For nearly two decades, quantum computing had a reputation as the technology that was always ten years out. Google broke that pattern in December 2024, when its Willow chip did something no quantum processor had done before on real hardware: adding more qubits made the system more accurate, not less.
Why it matters: That single result, called below-threshold error correction, is the property any large-scale, reliable quantum computer will eventually need. It means engineers can keep adding qubits to cancel out errors instead of compounding them. Two other giants have since made their own versions of the same leap, and together they’re pulling quantum computing out of the theoretical column.
Three companies, three different bets on the same finish line
Qubits, the basic unit of quantum information, are notoriously fragile. Heat, vibration, even stray radiation can wreck them faster than any classical computer bit would break down. Below-threshold error correction is the first real proof that this fragility problem is solvable at scale, not just in theory.
Google’s approach: precision first. Prove that a bigger surface code actually reduces the error rate. Willow did exactly that.
IBM’s approach: scale out. Its Nighthawk processor, 120 qubits, is targeting what researchers call verified quantum advantage, a quantum computer beating any classical supercomputer on a genuinely useful problem, by the end of this year. IBM says it hit a tenfold speedup in error correction roughly a year ahead of its own internal schedule.
Microsoft’s approach: something stranger. Its Majorana 1 chip, unveiled in February 2025, is the first processor built on topological qubits, which store quantum information in the global structure of a system instead of in individual particles. It’s less mature than Google’s or IBM’s designs, but if the physics holds up at scale, it could mean inherently lower error rates down the road.
Real money is already showing up
JPMorganChase says it has a working quantum algorithm delivering a theoretical exponential speedup for processing large financial datasets in real time, an early sign finance might be first in line for genuinely practical quantum use, not just research demos. Peer-reviewed papers on quantum error correction jumped from 36 in all of 2024 to more than 120 in just the first ten months of 2025.
Funding hasn’t caught up to revenue yet, though. One analysis found the industry pulled in roughly $2 billion in startup funding recently while generating under $750 million in actual sales, a gap that shows how much of quantum computing’s current value is still a bet on the future.
The uncomfortable side effect
Here’s the part that should worry more people than it currently does: a sufficiently powerful quantum computer would eventually crack the encryption protecting most of the internet, your bank, and classified government data. Security researchers who track the field say Willow and Nighthawk both strengthen the case that a “Q-Day”, a quantum computer capable of breaking today’s most common encryption, could arrive within a decade. Estimates cluster around 2033 to 2035, with Google itself warning it could come as early as 2029.
That timeline has already forced the U.S. government’s hand on a completely separate but closely connected front: a rushed migration to quantum-resistant encryption across federal systems.
The bottom line: Even the companies leading this race are careful not to overpromise. Microsoft talks about practical quantum computing arriving in years, not decades. IBM separates its 2026 advantage target from true fault tolerance, which it doesn’t expect to prove out at scale before 2029. The lab-bench breakthroughs are real. The everyday impact is still a few years out, encryption risk aside.
The ratio that actually explains the hype
The clearest way to understand why 2026 feels different: the ratio of physical qubits, the raw, noisy hardware, to logical qubits, the reliable, error-corrected unit computers actually need, has dropped fast. A few years ago, building one reliable logical qubit took roughly 1,000 physical qubits. Recent joint progress from IBM and Google has pushed that closer to 100 to 1, a change that dramatically improves the economics of scaling toward a genuinely useful machine.
Google, IBM, and Microsoft dominate the headlines, but they’re not alone. Quantinuum and IonQ are both betting on trapped-ion qubits, a different physical approach that some researchers argue carries inherently lower error rates, at the cost of slower operation speed. Nobody has settled which architecture ultimately wins, and the field currently resembles the early, chaotic years of classical computing, before a handful of dominant designs eventually crowded out the rest.
