the calculations yet.
What Comes Next
Iceberg Quantum secured $6 million in seed funding and is expanding globally with a new Berlin office. The company plans to partner with hardware providers to test their architecture in practical applications.
Meanwhile, QLDPC technologies are gaining ground. It's not just Iceberg working on them — research teams worldwide see quantum low-density parity check codes as a more efficient solution than traditional surface codes.
🎯 Frequently Asked Questions
When will we see practical 100,000-qubit quantum computers?
Current predictions point to the late 2020s. IBM targets 100,000 qubits by 2033, but if the Pinnacle architecture works, we might see cryptographically relevant quantum computers sooner.
Should businesses panic?
No panic, but preparation. The transition to post-quantum cryptography is already underway. NIST has approved the first post-quantum standards, and most major tech companies are gradually adopting them.
Are the results reliable without peer review?
The calculations look solid on paper, but other research teams haven't confirmed them yet. Publishing on preprint servers is common practice for rapid idea exchange, but findings must be confirmed by other research teams.
Iceberg Quantum's promise — 100,000 qubits to break RSA — if verified, will fundamentally change the conversation around quantum threats. We're no longer talking about some indefinite future, but concrete numbers and practical challenges. 2026 looks like the year quantum computers leave the lab and enter our daily security discussions.