Google Willow and the Race to Quantum Supremacy: 105 Qubits and Below-Threshold Error Correction

On December 9, 2024, Google’s Quantum AI division published a paper in Nature that sent shockwaves through the computing world. The paper, authored by a team led by Hartmut Neven and Julian Kelly at Google’s quantum computing laboratory in Santa Barbara, California, announced the Willow chip: a 105-qubit superconducting quantum processor that achieved what no quantum computer had ever done before. Willow demonstrated below-threshold quantum error correction, proving experimentally that increasing the size of a quantum error-correcting code actually reduced the logical error rate rather than amplifying it. This result, long predicted by theory but never achieved in practice, crossed a threshold that many physicists considered the single most important milestone on the road to fault-tolerant quantum computing. And it was accompanied by a benchmark that defied comprehension: Willow completed a random circuit sampling computation in under five minutes that would have taken the most powerful classical supercomputer on Earth an estimated 10 septillion years — a number that exceeds the age of the known universe by a factor of roughly 700 trillion.

To appreciate what Willow represents, it is necessary to understand where it came from. Google’s quantum computing program traces its origins to 2006, when Hartmut Neven, a German-born physicist and computer scientist, joined Google to lead its efforts in quantum machine learning. By 2013, Google had established a dedicated Quantum AI lab, initially in partnership with NASA and the Universities Space Research Association, housed at NASA’s Ames Research Center in Mountain View, California. The early work centered on a D-Wave quantum annealer — not a gate-based quantum computer but an optimization machine based on quantum tunneling. Google quickly recognized the limitations of quantum annealing for general computation and pivoted to building its own gate-based superconducting processors. In 2018, Google unveiled Bristlecone, a 72-qubit chip that demonstrated high-fidelity gates but fell short of the coherence needed for quantum supremacy. Then came Sycamore.

In October 2019, Google published a landmark paper in Nature reporting that its 53-qubit Sycamore processor had achieved quantum supremacy — performing a specific computation in 200 seconds that Google estimated would take IBM’s Summit supercomputer approximately 10,000 years. The claim was immediately contested. IBM published a blog post arguing that Summit could complete the same task in 2.5 days using optimized classical algorithms and enough disk storage, and that the term “quantum supremacy” overstated the result. Chinese researchers later demonstrated that tensor network methods on classical hardware could simulate Sycamore’s quantum circuits more efficiently than Google had assumed. But the fundamental point stood: Sycamore had performed a computation that was at least impractical, if not impossible, on classical hardware. It was a proof of concept, not a useful application, but it proved that quantum processors could operate in a computational regime qualitatively different from anything classical computers could reach.

Willow is the next generation of that program, and it represents a qualitative leap beyond Sycamore in every measurable dimension. The chip contains 105 transmon qubits — superconducting circuits based on Josephson junctions, cooled to approximately 15 millikelvin in dilution refrigerators. Google’s engineering achievement with Willow was not simply adding more qubits. The company fundamentally improved qubit coherence times, gate fidelities, and connectivity. Willow’s two-qubit gate error rates dropped below 0.5 percent, and the chip’s T1 coherence times — the time a qubit retains its quantum state before decaying — improved significantly over Sycamore. More importantly, Google redesigned the chip’s layout to support surface code error correction, the leading candidate for fault-tolerant quantum computing. Surface codes work by encoding a single logical qubit across a grid of many physical qubits, using repeated syndrome measurements to detect and correct errors without destroying the quantum information. The theoretical promise of surface codes is that below a certain physical error rate threshold, adding more physical qubits to the code makes the logical qubit more reliable, not less. Willow crossed that threshold.

The benchmark result — under five minutes versus 10 septillion classical years — warrants careful analysis. Random circuit sampling, the benchmark Google used, is not a commercially useful computation. It is a mathematical task specifically designed to be hard for classical computers but natural for quantum processors. Critics have rightly pointed out that beating classical computers at a task designed to favor quantum computers does not prove that quantum computers will be useful for real-world problems like drug discovery, materials simulation, or cryptanalysis. Google acknowledges this. The company’s published roadmap describes a phased progression from noisy intermediate-scale quantum (NISQ) demonstrations to what it calls “useful quantum computing” — quantum computations that solve real problems faster or better than any classical alternative. Google has targeted 2029 for this milestone, with an intermediate goal of building a logical qubit with an error rate below one in a million by 2027. Willow’s below-threshold error correction is the critical stepping stone: it proves that the physics works, that the scaling laws are favorable, and that the engineering is within reach.

Google’s Quantum AI laboratory in Santa Barbara is the nerve center of this effort. The lab occupies a purpose-built facility adjacent to the University of California, Santa Barbara, and houses some of the most advanced cryogenic and nanofabrication infrastructure in the world. The research team includes physicists, engineers, and computer scientists drawn from leading universities and national laboratories. Key figures include Julian Kelly, who leads the Willow hardware effort; John Martinis, who built the original Sycamore chip before leaving Google in 2020; and Sergio Boixo, who leads quantum theory and algorithms. The Santa Barbara lab fabricates its own quantum chips in-house, giving Google end-to-end control over the design-build-test cycle. This vertical integration — from qubit design through chip fabrication to cloud deployment — mirrors Microsoft’s approach with Azure Quantum and positions Google as one of only a handful of companies in the world with the capability to build quantum processors from scratch.

The cloud dimension of Google’s quantum program is inseparable from its government contracting ambitions. Google Cloud is one of the four primary vendors under the Joint Warfighting Cloud Capability (JWCC) contract, the $9 billion multi-cloud arrangement that replaced the canceled JEDI contract in December 2022. JWCC gives the Department of Defense access to cloud services from Microsoft Azure, Amazon Web Services, Google Cloud, and Oracle Cloud, spanning all classification levels from Impact Level 2 through Impact Level 6 and above. Google Cloud holds FedRAMP High authorization and has been expanding its government cloud capabilities through its Public Sector division. While Google has historically been more cautious about military AI work — famously withdrawing from Project Maven in 2018 after employee protests over drone targeting algorithms — the company’s participation in JWCC signals a pragmatic recalibration. Government cloud contracts represent hundreds of billions of dollars over the coming decade, and no major cloud provider can afford to cede that market entirely.

Google’s quantum roadmap intersects with its government cloud ambitions in ways that have received remarkably little public scrutiny. As quantum processors reach practical utility, they will be integrated into cloud platforms — the same cloud platforms that serve classified military networks. Google already offers quantum computing access through its Google Cloud console, allowing researchers and developers to run algorithms on its quantum hardware remotely. When Willow’s successors achieve fault-tolerant operation, that same cloud access model could extend to government and military users under JWCC. The potential applications in classified environments — optimization of logistics networks, simulation of novel materials for defense applications, and eventually cryptanalysis of adversary communications — are exactly the capabilities the Pentagon has been investing billions to develop.

The geographic footprint of Google’s infrastructure buildout adds another layer to this analysis. In 2024, Google announced a massive data center expansion in Pryor, Oklahoma, investing over $2 billion to expand an existing facility that has operated since 2011. The Pryor campus is located in the MidAmerica Industrial Park, one of the largest industrial parks in the United States, with direct access to low-cost hydroelectric power from the Grand River Dam Authority. Google’s expansion in Pryor is part of a broader pattern of hyperscale data center construction across the American heartland, a pattern that aligns with the infrastructure corridor being built for the Stargate Project — the $500 billion AI initiative announced in January 2025 by OpenAI, SoftBank, Oracle, and MGX, with its primary campus in Abilene, Texas. The clustering of massive computing infrastructure in a corridor stretching from Texas through Oklahoma is not coincidental. These facilities require reliable power, fiber connectivity, and physical security — the same requirements that define the locations of classified government computing installations. The proximity of commercial hyperscale data centers to existing government and military infrastructure nodes creates opportunities for hybrid classified-commercial computing architectures that few outside the defense contracting world fully appreciate.

Google’s Willow chip represents genuine scientific progress. The below-threshold error correction result is real, peer-reviewed, and significant. But scientific progress does not occur in a vacuum. It occurs within institutional contexts — corporate strategies, government contracts, infrastructure deployments — that determine how breakthroughs are used and who benefits from them. Google is simultaneously pursuing quantum supremacy in its Santa Barbara lab and expanding its government cloud presence through JWCC. It is building hyperscale data centers along the same infrastructure corridors that host classified government installations. And it is developing quantum processors that, within the next five years by its own estimate, will be capable of computations with direct national security applications. The investigation documented on this site traces these convergences — between commercial technology and military infrastructure, between quantum computing and classified cloud networks, between the companies that build the processors and the contracts that deploy them into the deepest layers of the defense establishment. Google Willow is one node in a much larger network. Understanding that network requires following not just the science but the money, the contracts, and the infrastructure.