Intel Just Paid $2.3 Billion for Quantum Software. Your Encryption Clock Just Got Louder
On October 26, Intel signed a definitive agreement to acquire Classiq, a quantum software company based in Tel Aviv. Intel did not disclose the price. Calcalist, Haaretz and Bloomberg, citing people familiar with the matter, put it at about $2.3 billion in cash and shares. If that number holds, it is the largest acquisition in quantum computing so far, and Intel's first major acquisition in four years.
The deal is expected to close in the first half of 2027, subject to regulatory approvals. Until then the two companies operate separately. That is the news. What matters more is what Intel actually bought, and what it tells the rest of us about the timeline we are working against.
What Classiq actually does
Classiq does not build quantum computers. It builds the layer between a human idea and a quantum circuit.
You describe what a quantum program should do in a high-level model, and Classiq's synthesis platform produces the circuit that a quantum computer runs. It handles the qubit allocation, the routing, the gate decomposition, the hardware constraints. Think of it the way chip designers think about EDA tools: nobody hand-draws a billion transistors anymore, and nobody serious will hand-write large quantum circuits either. Classiq is the compiler for that world.
The company was founded in 2020 by Nir Minerbi, Yehuda Naveh and Amir Naveh. It employs about 150 people and holds more than 60 patents around its synthesis technology. Its customer list reads like a roll call of industries that cannot afford to get the quantum timeline wrong: BMW, Citi, Rolls-Royce, HSBC, Mizuho. These are not research grants. They are enterprises paying to be ready.
Why Intel wants it
Intel's own quantum work is built on silicon spin qubits, the approach that fits most naturally with the semiconductor manufacturing Intel already owns. The company has disclosed a qubit chip tape-out on a process derived from its 18A node. What it did not have was the software layer that turns that hardware into something customers can actually use.
Intel's CEO Lip-Bu Tan said the combination "transforms Intel's Quantum Computing Group into a full-stack quantum computing business." That is the language of a company that watched itself miss mobile and then miss the first wave of AI accelerators, and decided not to miss a third shift. You can argue about whether the comparison is fair. You cannot argue that $2.3 billion is a casual bet.
One detail matters for everyone who is not Intel: the company says Classiq's software will stay hardware-agnostic, with its existing multi-platform integrations intact. The platform is supposed to keep supporting other quantum hardware after the deal closes. That promise is worth watching, because the moment a compiler like this becomes exclusive to one vendor's qubits, the whole field gets narrower. For now, the stated plan is the open one.
What this has to do with your encrypted messages
Here is the part I keep coming back to, because I build an encrypted messenger.
Nothing about this deal means a cryptographically relevant quantum computer exists today. It does not. The most aggressive public estimates still talk about thousands of logical qubits to break RSA-2048, and no machine close to that exists. Intel bought a software company, not a codebreaking machine.
But the deal does move the clock, and it moves it in a specific way. Quantum computers become dangerous to encryption in two stages. First, someone has to build the hardware. Second, someone has to make that hardware usable enough to run a real algorithm on it. Classiq works on the second stage. Every improvement in synthesis, every reduction in circuit depth and qubit count, lowers the bar for stage one. The Alice and Bob preprint from October 28 estimated that fewer than 23,000 physical cat qubits could factor RSA-2048. That number keeps drifting downward as the software gets better, not because the physics changed.
That is the same logic behind "harvest now, decrypt later." An attacker does not need a quantum computer today. They need your encrypted traffic recorded today, and a quantum computer at some point before that data stops mattering. Every week the quantum software stack gets better, the value of what is being recorded right now goes up.
The standards world already answered this. On October 27, one day after the Intel announcement, NIST published the final version of its transition report, IR 8547. Quantum-vulnerable public-key cryptography gets deprecated after 2030 and disallowed under NIST standards after 2035. For the most sensitive federal systems, post-quantum key establishment is due by the end of 2030 and post-quantum signatures by the end of 2031. These dates are not guesses about when a quantum computer arrives. They are a schedule for getting out of the way before it does.
What I did about it in my own project
I build a small end-to-end encrypted messenger, an experimental project I run myself. It uses a hybrid key exchange: ML-KEM-768, the NIST-standardized lattice KEM, combined with a classical elliptic-curve exchange, plus ML-DSA-65 signatures for identity. Key fingerprints are shown as safety numbers so two people can verify each other's keys out of band.
I will be straight about what that is and is not. It is a real, working implementation that I test on my own devices. It is not independently audited, it has no users beyond my own testing, and I do not pretend otherwise. I mention it because the decision it forced is the same decision every team building encrypted software faces right now: switch the key exchange early, while it is cheap, or keep shipping classical-only key exchange and hope the harvest does not include you.
Switching early was not hard. ML-KEM-768 keys and ciphertexts are a little over a kilobyte each, and encapsulation runs faster than an X25519 operation on most hardware. The hard part was not the algorithm. It was deciding that the migration was worth doing before anyone forced me to. Intel just spent $2.3 billion as its answer to that same question, at a slightly larger scale.
Three things worth doing now
First, take inventory. You cannot migrate cryptography you have not listed. Find every place your system uses RSA, ECDSA, EdDSA or Diffie-Hellman, including the places you forgot: old API clients, internal services, backup encryption, signed update channels.
Second, go hybrid first. Pair ML-KEM with the classical algorithm you already trust, the way browsers and Windows Schannel already do with X25519MLKEM768. Hybrid is a hedge: if the new algorithm breaks, the classical half still holds, and if the classical half falls to a quantum computer, the post-quantum half still holds. The standards bodies recommend it for exactly this reason, and it is what the real deployments are shipping.
Third, treat TLS 1.3 as a precondition, not an upgrade. The post-quantum groups in Windows, in browsers, in Cloudflare's stack, all ride on TLS 1.3. Anything still negotiating TLS 1.2 gets no quantum-resistant protection from any of this, no matter how many patches you install.
The signal under the noise
Acquisitions are not technical milestones. No qubit count was announced, no benchmark was published, and the deal still needs regulatory approval. But $2.3 billion is a serious amount of money to pay for a company whose product only matters if quantum computing becomes real and usable. Intel is betting that it will, and that the software layer is where the value concentrates first.
The safe way to read that bet is simple: the people with the most to lose from being late are moving now. If you build anything that encrypts data meant to stay secret for years, you are on the same timeline they are. The cheapest day to start migrating was the day the NIST standards were published in August 2024. The second cheapest day is today.
Sources: Intel's announcement of the definitive agreement (Oct 26, 2026), as reported by PostQuantum.com and Haaretz; NIST IR 8547 final (Oct 27, 2026), covered in this summary.
Top comments (0)