The Applied Quantum Technologies Institute (AQTI) issued a press release on March 2, 2026, under the headline “Cybersecurity Apocalypse in 2026 with a New Algorithm According to The Advanced Quantum Technologies Institute.” AQTI promoted a factoring method called the Jesse-Victor-Gharabaghi (JVG) algorithm. “Research extrapolations suggest it will require less than 5,000 qubits to break encryption methods used in RSA and ECC,” AQTI wrote.
The research is a preprint by Jesse Van Griensven Thé, Victor Oliveira Santos, and Bahram Gharabaghi of the University of Guelph’s School of Engineering. The authors posted the first version in October 2025 and the third, current version on February 14. Preprints.org labels the PDF “Not peer-reviewed version.”
In the preprint, the authors wrote that their projected values “should not be understood as realistic resource estimate for breaking RSA-sized keys.”
The preprint gives no qubit count for breaking RSA-2048. AQTI did not say whether the 5,000 qubits in its release were physical or logical. Physical qubits are the individual hardware devices in a quantum processor. In error correction, each logical qubit is encoded across many physical qubits to protect a long computation from errors.
Editor’s note: The Applied Quantum Technologies Institute is not affiliated with Applied Quantum, the quantum systems integration and consulting firm founded and led by Quantum Observer’s Editor-in-Chief Marin Ivezic.
The JVG method
In Shor’s algorithm, a quantum computer factors a number by finding the period of modular exponentiation. Modular exponentiation is the repeated multiplication of a base number, keeping only the remainder after each division by the number being factored. The period is the number of multiplications after which those remainders repeat.
The computer evaluates that function on many inputs at once and extracts the period with a quantum Fourier transform. That transform is a circuit that converts a repeating pattern into a measurement result from which the period can be calculated.
The JVG authors replaced the quantum Fourier transform with a quantum number theoretic transform. A number theoretic transform is a Fourier-type calculation done in modular integer arithmetic instead of complex numbers.
The authors wrote that the modular exponentiation values “were computed classically using NumPy and then encoded into the quantum circuit using amplitude encoding as a data-embedding step.” Computing a value classically means calculating it on an ordinary computer. NumPy is a numerical library for the Python programming language.
In amplitude encoding, researchers load a list of ordinary numbers into a quantum state as its amplitudes, one number for each possible measurement outcome. The amplitudes determine how likely each outcome is.
What the authors ran
The authors compared JVG with IBM’s reference implementation of Shor’s algorithm in IBM’s Qiskit Aer simulator and on ibm_torino, an IBM Heron processor with 133 physical qubits. Qiskit Aer is IBM software that simulates quantum circuits on an ordinary computer.
In the abstract, they reported that by using JVG in simulation, they cut runtime from 174.1 to 5.4 seconds, memory from 12.5 to 0.27 GB, and gate counts by about 99%.
On the IBM processor, they reported a runtime reduction from 67.8 to 2 seconds and a gate-count reduction of more than 98%. They wrote that the two methods used registers of different sizes and that the reported gate and depth reductions are end-to-end resource budgets. A register is a group of qubits treated as one unit in a computation.
The authors ran the Shor baseline in simulation on numbers of up to five digits and on the IBM processor on numbers of up to four digits. They reported that on the hardware, JVG “was able to execute circuits corresponding to a 75-digit composite number using a 70-qubit configuration.” The 70 qubits were physical qubits on the IBM processor.
An RSA-2048 modulus, the 2,048-bit number an attacker would have to factor to recover an RSA private key, has 617 decimal digits.
The RSA-2048 projection
In the abstract, the authors wrote that JVG would need “a projected quantum runtime of 11 hours for a factorization” of an RSA-2048 key under identical scaling assumptions. In the body of the preprint, they wrote that “one thousand attempts would scale the JVG estimates to roughly 11 hours.” They added that the projections “should therefore be interpreted as indicative scaling trends rather than precise forecasts.”
In their introduction, the authors cited a 2025 estimate by Google Quantum AI researcher Craig Gidney. In a preprint, Gidney estimated that a machine with less than a million noisy physical qubits could factor a 2,048-bit RSA integer in less than a week. He assumed a gate error rate of 0.1%, or about one error in every 1,000 operations.
What AQTI said
The release used the name Advanced Quantum Technologies Institute in its headline, first paragraph, and contact line, and Applied Quantum Technologies Institute in its source line and company description. AQTI’s website uses the Applied name.
“The lesson from JVG is that the timeline is accelerating not only because hardware advances, but also because algorithms improve,” Van Griensven Thé, the preprint’s first author, said in the release. “That is why post-quantum upgrades must be treated as urgent infrastructure work.”
Migration timeline
In a November 2024 draft transition plan, NIST proposed deprecating RSA and elliptic-curve algorithms at 112 bits of security strength after 2030 and disallowing them at all strengths after 2035. A 2,048-bit RSA key has 112 bits of security strength.
Update, November 12, 2026: On March 7, Scott Aaronson, a computer scientist at the University of Texas at Austin, wrote that the method computes the modular exponentiation values on an ordinary computer and that the number of those values grows exponentially with the size of the number being factored. He wrote that the method “can only look like it wins on tiny numbers” and is hopeless on large ones. The authors published the work in the Journal of Cybersecurity and Privacy on April 10, with a projection of 29 hours for ten thousand attempts and caveats of its own. On October 27, NIST finalized its transition plan, which deprecates RSA and elliptic-curve algorithms at every approved key size after 2030.
