Intel Agrees to Buy Quantum Software Maker Classiq Before It Has a Quantum Computer to Sell

October 26, 2026 – Intel Corporation announced a definitive agreement to acquire Classiq, a quantum computing software company based in Tel Aviv. The deal is subject to regulatory approvals and other customary closing conditions. Intel expects to close it in the first half of 2027. Until then, the two companies will operate separately.

Intel did not disclose the price or the form of payment. Calcalist reported a price of $2.3 billion in cash and stock. The Next Web reported the same figure from Bloomberg, which cited people familiar with the matter, and said Intel declined to comment on it. Globes put the effective price at about $2.2 billion after deducting roughly $100 million of cash on Classiq’s balance sheet. Calcalist also reported that Intel has allocated more than $100 million in retention grants for Classiq employees.

Intel said it would combine Classiq’s synthesis platform with its own spin-qubit manufacturing and quantum software into a full-stack quantum computing platform. With that platform, Intel said, it could add quantum processing units (QPUs) to its portfolio of CPUs, GPUs and application-specific chips. Intel compared Classiq’s role with that of electronic design automation (EDA) software, the tools engineers use to design chips without laying out each transistor by hand. Intel chief executive Lip-Bu Tan said that “joining forces with Classiq transforms Intel’s Quantum Computing Group into a full-stack quantum computing business.”

In the release, Intel said that after closing it intends to keep supporting Classiq’s existing integrations with other hardware makers and to keep the software independent of qubit modality and hardware. Intel also said it is the only company able to produce qubit chips and complex cryogenic control electronics at scale with the equipment and processes of conventional semiconductor manufacturing.

Nir Minerbi, Classiq’s co-founder and chief executive, said the company has worked to let developers, researchers and enterprises build quantum applications without deep expertise in quantum circuit design.

What Classiq Makes

Minerbi founded Classiq in 2020 with Yehuda and Amir Naveh. Developers describe what a quantum algorithm should do in Qmod, Classiq’s high-level modeling language, and use Classiq’s synthesis engine to generate an optimized gate-level circuit for the machine they choose. In the company description attached to Intel’s release, Classiq says dozens of global enterprises use its platform. It describes the platform as combining AI agents, the modeling language and design automation, with support for benchmarking and execution on different vendors’ machines.

In May 2025, Classiq raised $110 million in a Series C round, which it described as the largest round ever for a quantum software company. At the time, it named BMW, Citi, Deloitte, Rolls-Royce, Mizuho and Toshiba as customers. It listed direct integrations with quantum hardware from IBM, IonQ, QuEra, Quantinuum, OQC, AQT, Alice & Bob and Rigetti, and with simulators including Nvidia’s and Intel’s. Customers could also deploy through AWS Braket, Azure Quantum and Google Cloud. In November 2025, AMD Ventures, Qualcomm Ventures and IonQ joined a further strategic round, after which Classiq’s total funding stood above $200 million.

On September 30, 2026, Classiq announced a Fault Tolerance Engine. It describes the engine as translating logical programs into fault-tolerant execution plans, with estimates of physical qubits, error-correction cycles, code distance and runtime. Classiq gave no performance figures in the announcement.

According to Calcalist, Classiq employs around 150 people and has annual revenue in the tens of millions of dollars, well below a $100 million run rate. Calcalist estimated Classiq’s valuation in the 2025 Series C at $600 to $700 million. Neither company has disclosed Classiq’s financial results.

Tan’s Prior Investment in Classiq

When Classiq announced its $33 million Series B in February 2022, it listed personal investments from two veterans of chip design software: Lip-Bu Tan, then executive chairman of Cadence Design Systems, and Harvey Jones, a former chief executive of Synopsys. In the same release, Classiq said Cadence and Synopsys do for electronic design what Classiq does for quantum software design. Tan became Intel’s chief executive in March 2025.

SiliconANGLE reported that Tan also took part in the 2025 Series C. Classiq did not name him in its Series C announcement. A September 2026 Calcalist article described him as an existing investor at the time of that round.

Calcalist reported, citing sources close to Classiq, that Tan recused himself from discussions about the acquisition. Intel did not mention Tan’s holding in its announcement or say how the transaction was reviewed.

Intel’s Quantum Program

Intel says it has invested in quantum computing since 2015. It builds silicon spin qubits, which store information in the spin of a single electron held in a transistor-like structure. Intel’s best-known research chip, the 12-qubit Tunnel Falls, went to research laboratories in June 2023. Intel made the chip on 300-mm wafers at its D1 fab in Oregon, using extreme ultraviolet (EUV) lithography. The first recipients were the Laboratory for Physical Sciences, Sandia National Laboratories and the universities of Rochester and Wisconsin-Madison. At the launch, Intel said a next-generation chip based on Tunnel Falls was expected in 2024.

In May 2024, Intel reported wafer-scale results in a Nature paper. On one 300-mm wafer, 96% of 232 tested devices had all components functional, including the 12 quantum dots and four charge sensors. In one device made with isotopically purified silicon-28, Intel’s researchers measured single-qubit gate fidelities of about 99.9%.

Intel also designs cryogenic control chips. QuTech and Intel researchers used Intel’s Horse Ridge controller, operating at about 3 kelvin, to control silicon spin qubits and run a two-qubit algorithm. They published the results in Nature in 2021.

In January 2026, Q-NEXT, a US Department of Energy research center hosted at Argonne National Laboratory, deployed a 12-qubit quantum dot device built by Intel to study how such systems behave as qubits are added. In March 2026, Mayer Feldman, an Intel and Princeton researcher, presented results from Tunnel Falls run as a four-qubit exchange-only processor at the American Physical Society’s Global Physics Summit. In his abstract, Feldman describes a multi-round, distance-2 repetition code that detects bit-flip or phase-flip errors.

On July 22, 2026, Hitachi announced a silicon quantum computer project with Intel’s Japanese subsidiary and Japan’s National Institute of Advanced Industrial Science and Technology (AIST), under a research program of NEDO, a Japanese government agency. Using a process design kit from Intel, Hitachi will design qubit chips for Intel’s quantum-device processes, which are based on Intel 18A technology. Hitachi aims to launch a cloud service for external researchers through AIST in fiscal 2027. Its later targets are a 100-qubit prototype running quantum error-correcting codes in fiscal 2028 and a 1,000-qubit prototype in fiscal 2030. Hitachi’s fiscal year begins in April.

In an interview on the QuantumJane podcast, reported by Calcalist in September 2026, Tan said about 70 people work on quantum computing at Intel. He described quantum computing as part of Intel’s strategic direction for the next five to 10 years.

Intel will report third-quarter results on Thursday, October 29, after the close of US trading.

My Analysis

Intel has agreed to pay a reported $2.3 billion for the part of the quantum stack that customers work with directly, at a time when it sells no quantum computer and runs no public cloud service on its own quantum hardware. Whether Intel earns that money back depends heavily on whether Classiq’s customers keep trusting Intel to optimize Classiq as carefully for IBM, IonQ and Quantinuum machines as for an Intel QPU, once one exists.

Intel Is Buying the Software Layer Before It Sells Hardware

IBM, IonQ and Quantinuum each pair their software with commercial access to their own quantum computers. IBM sells cloud access to its superconducting processors and gives away Qiskit to attract developers to them. IonQ and Quantinuum sell time on their trapped-ion systems, directly and through the large cloud providers. Intel supplies research chips to national laboratories and universities, and a process design kit to Hitachi. It does not sell a quantum computer or run a public cloud service on its own hardware.

Intel’s researchers have kept working since Tunnel Falls: they built the device Argonne deployed in January and presented the four-qubit exchange-only result in March. On its quantum research page, though, Intel still calls Tunnel Falls “Intel’s most advanced silicon spin qubit chip to date.” I could not find a public release of the successor Intel promised for 2024.

Intel’s full-stack quantum business does not exist yet. What Intel has agreed to buy is a software business whose customers run their programs on other companies’ hardware and on simulators. An Intel QPU could become another Classiq target if Intel brings one to market.

I think the strategy is coherent. Intel will probably wait years for revenue from its quantum hardware, while the enterprises that may later buy quantum computing time are choosing their development tools now. If Intel waited for its own QPU before building a software business, it would arrive with hardware and few users. With Classiq, Intel would get enterprise customers, a team Calcalist puts at around 150 people, and a compiler that already targets many leading gate-based machines. A Classiq user could later retarget an existing model to an Intel QPU, within that machine’s qubit count and error rates.

I read the Hitachi project as a second part of the plan. In that model, Intel would fabricate qubit chips on its 18A-based process and supply the design kit. Partners such as Hitachi would build and run the machines, and developers would program them with Intel’s software. Intel has not described its strategy this way.

Tan’s interest in the software side predates his job at Intel. He ran Cadence from 2009 to 2021. In the QuantumJane interview, as quoted by Calcalist, he said Classiq wanted to be “the next Cadence” for quantum computing. As Intel’s chief executive, he is now acting on a thesis he backed as a private investor in 2022.

The EDA Comparison and the Ownership Question

Intel and Tan both describe Classiq as the quantum equivalent of EDA. On the technology, I think the comparison is fair. A Classiq user writes a functional description and gets back an optimized implementation for a specific target, much as a chip designer does with logic synthesis.

Cadence and Synopsys became the standard suppliers of chip design software without belonging to any chipmaker. TSMC, Samsung and Intel Foundry each certify the two vendors’ tool flows for their process technologies. A design team that moves between foundries still needs a new process design kit and substantial rework, but it can keep its EDA vendor. Classiq will instead belong to a company that plans to sell its own QPUs.

A hardware company can still support its competitors’ machines. IBM sells its own quantum computers, yet developers can use Qiskit with other vendors’ machines through provider plugins. Quantinuum also builds hardware, yet it publishes open-source TKET extensions for IBM and AWS Braket backends.

After closing, the engineers who decide how much work goes into Classiq’s IBM, IonQ and Quantinuum backends will work for a company that wants its own QPU to succeed. Several of Classiq’s partners also compete with Intel. IonQ invested in Classiq in November 2025, alongside the venture arms of AMD and Qualcomm, and builds customer applications with Classiq. In February 2026, Classiq, Comcast and AMD reported a network-routing trial in which they ran Classiq’s algorithms on IonQ hardware and in simulations accelerated by AMD Instinct GPUs. Each of those partners will soon work with software that a competitor owns.

In 2010 Intel settled an antitrust case with the US Federal Trade Commission without admitting a violation. The FTC had alleged, among other things, that Intel redesigned its compiler in a way that held back the performance of competitors’ processors, and did not disclose this to customers. Under the settlement order, Intel had to disclose that its compiler “may or may not optimize to the same degree for non-Intel microprocessors.” It also had to reimburse compiler customers who had relied on Intel’s representations and chose to recompile with another vendor’s compiler, up to $10 million in total. That case concerned CPUs and is sixteen years old. It is still the closest precedent I know for an Intel-owned compiler in a market where Intel also sells hardware.

I take Intel’s multi-platform commitment at face value for the next few years. Without third-party backends, Classiq’s customers could run their programs only on simulators, and few would stay. Customers will learn how firm the commitment is once Intel has a QPU of its own and its engineers have to decide where the next round of optimization work goes.

A customer who suspected bias could move work to Qiskit or TKET, at the cost of rewriting Qmod models in another framework. In my piece on Quantum Open Architecture, I described quantum computers assembled from layers that specialist vendors supply, and I listed Classiq among the independent providers of the software abstraction layer. If the deal closes, Intel will take one of the best-funded of those providers off the independent market.

Intel’s “Only Company” Manufacturing Claim

Intel’s manufacturing experience is real. Few organizations can put spin-qubit test chips through a production 300-mm line with EUV lithography, as Intel did for Tunnel Falls. Intel has also published wafer-scale yield data for those devices. With Horse Ridge, it has shown that it can design cryogenic control chips. Hitachi chose Intel’s 18A-based process for its own program. Intel has not, however, defined “at scale” or said how it knows that no other company can do the same. I find the “only company” part hard to reconcile with what other groups have published.

IBM agreed to buy HRL Laboratories in July and completed the purchase on August 26. As I covered in my analysis of HRL’s Nature cover paper, HRL controlled a 54-quantum-dot chip, configurable as up to 18 exchange-only qubits, with a controller chip of about 70 million transistors that operated inside the cryostat. HRL had the controller made in a commercial 130-nm CMOS process. Using up to seven qubits, HRL ran repetition-code error correction without real-time input from room-temperature electronics. Quantum Motion installed a full-stack silicon quantum computer, made with 300-mm processes at commercial chip foundries, at the UK’s National Quantum Computing Centre in September 2025.

Neither HRL nor Quantum Motion has demonstrated volume production. Each has shown part of what Intel describes as unique: HRL runs silicon spin qubits with cryogenic control electronics made at a commercial foundry, and Quantum Motion makes its qubits at commercial 300-mm foundries. Until Intel says what “at scale” means, I cannot treat its “only company” claim as established.

The US Defense Advanced Research Projects Agency (DARPA) runs the Quantum Benchmarking Initiative, in which it assesses whether any quantum computing approach can reach utility-scale operation by 2033. In November 2025, DARPA moved 11 companies into the program’s second stage, four of them with silicon-based qubits: Diraq, Photonic, Quantum Motion and Silicon Quantum Computing. On October 7, 2026, Diraq advanced to the final stage with Atom Computing, IBM and IonQ.

Intel does not appear on DARPA’s current participant list. DARPA says it lists only companies with an active scope of work, so I cannot tell from it whether Intel applied or how DARPA would rate Intel’s approach.

In my assessment, silicon spin qubits have one of the clearest routes to semiconductor-scale manufacturing, and Intel’s process capability is a real advantage on that dimension. Among the published results I know of, HRL has gone furthest on integration: it generated all of the qubit chip’s time-varying control signals with a cryogenic controller during multi-qubit error-correction experiments. With Classiq, Intel’s quantum group would get software and customers. It would get no additional qubits, no higher gate fidelity and no answer to the control-wiring problem.

Classiq’s Price Against Its Funding, Valuation and Revenue

The reported $2.3 billion is roughly three to four times the $600 to $700 million valuation that Calcalist attributes to Classiq’s 2025 Series C. It is roughly ten times the $200 million-plus Classiq has raised. At $100 million of annual revenue, the price would be 23 times revenue. Calcalist puts Classiq’s revenue well below that, so the actual multiple is higher. Neither company has disclosed the figures needed to calculate it.

Intel would be paying for position: the team, the customer relationships and an early place in quantum software. According to Calcalist, Intel has allocated more than $100 million in retention grants, which I read as Intel counting the team as a large part of what it is buying. I expect Intel to earn most of the return only after fault-tolerant machines exist. In my Quantum Utility Ladder article, I mapped which applications become possible at each size of fault-tolerant machine.

SiliconANGLE called this Intel’s second-largest acquisition after Mobileye. At least three of Intel’s earlier acquisitions were larger by announced deal value: Altera ($16.7 billion, announced 2015), Mobileye ($15.3 billion, announced 2017) and McAfee ($7.68 billion, announced 2010). Forbes Israel and The Jerusalem Post make a narrower claim that holds: this is Intel’s largest Israeli acquisition since Mobileye, itself an Israeli company.

Calcalist called the deal the most expensive acquisition of a quantum company to date, ahead of IonQ’s roughly $1 billion purchase of Oxford Ionics. At the reported price, it is also larger than IonQ’s purchase of the contract chip foundry SkyWater, announced at an equity value of about $1.8 billion and completed on July 31, 2026. Intel may disclose the price in a later filing or earnings report, but it has not said it will.

Tan’s Stake and Intel’s Governance Record

Tan invested in Classiq in 2022, three years before he became Intel’s chief executive, and Classiq announced the investment at the time. His interest in quantum design software follows from his years at Cadence. I have seen nothing to suggest the purchase is anything other than strategic.

Intel’s investors will still read the deal against earlier reporting. In December 2025, Reuters reported that Intel’s board had initially rejected Tan’s 2025 proposal to buy Rivos, an AI chip startup he chaired. According to Reuters, the board said Tan had a conflict of interest and lacked an AI strategy to justify the deal. After Meta showed interest in Rivos, Intel made its own offer, and Meta went on to buy the company. Reuters also reported, citing two sources, that Intel had adopted a policy barring Tan from attending or voting in decision meetings where he has a conflict.

In April 2026, Reuters reported that a late-March Intel securities filing referred to four unnamed companies whose financings were material to disclose because of their size and their benefit to Tan. Reuters identified them as EPIC Microsystems, 3D Glass Solutions, OPAQUE Systems and SambaNova. Intel told Reuters it maintains rigorous governance and conflict-of-interest policies. It also said it had been a shareholder in three of the four companies before Tan joined.

On Classiq, only Calcalist has reported a recusal so far, citing sources close to Classiq. Globes described Tan as the driving force behind the acquisition. Both accounts can be accurate if Tan set Intel’s quantum strategy and stayed out of the transaction itself. Only Intel can say whether that is what happened.

Intel could answer three questions in a short statement: whether Tan recused himself, which board committee approved the deal, and what Tan receives at closing. According to Intel’s 2026 proxy statement, the audit committee of Intel’s board oversees related-party transactions. If Intel says nothing earlier, its 2027 proxy statement is the most likely place for any required disclosure.

What Classiq Customers Should Do Before the Deal Closes

For enterprises with Classiq in their quantum programs, my recommendation is to keep using it. I would spend the months before closing on three things:

• Get backend support in writing. In a renewal signed now, a customer can set out which hardware targets stay supported, for how long, and how much notice it gets if one is dropped.
• Run a test exit. Teams can export synthesized circuits from Classiq to OpenQASM 2.0 and 3.0, QIR, Q# and several vendor formats, and run them through another toolchain. Qmod and the synthesis engine are Classiq’s own, so a team that leaves would keep its exported circuits but would have to rebuild its models and synthesis constraints in another framework. On one representative workload, a team can list everything it would have to rebuild.
• Benchmark the output. Teams running on IBM, IonQ or Quantinuum hardware should compare Classiq’s output with the output of the hardware vendor’s own toolchain on equivalent workloads, after each major Classiq release or backend change.

Hardware Timelines and Industry Structure

Intel announced no new hardware result with the deal: no qubit counts, no fidelities and no logical-qubit demonstrations. I see nothing in it that would move timelines to fault-tolerant machines or to a cryptographically relevant quantum computer. Tan himself places quantum computing in Intel’s strategy for the next five to 10 years.

Researchers and compiler developers can still shorten those timelines. The number of physical qubits a fault-tolerant program needs depends on how developers compile it for error correction. With better synthesis, a developer can run the same algorithm on fewer qubits and in a circuit with fewer layers of gates. In a 2025 paper, Craig Gidney cut the estimate he and Martin Ekerå had published in 2019 for factoring RSA-2048, from 20 million noisy qubits to under a million, with the same hardware assumptions. He used new arithmetic and error-correction techniques, and accepted a longer runtime: under a week instead of eight hours. Classiq has published no comparable result, and no performance figures for its Fault Tolerance Engine.

Large companies in quantum computing spent 2026 buying the layers they lacked. IonQ completed its purchase of the chip foundry SkyWater in July. IBM announced Anderon, a quantum wafer foundry, in May, and completed its purchase of HRL in August. Intel has now agreed to buy Classiq. Fewer independent companies remain between the hardware makers and the users. Of those that remain, Q-CTRL, Riverlane and Qedma are the ones I would watch for the next acquisition.

From the author