Why Shor Won't Kill RSA Overnight: The Engineering Reality
> Prerequisite navigation: You'll want episode 038 (resource estimates), 023 (magic-state distillation), 006 (decoherence), and 030 (the NISQ boundary). > Next up: episode 040, "Two Great Lineages: Amplitude Amplification and the Hidden Subgroup" (wrapping up batch four).
The laws of mechanics long ago proved that a thousand-ton bridge *could* be built across some gorge. But blueprints, budgets, steel, and schedules? None of that exists. When passersby ask "when does the bridge open," the engineer's answer is always the same: laws don't set schedules; engineering does.
"Will Shor kill RSA?" is the same question. Episode 037 covered the math (the principle works; the throne has been forged). Episode 038 covered the cost (millions of qubits, tens of billions of magic states). This post finishes the remaining three topics: exactly where the chasm lies, how to read the progress bar without being fooled, and why cryptographers stopped panicking long ago.
Three Steps Across the Chasm
From today's machines to "able to break RSA-2048" isn't one slope—it's three flights of stairs:
Step one: quantity. Today's best machines: a few hundred to a thousand-plus physical qubits. The requirement: millions. That's three orders of magnitude—in bit-count terms, the entire road from "the first vacuum-tube computer" to "a smartphone."
Step two: quality. Worse, the requirement isn't millions of *today's* qubits—it's millions of physical qubits to sustain a few hundred error-corrected logical qubits (batch six: one logical qubit needs ~1,000 physical qubits as bodyguards, and physical error rates must first be pushed below threshold). The qubits on today's machines are "noisy raw material." Piling up quantity alone doesn't help: Shor's depth-n³ circuit dies on any machine without error correction (a restatement of episode 030's NISQ boundary).
Step three: rations. Episode 038's magic-state accounting: tens of billions of T-gates must be purified in bulk by distillation factories (episode 023)—and a distillation factory is itself a machine that must be designed, validated, and scaled up.
Stack the three steps together and you get the full meaning of episode 038's "three orders of magnitude short": not a 1000× "build more," but a 1000× "build better and sustain."
How to Read the Progress Bar: Count Logical Qubits, Not Headlines
The popular-news progress bar is "how many qubits we have"—and that one lies. The engineering-honest progress bar is the number of logical qubits: error-corrected qubits that can run deep circuits. A few real milestones from 2024–2026 (timeliness note: written from a November 2026 vantage point): the superconducting route demonstrated "error correction that improves as you scale" (the inflection point where adding qubits *lowers* error rates appeared for the first time); the neutral-atom route connected dozens of logical qubits into small networks; IBM's public roadmap commits to delivering around 200 fault-tolerant logical qubits by roughly 2029.
Against the requirement (thousands of high-quality logical qubits to break RSA-2048), the bar is a few thousandths full. Honest conclusion: the probability of breaking RSA-2048 before 2035 is very low; it rises significantly over 2035–2045; the specific year, nobody knows. Engineering history teaches two lessons at once: linear extrapolation is laughable (breakthroughs often come in jumps), but don't treat every step as a permanent cliff either.
Meanwhile, the Other Side Is Already Moving: Cryptography's Bridge Swap
The urgency that actually matters comes from the cryptographic community itself—and their response isn't panic; it's swapping the bridge:
One: the new locks are built and shipping. In 2024, the US standards agency formally published a set of post-quantum cryptography standards (lattice- and hash-based constructions). Their mathematical foundations are not the factoring problem—Shor's phase-estimation machine has no purchase on them. Hybrid schemes (classical + post-quantum double locks) are already deploying in browser and cloud TLS connections—your online shopping traffic today very likely already carries the new locks.
Two: the real deadline is "harvest now, decrypt later." This is the most important puzzle piece in the whole post: ciphertext intercepted and archived today will be replayed and decrypted the day the quantum machine arrives. Even if the machine is ten years out, data that "must stay secret for ten years" (medical records, state secrets, long-horizon trade secrets) should switch locks now—because eavesdroppers don't need to wait for the machine to start archiving. Your personal shopping passwords needn't worry (that data expires and becomes worthless); the hard fight is institutional migration of long-lived sensitive data, and national cryptographic agencies' roadmaps generally target completing migration by 2030–2035.
Three: RSA itself is exiting gracefully. Throughout tech history, most ciphers aren't "killed overnight"—they're replaced by standardization: old algorithms get deprecated, banned, retired, each step on a schedule. Shor's role is to stamp the mathematical final chapter on that schedule: RSA's foundation went from "never been broken" to "proven breakable"—security with a deadline.
Taking the Analogy Apart
What the bridge analogy borrows, and where it breaks:
- Disanalogy one: if a bridge never gets built, you detour; if the cryptographic bridge swap is slow, the cost is failure of stored secrets—and the failure is retroactive (all archived ciphertext is hit). So the urgency on "the other side" comes not from quantum progress but from the act of archiving itself.
- Disanalogy two: there's only one bridge; cryptographic "bridges" are countless and independent—proving the mathematical foundation breakable doesn't mean every bridge collapses at once. Migration is bridge-by-bridge, segment-by-segment engineering. The analogy's "bridge" refers to the mathematical-foundation layer, not to specific systems.
What to Carry Away
1. Three stacked steps: quantity (thousands → millions), quality (noisy → fault-tolerant, ~1,000 physical qubits per logical qubit), rations (magic-state distillation factories). "Three orders of magnitude short" is three steps stacked, not one slope. 2. Read the progress bar in logical qubits (November 2026 vantage: the inflection point has appeared, roadmaps are public, still a few thousandths from breaking RSA-2048). Guard against two errors when predicting dates: don't extrapolate linearly, and don't mistake a step for a cliff. 3. Cryptography already swapped bridges: post-quantum standards shipped in 2024, hybrid encryption is deployed, and the truly hard deadline is harvest-now-decrypt-later—long-lived sensitive data should switch locks now. Shor's role is to stamp RSA's mathematical final chapter.
Hook for the next stop: batch four's finale. Deutsch–Jozsa, Bernstein–Vazirani, Grover, Simon, QFT, Shor—six weapons laid on one table. Time to step back and see the panorama: all quantum algorithms fall into two great lineages—one descended from amplitude amplification, the other from the hidden subgroup. Once you see the family tree, you know which problems quantum computers excel at, and which they'll never help with. Episode 040: "Two Great Lineages: Amplitude Amplification and the Hidden Subgroup."
*From Coins to Error Correction · Quantum Computing Column, episode 039*