Quantum computing has spent two decades as the technology that is always ten years away. In December 2024, Google did something that made that framing harder to sustain: it unveiled the Willow chip, a 105-qubit processor that cracked a problem the field had been stuck on for nearly 30 years, and benchmarked it against a task that would take the world's fastest classical supercomputer longer than the age of the universe to complete.
That is not a metaphor. The number is 10 septillion years — a 1 followed by 25 zeroes. Willow solved an equivalent problem in under five minutes.
Since then, Google's Quantum AI program has moved fast. In March 2026, it added a second hardware modality — neutral atom computing — to run alongside its superconducting chip programme. The goal, unchanged since the programme's founding, is a large-scale error-corrected quantum computer. The timeline — by 2029 — is now treated as a serious engineering target rather than a research aspiration.
Here is what Willow actually achieved, what the six-milestone roadmap looks like from the inside, and what Google is building toward.
What Willow Actually Did
Before Willow, the fundamental problem with quantum computers was this: the more qubits you added, the more errors you introduced. Errors in quantum systems are not like errors in classical computing — they do not just corrupt data, they destroy the quantum states that give quantum computers their power in the first place. Adding more qubits to fix errors introduced more errors than they fixed. The field had known about this ceiling for almost 30 years and had been trying to break through it with a concept called quantum error correction.
Willow broke through it.
Using a technique called surface code error correction, Willow demonstrated that as it scaled up in qubit count, its error rate went down, not up. This is called operating below the threshold — the point at which adding more physical qubits to form a single logical qubit actually makes that logical qubit more reliable rather than less. It is the foundational prerequisite for building a quantum computer that can run long enough and accurately enough to solve real-world problems.
This is what Google calls Milestone 2 on its roadmap. It took five years to get there from Milestone 1.
Understanding the underlying physics behind why this is difficult requires grasping what quantum entanglement actually is — the fragile correlations between qubits that quantum computers exploit, and that environmental noise destroys. Error correction is the engineering discipline of protecting those correlations long enough to be useful.
The Six-Milestone Roadmap
Google Quantum AI organises its entire programme around a public six-milestone roadmap. Each milestone represents a concrete engineering proof point on the path to the final goal: a fault-tolerant quantum computer capable of running algorithms that matter.
| Milestone | Description | Status |
|---|---|---|
| 1 | Beyond-classical computation | ✅ 2019 — Sycamore (53 qubits) |
| 2 | Below-threshold error correction | ✅ Dec 2024 — Willow (105 qubits) |
| 3 | Long-lived logical qubit | 🔄 In progress |
| 4 | Logical gate between two logical qubits | ⏳ Pending |
| 5 | Scale-up to many logical qubits | ⏳ Pending |
| 6 | ~1 million physical qubits; large-scale error-corrected machine | 🎯 Target: ~2029 |
A few things stand out in this table. First, Milestone 1 (Sycamore, 2019) was a demonstration of quantum computers performing a task that classical computers could not practically replicate — but it was a contrived task with no real-world application. Its value was proof of concept, not utility.
Second, Milestone 2 (Willow, 2024) was the first time Google demonstrated the physical foundation for useful quantum computing: below-threshold error correction. Without this, none of the remaining milestones are achievable.
Third, the gap between Milestone 2 and Milestone 6 is enormous. A million physical qubits, bundled into a much smaller number of stable logical qubits, capable of running algorithms for drug discovery, materials science, or cryptography. The distance between 105 noisy physical qubits and one million error-corrected ones is where most of the hard engineering work actually lives.
The target date of 2029 was stated explicitly by Google in late 2024. It is an ambitious claim. Whether it holds will depend on progress at Milestones 3 and 4.
2026: Google Opens a Second Hardware Lane
In March 2026, Google Quantum AI announced something unexpected: it was adding neutral atom computing as a second hardware modality to run alongside its superconducting qubit programme.
The two technologies approach the same goal from different angles. Superconducting qubits — the technology Willow is built on — excel at circuit depth: they can execute more gate operations per unit of time, which matters for certain classes of quantum algorithms. Neutral atom systems, in which individual atoms are held in place by laser tweezers and used as qubits, excel at scaling qubit count: atom arrays can currently pack many more qubits into a usable configuration than superconducting chips can.
Google describes this as a space-time trade-off — superconducting qubits win on the time (depth) dimension, neutral atoms win on the space (count) dimension. Rather than betting exclusively on one approach, the 2026 announcement positions Google to pursue whichever modality produces the next major result.
The neutral atom programme will focus on three areas: adapting error correction protocols to work with atom array connectivity patterns, using Google's compute infrastructure to simulate and optimise hardware designs before building them, and developing application-scale fault-tolerant performance benchmarks. It is not a consumer product announcement. It is an R&D hedge of significant scale.
The Research Access Expansion
Alongside the hardware work, Google Quantum AI has opened access to Willow to a select cohort of external researchers through its Willow Early Access Program. Organisations can submit research proposals to work directly on the Willow processor, with priority given to high-impact scientific and computational problems.
In January 2026, Google announced a partnership with the UK National Quantum Computing Centre (NQCC), widening access for UK-based researchers. This pattern — building an ecosystem of external collaborators rather than running purely closed internal research — mirrors how cloud computing developed, with early access partners generating results that validate the technology's direction.
For the broader implications of what quantum computers could eventually break — and why post-quantum cryptography is already being deployed — the encryption and quantum computing connection is worth understanding now, not when the hardware arrives.
What Quantum Computers Are Actually For
A common misconception is that quantum computers will replace classical computers. They will not. Even a fully error-corrected quantum computer running Milestone 6 hardware will not be faster at browsing the internet, running spreadsheets, or streaming video. Classical computers are already optimal for those tasks.
What quantum computers are specifically suited to are problems involving enormous combinatorial search spaces, quantum simulation, and certain mathematical operations that are intractable for classical hardware. The canonical applications include:
- Drug discovery and materials science: Simulating molecular behaviour at the quantum level — something classical computers can only approximate — to identify promising drug candidates or new materials
- Cryptography: Running Shor's algorithm to factor large integers, which underpins most current public-key cryptography (this is the threat that post-quantum cryptography is designed to pre-empt)
- Financial optimisation: Portfolio optimisation and risk modelling problems that scale exponentially with classical approaches
- Climate and logistics modelling: Certain classes of optimisation problems that affect routing, resource allocation, and atmospheric simulation
None of these are available today. They require the error correction and scale that the later milestones on Google's roadmap aim to deliver. The honest timeline from the current state of hardware to commercially useful quantum advantage on real problems is still measured in years, not months.
For a plain-English grounding in how qubits and superposition actually work before diving into the roadmap details, the quantum computing explainer covers the fundamentals without the jargon.
Addressing the Shutdown Myth
In early 2026, a wave of online posts claimed that Google had "shut down" or "cancelled" Willow. This was misinformation.
Willow is a research chip, not a consumer product with a launch event and a cancellation notice. Research hardware programmes operate quietly between technical milestones — there are no press conferences when a team is running error correction experiments or debugging qubit coherence times. When public communication goes quiet, it means the team is doing the work, not that the project is dead.
Google's Quantum AI programme is active, funded, and running both Willow access programmes and the 2026 neutral atom initiative in parallel. The roadmap is public. The research publications continue. The "shutdown" narrative was a misreading of research-programme communication norms by audiences accustomed to product cycles.
What Comes After Willow
The next milestone — Milestone 3, a long-lived logical qubit — is the current engineering target. A logical qubit is not a single physical qubit; it is a collection of physical qubits whose correlations are managed by error correction protocols to create a stable, persistent quantum state. Demonstrating a logical qubit that stays coherent long enough to be useful is the proof that error correction works at a level above the threshold demonstration Willow provided.
After that, Milestone 4 requires demonstrating a gate between two logical qubits — the basic operation of quantum computation at the logical level rather than the physical level. This is where the architecture for a real quantum algorithm starts to become visible.
Milestones 5 and 6 involve scaling. The final machine targets approximately one million physical qubits bundled into a much smaller number of highly reliable logical qubits. Google's 2029 target covers completing this arc.
The scale of ambition is real, and so is the uncertainty. The path from 105 physical qubits to one million is one of the most demanding engineering challenges in the history of computing. Google is further along it than any organisation has been before. Whether that translates into a commercially useful quantum computer before the end of the decade is what the next three years will determine.
Frequently Asked Questions
What is Google's Willow chip?
Willow is Google's 105-qubit quantum processor, announced in December 2024. It was the first chip to demonstrate below-threshold quantum error correction — meaning its error rate decreases as more qubits are added, solving a 30-year challenge in the field.
How fast is the Willow quantum computer?
Willow completed a specific benchmark computation in under five minutes. An equivalent computation on the world's fastest classical supercomputer would take approximately 10 septillion years — a number larger than the age of the universe.
What is Google's quantum computing roadmap?
Google's Quantum AI programme follows a six-milestone roadmap. Milestones 1 (2019, Sycamore) and 2 (2024, Willow) are complete. The remaining milestones cover long-lived logical qubits, logical gates, scale-up, and ultimately a large-scale error-corrected quantum computer targeting approximately 2029.
Did Google shut down Willow in 2026?
No. Willow is an active research platform. The "shutdown" claim was misinformation. Google launched the Willow Early Access Program for external researchers in 2025-26 and announced a UK National Quantum Computing Centre partnership in January 2026 — both signs of an active, expanding programme.
What is neutral atom quantum computing?
In neutral atom quantum systems, individual atoms held in place by laser tweezers act as qubits. In March 2026, Google added neutral atom hardware as a second research modality alongside its superconducting chip programme. Neutral atoms are better suited to scaling qubit count; superconducting qubits excel at circuit depth. Google is running both in parallel.
Sources
- Google Willow quantum chip announcement — Google Blog
- Google Quantum AI Adopts Dual-Modality Strategy with Neutral Atom Expansion — Quantum Computing Report
- Google Expands Quantum Efforts to Include Neutral Atom Systems — HPCwire
- Google Quantum AI: The Complete 2026 Guide — Quantum Zeitgeist
- Google Quantum AI Roadmap — quantumai.google
