Roadmap to a large error-corrected
quantum computer
Google Quantum AI's mission is to lead development of large, error-corrected quantum computing and make its benefits universally accessible and useful
Logical qubit prototype
✔
M1 (2019)
M2 (2023)
M3 (2025+)
M6
# physical qubits
M5
M4
[logical qubit error rate]
54 [-]
102 [10-2]
103 [10-6 ]
104 [10-6 ]
105 [10-8]
106 [10-13]
Beyond classical benchmark
✔
1 long-lived logical qubit
Tileable module (logical gate)
Engineering
scale up
Error-corrected (EC) quantum computer
Our ~10 year roadmap to building the first large error corrected quantum computer defines our critical path of scientific demonstrations and engineering scale up milestones.
Logical qubit prototype
✔
M1 (2019)
M2 (2023)
M3 (2025+)
M6
# physical qubits
M5
M4
[logical qubit error rate]
54 [-]
102 [10-2]
103 [10-6 ]
104 [10-6 ]
105 [10-8]
106 [10-13]
Beyond classical benchmark
✔
1 long-lived logical qubit
Tileable module (logical gate)
Engineering
scale up
Error-corrected (EC) quantum computer
Our roadmap philosophy is to frontload invention, developing the scientific proofs-of-concept which are needed to make quantum computers work at scale
M2 (2022)
M3 (WIP)
M1 (2019)
(Nature)
(Nature)
Logical error rate per cycle
10-2
10-3
10-4
M2
M3
10-5
0 5 10 15 20
10-6
10-7
Code distance
0 5 10 15 20 25
Quantum error correction cycle, t
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
Logical error probability, pL
Cross-entropy benchmarking
fidelity, 𝐹XEB
12 14 16 18 20
Number of cycles, m
10-1
10-2
10-3
We are in the process of building a long lived logical qubit (M3)
Running the Milestone 1 Random Circuit Sampling benchmark on our 2023 quantum hardware, would take the world’s best supercomputer 1B years to reproduce.
We believe Effective Quantum Volume to be the most honest and useful measure of the capabilities of today’s quantum chips
The usefulness of a quantum chip is defined by the size of program you can run on it without noise degrading the quantum state of the output.
I.e. the number of quantum operations (i.e. gates) that can contribute to a measurement outcome.
The usefulness of a quantum chip cannot be determined by counting qubits alone. Qubit counts need to be coupled with information on eg qubit quality and error rates to assess the capabilities of a quantum chip.
Coherence,
Gate Fidelity
Qubit Quality
Overall error rate, Error correction
Error Rate
Number of qubits
Qubit Count
Operations per unit of time
Speed
How qubits can interact
Connectivity
As the industry progresses towards large scale quantum computing, it is critical that hardware companies work together and with suppliers to develop the quantum supply chain.
Scaling quantum computers requires lower component cost per qubit. Hardware companies can support suppliers’ growth to improve yield and achieve economies of scale.
Cost drivers: Cryostats | Wiring | Electronics
Google Quantum AI's mission is to lead development of large, error-corrected quantum computing and make its benefits universally accessible and useful
Logical qubit prototype
✔
M1 (2019)
M2 (2023)
M3 (2025+)
M6
# physical qubits
M5
M4
[logical qubit error rate]
54 [-]
102 [10-2]
103 [10-6 ]
104 [10-6 ]
105 [10-8]
106 [10-13]
Beyond classical benchmark
✔
1 long-lived logical qubit
Tileable module (logical gate)
Engineering
scale up
Error-corrected (EC) quantum computer
We believe that today’s experimental processors best serve the world as a research tool for advancing the field towards the era of commercially useful quantum computing.
experimental quantum processors
Our processors are already being used for new discoveries
Formation of robust bound states of interacting microwave photons
(Morvan et al., Nature 2022)
Noise-resilient edge modes on a chain of superconducting qubits
(Mi et al., Science 2022)
Quantum advantage in learning from experiments
(Huang et al., Science 2022)
Non-Abelian braiding of graph vertices in a superconducting processor
(Andersen et al., Nature 2023)
Traversable wormhole dynamics on a quantum processor
(Jafferis et al., Nature 2022)
Measurement-induced entanglement and teleportation on a noisy quantum processor
(Hoke et al., in review at Nature)
Unbiasing fermionic quantum Monte Carlo with a quantum computer
(Huggins et al., Nature 2022)
Dynamics of magnetization at infinite temperature in a Heisenberg spin chain
(Rosenberg et al., in review at Science)
Selected quantum applications research results
Google Quantum AI results
Selected other group results
Google Quantum AI ongoing research
Our next challenge is to demonstrate a useful beyond- classical computational application to cross over the quantum utility frontier
We offer a full stack Quantum Research Platform (hardware + software) to a small circle of expert users to run quantum programs on Google’s quantum hardware
the only beyond classical offering
a product designed to remain at the cutting edge of development
Beyond Classical
Chip Performance
Evergreen
Experimental
built by and for researchers
Large error-corrected quantum computers will achieve commercial and societal impact if the global quantum ecosystem collaborates on and funds fault-tolerant applications development
O(√N) vs. O(N/2)
O(N3) vs. O(2N)
O(NC) vs. O(2N)
O(NC) vs. O(2N)
unstructured search
(Grover’s algorithm)
factorization
(Shor’s algorithm)
quantum simulation
(e.g. physics, chemistry)
quantum ML
(esp. with quantum data)
O(NC) vs. O(2N)
differential equations
(e.g. oscillators)
Work with us in Qᴜᴀʟᴛʀᴀɴ, an open source library which accelerates
fault tolerant applications development! See github.com/quantumlib