Fault tolerance dominated the recent Quantum World Congress 2026, and so did a sharper debate over proof. The industry has largely stopped competing on raw qubit counts. Instead, vendors now compete on whose logical qubits, decoders, and benchmarks hold up under scrutiny, and buyers benefit from it.
The sold-out event brought together 3,000 attendees, 1,000 organizations, and delegations from over 30 countries in College Park, Maryland, just a few blocks from IonQ‘s headquarters and at the center of the state’s Capital of Quantum initiative.
Platinum sponsors included Atom Computing, Classiq, D-Wave, IBM, Infleqtion, IonQ, Microsoft, and Quantinuum.
Fault Tolerance Claims Meet a Skeptical Room
Every major hardware vendor now has a fault-tolerance roadmap. At QWC 2026, the conversation shifted to evidence, and vendors openly graded one another’s homework.
IonQ Runs Its Error-Correction Decoder on a Single CPU
IonQ announced a significant step toward practical fault-tolerant quantum computing, demonstrating the industry’s first end-to-end, real-time quantum error-correction decoder running on a single standard CPU (the Arm-based Apple M4 Max).
The core problem IonQ is addressing is that quantum error correction creates a substantial classical-computing workload. As a quantum computer runs, the control system must continuously interpret error-syndrome data and determine which corrections are required.
If that decoding cannot keep pace with the quantum hardware, the quantum computer must slow down or wait. IonQ says its decoder can perform this work continuously without creating a meaningful bottleneck.
The results showed:
- IonQ tested a dual-decoder architecture on benchmark circuits representing as many as 408 logical qubits.
- The simulated system included 88 memory blocks and magic-state factories, components required for large-scale fault-tolerant computation.
- The benchmarks covered more than 31.5 million quantum operations, which IonQ describes as “MegaQuOp” scale.
- Under the modeled noise conditions, the decoder added as little as 0.02% execution-time overhead, or “stretch” (the extra runtime that decoding introduces to a circuit).
IonQ’s strongest claim here concerns architectural scalability. The company is showing that the classical side of error correction does not need to become disproportionately expensive as logical-qubit counts increase.
If that holds up on increasingly larger physical systems, it removes one more potential bottleneck on the path to fault-tolerant quantum computing.
Infleqtion Entangles 30 Logical Qubits on Sqale
The neutral-atom vendor announced that it has encoded 30 logical qubits into 80 physical atoms, achieving an 8:3 overhead, and ran a benchmark circuit of roughly 1,000 physical operations that included four logical CCZ gates.
Two techniques stood out:
- An AI-discovered entangling gate reduced the number of two-qubit gates required per logical entanglement from eight to four.
- Software-based loss correction quadrupled the number of usable shots.
Infleqtion targets 100 logical qubits by 2028 and 1,000 by 2030.
Quantinuum Pushes Back
Quantinuum challenged many of the announced fault-tolerant claims in its post-event commentary. The company argued that decoding in simulation is distinct from real-time error correction on live hardware and that adding logical qubits does not, by itself, deliver fault tolerance.
The company cited its Helix architecture on Helios, which reports a logical error rate of 2.8×10⁻⁴ against a physical two-qubit error rate of 8×10⁻⁴, without post-selection, and 64 error-detected logical qubits.
Benchmarks Become the Battleground
Vendor-selected metrics make hardware comparisons nearly impossible for procurement teams. QWC 2026 brought standardized, system-level measurement to the forefront, with several efforts converging during the week.
QUOPS from Sandia, Quantinuum, and NVIDIA
Quantinuum is backing a new system-level quantum benchmark called QUOPS (Quantum Universal Operations Performance System), designed to make performance comparisons across very different quantum architectures more meaningful.
The core idea is straightforward: instead of judging a quantum computer primarily by qubit count, gate fidelity, or gate speed, QUOPS asks how much useful computation the complete system can successfully execute, and how quickly it can execute it.
This becomes increasingly important as the industry moves toward fault-tolerant systems, where error correction, decoding, compilation, routing, logical gates, and mitigation can significantly affect real-world performance.
QUOPS reports two primary metrics:
- Q captures the largest random circuit a system can run above a success threshold.
- Ω captures effective operations per second.
Together, they provide something analogous to “capacity plus throughput,” indicating how large a computation the machine can reliably handle and how quickly it can perform it.
The published QUOPS analysis compares systems from Quantinuum, Google , and IBM. It clearly illustrates the architectural trade-offs:
- Google Willow and IBM Boston: higher execution rates because superconducting gates are fast, but smaller demonstrated QUOPS-capability regions.
- Quantinuum Helios: slower execution rate because trapped-ion gates are slower, but it supports a larger computational capability region and a higher Q score.
- Error mitigation lets all systems trade execution speed for larger usable circuits, and QUOPS makes that cost visible through Ω.
The most important aspect of the announcement is that QUOPS doesn’t try to declare one architecture universally “better.” Rather, it exposes the size-versus-speed trade-off in a way buyers can compare.
Sandia National Laboratories initially developed QUOPS, with input from Quantinuum and NVIDIA.
Independent Cross-Stack Testing on IBM Heron
A benchmark released at the event applied the same demand for independent measurement to the software stack, comparing commercial error-management tools on IBM’s 156-qubit Heron processor:
- Q-CTRL and Qedma Quantum Computing QESEM both reduced errors compared with native Qiskit Runtime.
- Qedma achieved the highest precision at the cost of more QPU time.
- Q-CTRL balanced error reduction against runtime.
Separately, Q-CTRL reported a 100-qubit Quantum Fourier Transform on Heron R3.
DARPA Capital Quantum Benchmarking Hub
Independent evaluation needs an institutional home, and DARPA’s hub in College Park fills that role. The organization validates quantum systems for national security and commercial applications.
Microsoft will use its new Maryland research center to work directly with DARPA’s test and evaluation teams.
Commercialization: Integrators and Named Workloads
On the commercial side, QWC 2026 showed quantum reaching customers through delivery partners, specific workloads, and on-premises systems. Optimization and logistics accounted for most of the production-oriented news.
CGI Adds D-Wave to Its Services Portfolio
CGI will integrate D-Wave’s Advantage2 system and hybrid solvers into its technology services portfolio. The partnership focuses on production planning, energy grid optimization, and asset maintenance scheduling across logistics, transportation, and retail.
The deal builds on more than a year of joint work and gives D-Wave a global systems integrator channel.
CGI Federal Brings Quantum to Military Logistics
CGI Federal, the Defense Logistics Agency, and the University of Tennessee are partnering on the Quantum Pathfinder initiative, which applies AI and quantum methods to warehouse management, reverse logistics, and inventory positioning within military supply chains.
Classiq and IonQ Optimize a Gas Transmission Network
Classiq, Israel Natural Gas Lines, and IonQ published a hybrid QAOA workflow that maximizes throughput across a natural gas transmission network.
The partners validated it on IonQ’s Forte-1 system.
IonQ Lands On-Premises and Defense Work
Gate-model vendors continue to sell on-premises systems to universities and governments:
- Florida International University contracted for an IonQ Superion 256 trapped-ion system, the first 256-qubit machine in Florida, with delivery scheduled for late 2027.
- IonQ also published research demonstrating that quantum generative models outperform classical methods in satellite radar change detection at MCAS Miramar.
Maryland is the “Capital of Quantum”
Holding the event in College Park tied its demands for validation and commercialization to the infrastructure needed to support them. Maryland used the week to show off the research facilities, industry tenants, and federal evaluation capacity behind its quantum branding, and the wider region added training infrastructure of its own.
- Microsoft Quantum Research Center opens: Microsoft cut the ribbon on a 15,000-square-foot facility in the University of Maryland’s Discovery District, one year after announcing it at QWC 2025. The company also funded new quantum postdoctoral fellowships at UMD and opened a private preview of its Quantum Development Kit for analytics, including a Quantum Tensor PCA library.
- A denser Discovery District: Governor Wes Moore noted that IQM, Quantum Motion, and Riverlane have all added Maryland operations since Microsoft’s commitment. The state and UMD have now secured more than $500 million in pledges toward the $1 billion Capital of Quantum goal.
- A quantum computer in a public high school: Fairfax County Public Schools, Connected DMV, and FCEDA will install a room-temperature, four-qubit XeedQ XQ1e system at Skyview High School later this year, making it the first public high school to host a physical quantum computer. XeedQ previously won the QWC startup pitch competition.
Maryland now concentrates research talent, private capital, and federal evaluation capacity within a short drive. That proximity can shorten the path from validated results to commercial deployment, making it hard for other state quantum hubs to replicate quickly.
A survey released on opening day added a caveat to regional ambitions. Of 160 quantum companies across 15 countries, 90% rely on cross-border supply chains, and 74% serve international customers, complicating any push for protectionist trade policy.
What Comes Next
QWC 2026 raised the bar for what a quantum announcement must demonstrate. Over the next year, expect buyers and agencies to include neutral benchmarks such as QUOPS in requests for proposals, and expect vendors to face pressure to run decoders, logical qubits, and error correction together on live hardware.
The companies that publish verifiable, end-to-end results first will help set the terms for the first true fault-tolerant procurement cycle. QWC 2026 shows this is starting to happen, with an industry rapidly moving toward commercialization.
