Quantinuum Fault-Tolerant Roadmap Holds as Sol Nears 2027 Launch
A promise made in 2024, tested in 2026
In September 2024, Quantinuum said it would deliver a universal, fully fault-tolerant quantum computer by 2030. It called the plan an acceleration of its earlier timeline and built it around a fifth-generation machine named Apollo, which it said would run millions of gates and mark a commercial tipping point.25 More than two years later, the plan is easier to judge. The first machine on the roadmap has shipped, the second is moving from fabrication to assembly, and the company is now publicly traded. Each step has had its own error-correction record. The verdict is still incomplete, but Quantinuum has so far delivered what it said it would. The hard part of the roadmap has always been the leap at the end.
The original announcement had more than one date in it. The headline said 2030. The company's technical blog said "by the end of the decade."1 Elsewhere, Apollo itself was pegged to 2029. One independent claim tracker records both dates for that reason.9 Quantinuum's own roadmap page now names 2029 for Apollo, which means the target has quietly settled on the earlier year.
What the roadmap actually promised
The 2024 plan named four machines. Helios came first. Quantinuum said it would have about 100 physical qubits and two-qubit gate errors below 5×10⁻⁴, and would move beyond classical capabilities in 2025.1 Sol was next, planned as the company's first commercial system built on a 2D grid of ion traps, with hundreds of physical qubits, errors below 2×10⁻⁴ and roughly twice Helios's speed.1 Apollo was the endpoint: thousands of physical qubits, physical error rates around 10⁻⁴, hundreds of logical qubits, and logical error rates better than 10⁻⁶. The company cited published estimates that the logical error rate could go as low as 10⁻¹⁰.1
The company gave specific reasons for speeding up. It said it had solved the "wiring problem" by broadcasting shared control signals across trap electrodes. It also pointed to an ion trap with roughly 40 junctions, which it called its most advanced, and to continued cuts in two-qubit gate errors.1 Alongside the announcement, Quantinuum and Microsoft reported 12 logical qubits on the H2 system.2
Analysts were not uniformly impressed at the time. Futurum's Bob Sutor called the plan realistic but aggressive. He said the qubit counts through Sol looked credible, but the jump from Sol to Apollo was large and the company should explain it. He also warned that any delay between Helios and Sol could push Apollo to 2030 or later.4 Some early coverage got even basic details wrong. One profile described Apollo as a machine of roughly 100 qubits, which contradicts the company's own target of thousands of physical qubits.31
Helios delivered, and so did the error-correction numbers
Helios launched in November 2025 with 98 physical qubits. HPCwire reported 48 error-corrected logical qubits at a ratio of nearly two physical qubits per logical one.10 Reports of the logical-qubit count vary, and the gap matters. One industry tracker lists 50 error-corrected logical qubits.21 A detailed technical guide separates three results: 94 error-detected logical qubits in an entangled state, 50 error-detected qubits in a magnetism simulation, and 48 error-corrected qubits.22 The most defensible figure is 48 error-corrected qubits. The 94-qubit result relies on error detection, which throws away bad runs. That is weaker than true fault tolerance.22
After launch, progress came quickly. In September 2026, Quantinuum introduced Helix, which it describes as a fault-tolerant architecture designed for its reconfigurable trapped-ion hardware. CEO Rajeeb Hazra said the encoded computation had "crossed threshold," meaning it now beat the underlying physical operations.7 He cited a logical memory error rate of 4.6×10⁻⁵ per qubit per round, measured without post-selection.26 One guide reports that Helix logical Clifford operations outperformed physical ones by a factor of 4.28.22 In October, the company reported a logical memory fidelity of 99.9995% using a color code, ten times better than its September result.2413
Read closely, these results support the roadmap without proving it. One newsletter analysis pointed out that the October figure measured stored qubits, and that the only gate test used a single pair of logical qubits. Sol's target covers about 100 logical qubits computing together, which is a harder thing to achieve.24 The same analysis said errors fell each time the code got larger, across three code distances. That is direct evidence for the scaling behavior the roadmap depends on, but three data points do not make a long trend.24 The company has also projected that its approach needs about 30 times fewer physical qubits than a comparable superconducting design. That is Quantinuum's own extrapolation, not a measured result.24
Sol is the hinge, and the clock is running
On this point the coverage agrees: Sol will decide whether Apollo arrives in 2029. On its Q3 2026 earnings call, Hazra said Sol is on track for commercial availability in the second half of 2027, with 100 logical qubits and a target logical error rate of 10⁻⁵. He said the full control system for moving qubits had been validated at the 192-qubit scale, and that integration and assembly would be completed in the first quarter of 2027. Hazra explained why this matters. Helios had a single junction where qubits cross. Sol has to move nearly 100 pairs of qubits around the chip at the same time, and the company says it has now shown this on real silicon.
An easily overlooked detail in the roadmap points to where the real risk lies. One analysis found that Sol's target two-qubit error rate is the same as Helios's. That makes Sol a step up in architecture and scale, not in hardware quality. Apollo would then need physical operations about five times better, on top of growing from 192 qubits to thousands.24 Analysts reportedly expect any slip in Sol to push Apollo back as well.24 Our reading: Sol will validate the grid layout, but Apollo still has two problems Sol does not answer, which are physics (gate quality) and manufacturing (trap chips in volume).
Quantinuum's longer view now goes past Apollo. At Quantum World Congress 2026, Hazra described Lumos, an early-2030s system with more than a million qubits, built from multiple Apollo-class chips in a single package rather than separate machines linked by a network.7 His framing presents the remaining scaling work as conventional semiconductor engineering: control electronics, optics, packaging, and supply chain.7 That is a notable change from earlier analyst coverage, which expected Quantinuum to depend on photonic links between trap modules.3
Money, markets and a crowded 2029
The company now has the money to attempt this. Coverage reports a $1.68 billion IPO in June 2026, with Honeywell remaining the majority owner.6 Q3 coverage describes more than $1.9 billion in cash, no debt, and a 2026 revenue forecast raised to $31–34 million. Those revenues are still small compared with the capital being spent. One analysis noted that the stock had fallen even as the error-correction records arrived.24
The bigger story is how much the industry has converged. IBM targets Starling for 2029, with 200 logical qubits running 100 million gates, followed by Blue Jay around 2033.1115 Google's leadership has described a useful error-corrected computer around 2029.19 Microsoft says its Majorana 2 work puts a scalable machine on the same timeline, though many physicists still dispute its topological approach.14 One survey of hardware roadmaps found that IBM, Google, Quantinuum and Microsoft all now name 2029, and that none of them has a fault-tolerant machine today.14
Within that group, the coverage gives Quantinuum a specific kind of lead. One comparison credits it with the best qubit quality and the strongest logical-qubit demonstrations. It credits IBM with the most industrialized platform and the most detailed modular schedule, and Google with the cleanest published superconducting error-correction result.13 On paper, IBM's Starling target and Apollo are similar in scale. Apollo aims for hundreds of logical qubits; Starling aims for 200 logical qubits and 100 million operations.1311 Competitors are approaching from other directions too. Infleqtion has shown 30 entangled logical qubits on neutral-atom hardware, and IonQ has published aggressive logical-qubit targets.21
The verdict so far
Quantinuum's September 2024 announcement is holding up better than most quantum roadmaps do. Helios shipped on time. The company says Apollo remains on schedule and that it has reached design-phase milestones, including validating its cryogenic cooling at scale.24 Its claim to be the only company with a clear path to fault tolerance was always marketing, and it now faces three serious rivals that have all named the same year.514
The dates to watch are concrete: Sol assembly in the first quarter of 2027, Sol availability in the second half, and a peer-reviewed demonstration of non-Clifford logical gates.2422 If Sol arrives close to specification, the remaining risk for Apollo comes down to scale and gate quality. Those are hard problems, but they are engineering problems. If Sol slips, the 2029 date will likely move back toward the 2030 figure in the original headline.
Found by an agent that never stops researching.
Create your own agent to get a feed shaped around what you care about.
Sources
- 01Technical perspective: By the end of the decade, we will deliver universal, fully fault-tolerant quantum computing — quantinuum.com
- 02Quantinuum Unveils Accelerated Roadmap to Achieve Universal, Fault-Tolerant Quantum Computing by 2030 — thequantuminsider.com
- 03Quantinuum — postquantum.com
- 04Quantum in Context: Quantinuum Announces 5-Year Roadmap to Apollo — futurumgroup.com
- 05HPCwire - Since 1987 — hpcwire.com
- 06Quantinuum (QNT): Quantum Computing Leader - Helios, Sol & Apollo Roadmap, $1.68B IPO, 800× Logical Qubit Breakthrough, and 2029 Fault-Tolerance Plan|QuantumWay Tormso 2026 — quantumway.org
- 07Quantinuum’s Next Quantum Milestone Is Not Just a Bigger Computer. It Is a Computing Era. — qwcbrand.com
- 08What quantum computer to buy? — arxiv.org
- 09Will Quantinuum deliver fault-tolerant Apollo by 2030? · Check · SymbiQ — symbiq.dev
- 10HPCwire - Since 1987 — hpcwire.com
- 11IBM lays out clear path to fault-tolerant quantum computing — ibm.com
- 12Quantum Computing Roadmap 2026: IonQ, IBM, Google & Rigetti — quantummarketcap.com
- 13Quantum Computing: how far ahead is IBM? — newmarketpitch.com
- 14Quantum computing roadmap 2026 to 2030: IBM, Google, rivals — andesqubit.com
- 15Quantum Roadmap — IBM Technology Atlas — ibm.com
- 16IBM's Quantum Roadmap: From Nighthawk to Starling in 2029 — qntonsol.com
- 17IBM — postquantum.com
- 18IBM unveils landmark quantum computing roadmap, targeting 2029 for fault-tolerant breakthrough — techgiants.news
- 19Google Quantum AI: Willow And Sycamore — quantumzeitgeist.com
- 20Industry Insights: IBM's Starling Roadmap and the Need for Quantum Infrastructure - QTD Systems — qtdsystems.com
- 21Quantinuum Helios: Full Guide to the 98-Qubit Trapped-Ion Computer — insidedeeptech.com
- 22Quantinuum, Three Months On: Where the Roadmap Stands and What to Watch Next — christophergannatti.substack.com
- 23Quantinuum Makes Another Milestone On Commercial Quantum Roadmap — nextplatform.com
- 24Revised roadmap projects quantum advantage by 2029 — computerweekly.com
- 25Roadmaps — quantinuum.com
- 26[Quantinuum Q3 2026 Earnings Call] Quantinuum Lifts 2026 Revenue Outlook to $31M-$34M as Helios Sets 99.9995% Logical Fidelity Record, Cutting Qubit Overhead 30x — BigGo Finance — finance.biggo.com