Quantum Computing in 2026/2027: What Is Real and What Is Hype?

Quantum computing in 2026 is real science but not yet a useful business tool. Error correction now works in the lab, and a few experiments appear to beat the best classical computers on narrow physics problems. Machines that solve valuable problems are targets for 2029 to 2033, while the risk to today's encryption is already setting deadlines.

This guide separates what's real from what's hype: where the technology stands, which advantage claims hold up, what quantum computers will be used for first, the roadmaps to 2033, what to expect in 2027, when encryption is at risk, and what it all means for the UAE and for property and data-centre owners.

Key takeaways

• Error correction works, at small scale. Google's Willow chip showed errors falling as its code grew, and Quantinuum's Helios runs 48 error-corrected logical qubits that outperform its physical ones.
• Advantage claims are narrow and contested. The strongest are on physics problems. Three earlier candidates on a community tracker were overturned by classical methods, one in 12 minutes on a laptop.
• Useful machines are targets, not products. IBM aims for its Starling system in 2029 and Blue Jay in 2033; DARPA now thinks someone will build a utility-scale quantum computer by 2033.
• Chemistry and materials come first. Researchers argue useful advantage needs small-data problems with big speedups, which rules out most optimisation, finance and AI claims for now.
• Security can't wait. Estimates for breaking RSA and elliptic-curve cryptography keep falling. The US requires post-quantum key exchange on high-value federal systems by the end of 2030, and Google set 2029 for its own migration.

What is quantum computing, and where does it stand in 2026?

A quantum computer processes information with qubits, which use superposition and entanglement to tackle certain calculations in ways ordinary computers can't. In 2026 the field is moving from noisy experiments to early error correction. According to Google, machines with error rates low enough to matter have on the order of only 100 to 1,000 qubits.

Definition

Physical vs logical qubit — a physical qubit is a single hardware qubit, which makes errors. A logical qubit is an error-corrected qubit built from many physical qubits — "composed of hundreds of physical qubits", in Google's description — that together store one qubit of information reliably. Fault-tolerant machines that do useful work need hundreds to thousands of logical qubits.

Several hardware approaches are racing, and none has won:

Approach	Leading players	Where it stands in 2026
Superconducting circuits	Google, IBM, AWS	Error correction below threshold; IBM's Nighthawk runs circuits of 7,500+ gates
Trapped ions	Quantinuum, IonQ	Helios: 98 qubits at 99.921% two-qubit fidelity; IonQ reports 99.99%
Neutral atoms	QuEra, Atom Computing, Google (from 2026)	Arrays of 6,100 atoms; 3,000 atoms held continuously for over two hours
Photonics	PsiQuantum, Xanadu	Building sites in Chicago and near Brisbane; no public utility-scale system
Silicon spin qubits	Diraq, Quantum Motion, Silicon Quantum Computing	In DARPA's benchmarking programme
Topological qubits	Microsoft	Early claims disputed
Quantum annealing	D-Wave	Special-purpose machines; supremacy claim contested

What is real in quantum computing today?

Four results stand up to scrutiny: error correction that improves with scale, logical qubits that beat physical ones, atom arrays in the thousands, and at least one peer-reviewed, verifiable beyond-classical physics experiment.

Error correction below threshold. In December 2024 Google published in Nature that its Willow chip suppressed logical errors by a factor of 2.14 each time the error-correcting code grew by one step, reaching a 101-qubit code with 0.143% error per cycle. The error-corrected memory lasted 2.4 times longer than the chip's best physical qubit. That's the milestone the field had chased for decades: adding qubits made the system more reliable, not less.

Logical qubits that beat physical ones. Quantinuum's Helios, launched in November 2025, has 98 trapped-ion qubits with 99.921% two-qubit fidelity. The company reports 48 error-corrected logical qubits performing better than physical ones, and 94 error-detected logical qubits entangled across the machine.

Scale in neutral atoms. In September 2025 a Caltech team held 6,100 atoms as qubits in superposition for about 13 seconds, and a Harvard-led team kept an array of more than 3,000 atoms running for over two hours by reloading 300,000 atoms a second.

A verifiable physics advantage. In October 2025 Google reported in Nature a measurement on Willow that took about two hours and would take a classical supercomputer an estimated 13,000 times longer, a conclusion reached after roughly 10 person-years of classical "red teaming" trying to beat it.

Industrial money. Quantinuum listed on Nasdaq in June 2026, selling 28 million shares at $60 each, $1.68 billion gross. Its prospectus shows the scale of the business behind the headlines: $30.9 million of revenue and a $192.6 million net loss in 2025.

What is hype in quantum computing?

The hype sits in headline numbers without context, claims that haven't survived review, and business uses that aren't proven yet.

Claim	What's actually true
"Willow did in five minutes what a supercomputer would need 10 septillion years for"	True for random circuit sampling, a benchmark with no practical use. Google's own molecule experiment in 2025 was "not yet beyond classical"
"D-Wave achieved quantum supremacy on a useful problem"	Contested. Tensor-network researchers published a classical challenge in Science in 2026, and an EPFL team matched the annealer's accuracy with classical simulation
"Microsoft has a topological quantum computer"	Nature's editors stated the published results do not represent evidence of the Majorana zero modes the approach depends on
"A 6,100-qubit quantum computer"	The Caltech atoms weren't yet entangled; entanglement is the next step before full quantum computations
"The UAE has a 500,000-qubit quantum computer"	TII's 500,000-qubit result was a quantum annealing simulation on NVIDIA GPUs, not quantum hardware
"Quantum already beats banks' trading models"	HSBC and IBM reported up to 34% better prediction of bond-trade fills than common classical techniques used in the industry, a promising trial rather than a proven advantage
"More qubits means more power"	Error rates and logical qubits matter more than raw counts

Common misconception

"Quantum computers are already faster than supercomputers." Only for specific, carefully chosen tasks. On random circuit sampling and a few physics measurements, quantum machines appear to win; on almost everything businesses run, classical computers are faster, cheaper and more reliable, and will be for years.

Has quantum advantage been achieved?

Not in any way that matters commercially yet. A handful of experiments on narrow physics problems appear to beat the best classical methods, and IBM says it delivered advantage in 2026, but those claims are still being checked, and several earlier ones were overturned.

Quantum advantage means doing an information-processing task more efficiently, more cheaply or more accurately than classical methods alone. The problem is the moving target: classical algorithms keep improving. That's why IBM, Algorithmiq, the Flatiron Institute, BlueQubit and others run a public Quantum Advantage Tracker, where candidates stay "active" until classical methods catch up and become "superseded" when they do.

What the data shows

Quantum Advantage Tracker, September 2026: 6 active candidates (4 in observable estimation, 2 in classically verifiable problems) and 3 superseded ones. One superseded candidate — a 56-qubit, 1,917-gate "peaked circuit" — was solved classically by a Moderna researcher in May 2026 in 12 minutes on a consumer Apple laptop.

IBM's claim rests on three preprints posted in late July 2026 with partners including the University of Chicago, Qedma, RIKEN and BlueQubit, framed as advantage through "trusted" computation with built-in validation. IBM's roadmap page now says it and its partners delivered advantage as promised in 2026. The papers hadn't completed peer review when this was written, and they address physics problems, not business workloads.

What will quantum computers be used for first?

Simulating molecules and materials, where nature itself is quantum. Researchers at ETH Zurich and Microsoft argue that useful advantage needs problems with small data inputs and algorithms with more-than-quadratic speedups — conditions most proposed business uses don't meet — while seeing "a huge potential" in materials science and chemistry.

Use	Outlook	Why
Chemistry and materials: drugs, batteries, catalysts	Most promising	Nature is quantum, inputs are small and speedups can be large
Physics simulation	Producing results now	Where the 2025–2026 advantage claims sit
Breaking today's public-key cryptography	Real, but needs large fault-tolerant machines	Known algorithms give big speedups; resource estimates are falling
Optimisation and logistics	Unproven	Most proposed algorithms don't achieve the speedups needed to beat classical methods
Finance	Early trials	HSBC's bond-trading trial is the best-known result
AI and machine learning	Unlikely soon	Big-data workloads are a poor fit for quantum hardware

Quantum computers will work alongside classical ones rather than replace them. NVIDIA's NVQLink interconnect, announced in October 2025 with 17 quantum hardware builders and nine scientific labs, is built to couple quantum processors to GPU supercomputers, with each doing what it's good at. The GPU side of that picture is covered in the AI infrastructure guide.

What are the limits of quantum computing?

The limits are errors, scale and applications. Qubits are fragile, error correction multiplies the hardware needed, the largest arrays aren't yet running deep computations, and few problems are known to benefit.

• Errors. Even Helios, among the most accurate machines on sale, averages about eight errors per 10,000 two-qubit operations, according to its technical paper.
• Overhead. Each logical qubit consumes hundreds of physical ones, so a machine with a thousand logical qubits needs hundreds of thousands of physical qubits working together.
• Depth, not just size. Google says neutral atoms still have to demonstrate deep circuits with many cycles, while superconducting machines still have to scale to tens of thousands of qubits.
• Fragile advantage. Claims can be overturned by better classical algorithms, as the tracker shows.
• Few applications. Most proposed uses don't have the speedups needed to pay back the overhead.
• Economics. Revenue is small next to the capital raised; Quantinuum lost $192.6 million on $30.9 million of revenue in 2025.

What is the roadmap to useful quantum computers?

Companies target fault-tolerant machines between 2029 and 2030 and larger, utility-scale systems by 2033. These are company targets, not delivered products, and several earlier roadmaps have slipped.

From noisy qubits to useful machines

1. 01Noisy
   • Tens to hundreds of qubits
   • No error correction
   • Benchmarks only

   Where the field started

2. 02Early error correction
   • Below-threshold codes
   • Tens of logical qubits
   • Physics advantage claims

   2024–2026 — where it is now

3. 03First logical-qubit machines
   • Magne: 50 logical qubits
   • QuEra Libra on AWS
   • IBM: 10,000-gate circuits

   2027–2028 targets

4. 04Fault tolerant
   • IBM Starling: 200 logical
   • IonQ: 80,000 logical
   • Google: commercially relevant

   2029–2030 targets

5. 05Utility scale
   • IBM Blue Jay: 2,000 logical
   • DARPA's 2033 test

   Value exceeds cost

Company or programme	Target	When	Status
IBM Nighthawk	Circuits of 7,500 two-qubit gates, then 10,000 and 15,000	2026, 2027, 2028	Circuits of 7,500+ gates reported in August 2026 (company)
Atom Computing and Microsoft: Magne, Copenhagen	50 logical qubits, more than 1,200 physical	Around the turn of 2026/27	Target
QuEra	Libra, its first fault-tolerant machine, on Amazon Braket; a larger "gigaquop" system next	2028; 2028–2029	Target
IBM Starling, Poughkeepsie	200 logical qubits running 100 million operations	2029	Target
Google	Commercially relevant superconducting quantum computers	By the end of the decade	Target
Quantinuum Apollo	Universal, fully fault-tolerant computing	By the end of the decade	Target
IonQ	800 logical qubits in 2027; 80,000 logical and 2 million physical by 2030	2027–2030	Target
IBM Blue Jay	2,000 logical qubits running a billion operations	2033	Target
PsiQuantum	A utility-scale, fault-tolerant machine near Brisbane	No public date	Company claim
DARPA Quantum Benchmarking Initiative	Test whether any approach reaches utility scale, meaning value exceeds cost	By 2033	11 companies in Stage B

The most useful independent signal is DARPA's. Its benchmarking initiative evaluates companies' designs rather than their press releases, and in March 2026 its managing director said it "now seems likely" that someone will build a utility-scale quantum computer by 2033, though it remains unclear who.

What will quantum computing look like in 2027?

Expect more logical qubits, the first machines built around them, more advantage claims that are then tested, and security deadlines that force action. Don't expect broadly useful business computing in 2027.

• First logical-qubit machines. Magne, with 50 logical qubits, is due to be ready for its first tasks around the turn of 2026/27, and IonQ plans first deliveries of its Superion systems in 2027.
• Deeper circuits. IBM targets 10,000-gate circuits on Nighthawk in 2027, and IonQ targets 800 logical qubits.
• More claims, more checks. Advantage claims will keep arriving, and some will be overturned by classical methods, as three already have been.
• Standards and policy. NIST expects to finalise HQC, its backup post-quantum encryption standard, in 2027, and the US must update its National Quantum Strategy within 180 days of a June 2026 executive order.
• Security budgets. Post-quantum migration work, not quantum computing itself, is where most organisations will spend.

When will quantum computers break encryption?

No one knows, but the estimates keep shrinking. In 2025 Google estimated that RSA-2048 could be broken by about a million noisy qubits running for a week, 20 times fewer qubits than its 2019 estimate. In 2026 it estimated that 256-bit elliptic-curve cryptography, used by cryptocurrencies, could fall to fewer than 500,000 physical qubits in minutes.

Today's best machines have hundreds of high-quality qubits, so there's still a large gap. But a 2024 survey of 32 experts for the Global Risk Institute put the average likelihood of a "cryptographically relevant" quantum computer within 10 years at about 34% on an optimistic reading, up from 31% a year earlier, and about 19% on a pessimistic one.

Definition

Harvest now, decrypt later — adversaries collect encrypted data today, intending to decrypt it once a capable quantum computer exists. NIST cites this threat as the reason migration can't wait: data that must stay confidential for years is exposed now.

The replacement algorithms already exist. NIST published its first three post-quantum standards in August 2024 — ML-KEM for key exchange and ML-DSA and SLH-DSA for signatures — and the deadlines are set:

Date	Deadline	Who
2027	HQC backup key-exchange standard finalised	NIST
2029	Google's own post-quantum migration	Google
31 Dec 2030	Post-quantum key exchange on US high-value and high-impact federal systems	US executive order, June 2026
After 2030	RSA-2048 and equivalent elliptic-curve keys deprecated	NIST IR 8547 (draft)
31 Dec 2031	Post-quantum signatures on the same systems	US executive order, June 2026
After 2035	All quantum-vulnerable public-key algorithms disallowed	NIST IR 8547 (draft)

Common misconception

"Quantum computers will break encryption next year" — or "it's decades away". Both are wrong. Current estimates need roughly half a million to a million high-quality physical qubits, far beyond today's machines, but experts put meaningful odds on it within a decade, and harvest-now attacks make the risk real today for long-lived data.

What does quantum computing mean for the UAE?

The UAE builds its own small quantum computers and is preparing for the security risk. Abu Dhabi's Technology Innovation Institute (TII) runs superconducting quantum processors of 5 to 25 qubits, offered through a cloud service since February 2026, and the national encryption policy includes a post-quantum section.

Area	Where the UAE stands
Hardware	TII's Quantum Research Center began building the UAE's first quantum computer in August 2021; its lab now operates several superconducting systems of 5 to 25 qubits, including in-house chips
Access	A cloud service for TII's quantum processors since February 2026, initially for TII partners, later integrated with NVIDIA's CUDA-Q platform
Simulation	A 500,000-qubit quantum annealing simulation on NVIDIA GPUs, which is classical computing, not quantum hardware
Security policy	The UAE Cyber Security Council's National Encryption Policy includes a post-quantum cryptography section on protecting non-public information from future quantum threats
Academia	Khalifa University hosted the GCC's first international conference on quantum science in December 2025
Region	Quantinuum signed a non-binding memorandum with Saudi Aramco in Riyadh in August 2026

What this means

For UAE organisations, the actionable part of quantum computing in 2026 is security, not computing. The national encryption policy already points to post-quantum preparation, and anything with long-lived confidential data — government, banking, health and property records — should be on the migration list.

What does quantum computing mean for real estate and data centres?

Two things. Quantum computers are starting to become data-centre tenants, and property records protected by today's encryption will need a migration plan.

Quantum machines as tenants. IBM will build Starling in a new IBM Quantum Data Center in Poughkeepsie, New York. PsiQuantum is building in Chicago and at Moreton Bay near Brisbane, and Magne will sit in Copenhagen. IonQ says its new Superion system is built for data centres, fits a standard server-rack footprint and draws less power than a rack of GPUs — a vendor claim, but a sign of where the industry wants to go. For landlords, quantum is a small, specialised tenant category next to AI; the economics of the asset class are in data center real estate investment.

Encryption exposure. Title deeds, leases, mortgages and tenant records are kept for decades, building systems run for 10 to 20 years, and blockchain-based property tokens rely on elliptic-curve signatures of the kind Google's 2026 estimate addresses. Owners and platforms, including those behind tokenized real estate in Dubai, should ask vendors how and when they'll move to post-quantum standards.

How should businesses prepare for quantum computing?

Treat quantum as a security project now and a computing option later.

1. Inventory your cryptography. List where public-key encryption and signatures are used. The US order calls for a "cryptographic bill of materials", a useful model for anyone.
2. Prioritise long-lived data. Anything that must stay confidential beyond 2030 is exposed to harvest-now attacks today.
3. Ask vendors for their post-quantum plan. Look for support for NIST's ML-KEM and ML-DSA standards and a migration date.
4. Discount advantage claims. Ask what classical method a result was compared with, and whether independent groups have tried to beat it.
5. Experiment cheaply. Cloud services from the major vendors — and, for its partners, TII in the UAE — let teams test chemistry or optimisation ideas without buying hardware.
6. Watch the milestones. Logical-qubit counts, error rates and DARPA's evaluations say more than raw qubit numbers.

Quantum is the long-horizon member of this site's next-generation computing series; the nearer-term network shift is covered in 6G explained.

Final takeaway

Quantum computing crossed a real threshold in 2024–2026: error correction works, logical qubits beat physical ones, and a few physics experiments appear to outrun classical supercomputers. It hasn't crossed the commercial threshold. Useful, fault-tolerant machines are company targets for 2029 to 2033, and many claims along the way won't survive scrutiny. The part that can't wait is security: the standards exist, the deadlines are set, and data harvested today could be read tomorrow. For most organisations in the UAE and elsewhere, 2027 is the year to plan the post-quantum migration and watch the machines, not to buy them.