With Amazon Braket, you can get started exploring quantum computing with the AWS Free Tier and the open-source Amazon Braket software development kit (SDK). Learn more about the AWS Free Tier with Amazon Braket below.
Beyond the AWS Free Tier, there are no upfront charges, and you only pay for the AWS resources you use. Amazon Braket provides access to quantum computers, quantum circuit simulators, fully managed execution of hybrid quantum-classical algorithms, and Jupyter notebook development environments.
You will be billed separately for use of each of these capabilities, as well as other AWS services that you use with Amazon Braket such as Amazon Simple Storage Service (S3) for storing the results of quantum computations.
If you need expert help as you explore quantum computing, our Amazon Quantum Solutions Lab professional services team and our technology and consulting partners can help. You can request information here.
There are two pricing components when using a quantum computer, or quantum processing unit (QPU), on Amazon Braket: a per-shot fee and a per-task fee.
A shot is a single execution of a quantum algorithm on a QPU. For example, a shot is a single pass through each stage of a complete quantum circuit on a gate-based QPU from IonQ, Rigetti, or OQC. When you use a D-Wave quantum annealer, a shot is when you obtain a result sample of a quantum annealing problem. The per-shot pricing depends on the type of QPU used. The per-shot price is not affected by the number or type of gates used in a quantum circuit or the number of variables used in a quantum annealing problem.
A task is a sequence of repeated shots based on the same circuit design or annealing problem. You define how many shots you want included in a task when you submit the task to Amazon Braket. Per-task pricing is the same across all QPUs.
With Amazon Braket, you have a choice of quantum circuit simulators to run and test quantum algorithms, including the free local simulator in the Amazon Braket SDK, and three fully managed on-demand simulators: State Vector 1 (SV1), Density Matrix 1 (DM1), and Tensor Network 1 (TN1). SV1 is a general purpose quantum circuit simulator, DM1 enables you to simulate the effect of noise on your circuits, and TN1 specializes in larger scale simulations for certain types of quantum circuits. All on-demand simulators automatically scale AWS computing resources to deliver high-performance execution of your quantum algorithms. The cost of using the Amazon Braket on-demand simulators is based on the duration of each simulation task. When you use on-demand simulators, you are billed at a per-minute rate, in increments of one millisecond, for the time your simulation takes to run, with a minimum billing duration of 3 seconds per simulation.
With Amazon Braket Hybrid Jobs, you have the additional flexibility to use embedded simulators, designed with high performance and ultra-low latency in mind. These simulators can be embedded within the same jobs container as your application code. Today, Amazon Braket Hybrid Jobs supports embedded simulators from PennyLane or the option to embed your own circuit simulator as a container using our bring-your-own-container feature. Since embedded simulators run on the same hybrid jobs instance as your application code, your use of embedded simulators could impact the duration of your use of a hybrid jobs instance. The cost for using embedded simulators forms part of your overall bill for using the hybrid jobs instance and is not charged for separately. For more details on the pricing for hybrid jobs instances, refer to the Amazon Braket Hybrid Jobs Pricing Page.
The cost of using the Amazon Braket on-demand and embedded simulators is based on the duration of each simulation task. You are billed at a per-minute rate, in increments of one millisecond, for the time your simulation takes to run. When you use on-demand simulators, you are billed for a minimum of 3 seconds per simulation.
Embedded simulators are only available with Amazon Braket Hybrid Jobs. For more details on the pricing for embedded simulators, refer to the Amazon Braket Hybrid Jobs Pricing Page.
Quantum Circuit Simulators
Amazon Braket Hybrid Jobs provides fully managed execution of hybrid quantum-classical algorithms, combining AWS classical compute resources based on Amazon EC2 (the 'job instance') with Amazon Braket quantum processing units (QPUs) or quantum circuit simulators. For variational algorithms, QPUs act as co-processors to classical computers, similar to how a GPU is used in training machine learning models. Such hybrid algorithms rely on rapid, iterative computations between classical computers and QPUs or simulators. With the Hybrid Jobs feature, Amazon Braket coordinates the necessary quantum and classical compute resources, executes your algorithm, and releases the resources once the job is completed so you only pay for what you use. The feature also helps your algorithms run quickly and predictably, through on-demand priority access to QPUs on Braket. In order to provide ultra-low latency run times for variational algorithms such as quantum machine learning, Amazon Braket Hybrid Jobs supports high-performance embedded simulators from PennyLane, such as the lightning.qubit and lightning.gpu simulator. These simulators are embedded directly in the same container as your application code, and support advanced features such as the adjoint method to reduce the number of circuit evaluations required when computing gradients. When running hybrid jobs, you can use sample notebooks or write your algorithms using the SDK or PennyLane. You can also use your own custom software stack for your algorithm with a ‘bring-your-own-container’ capability.
There are two components to pricing for the Hybrid Jobs feature: charges for the use of classical resources, and charges for the use of quantum computers or quantum simulators. First, for the use of job instances, you will be charged for the duration the job is running, based on the instance type you use. The Hybrid Jobs feature uses the ml.m5.xlarge instance as the default instance type, or you can choose a different instance type when creating a job. You also have the option to add extra data storage within the job instance at an additional cost. For pricing of these instances and supplementary instance storage, please refer to the "Job Instances" pricing table below. If you are using one of the on-demand simulators (SV1, DM1, TN1) or a quantum computer available on Amazon Braket for a portion of your Hybrid Job, then you will be charged for the execution of quantum tasks that are created as part of your job. Pricing for these tasks is the same whether they are run as part of a Hybrid Job or not. Please refer to the “Quantum Computers” and “Simulators” tabs to learn more about pricing for quantum tasks on the different devices. If you are using an embedded simulator, then you only pay for the classical CPU or GPU resources you use for the duration of the job, based on the pricing table below.
Amazon Braket provides fully managed Jupyter notebooks so you can build your quantum algorithms, share code, test your algorithm, and visualize your results. When you create a notebook in Amazon Braket, it will be hosted and billed by Amazon SageMaker, the AWS machine learning service. With Amazon Braket managed notebooks, you can choose your preferred instance type to run each notebook and you will be billed per hour for usage, according to pricing for the selected instance type. For usage under one hour, cost is prorated based on the hourly price. The cost of using each notebook will be listed in the Amazon SageMaker portion of your bill.
For details on the pricing of different notebook instance types, refer to the Amazon SageMaker pricing page. Select the On-Demand Notebook Instance tab, select the relevant region, and refer to the instance type that you plan to use.
AWS Free Tier
Amazon Braket is part of the AWS Free Tier. The AWS Free Tier gives you one free hour of quantum circuit simulation time per month during the first twelve months of use. This applies to simulation time on SV1, DM1, TN1, or any combination of those three managed quantum circuit simulators, or use of simulators through the Hybrid Jobs feature. Standard rates for incremental usage apply after the free hour has been used. Use of other Amazon Braket resources such as managed notebooks, hybrid job instances, and quantum computers will incur costs.
Example 1: on-demand simulator
Let’s say a developer uses the Amazon Braket SDK running in a local environment to design a quantum circuit to use 30 qubits, and then simulates the circuit using the Amazon Braket on-demand simulator SV1 ($0.075 per minute) in the AWS US East (N. Virginia) region. The simulation takes 4 minutes to run. The developer is running this simulation after using Amazon Braket for 10 months, with no other simulation usage so far that month.
This workload does not use a managed notebook, and the simulation time is covered by the AWS Free Tier, which provides one hour of included on-demand simulation time per month for the first twelve months of using Amazon Braket.
Total charges for first 12 months: $0.
If the developer runs the same simulation after using Amazon Braket for twelve months, standard SV1 pricing ($0.075 per minute) applies:
($0.075 / minute) x 4 minutes =$0.30
Total charges after first 12 months: $0.30
Example 2: Quantum annealer
A researcher runs a quantum annealing problem on the D-Wave Advantage quantum computer in the AWS US West (Oregon) Region. This task includes 2,000 result samples of the same annealing problem. The cost to run this task includes a per-task charge of $0.30, plus 2,000 shots at a per-shot price of $0.00019.
The cost to run this problem:
Task charges: 1 task x ($0.30 / task) = $0.30
Shots charges: 2,000 shots x ($0.00019 / shot) = $0.38
Total charges: $0.68
Example 3: Gate-based QPU
A scientist runs a quantum algorithm on the Rigetti M-2 quantum computer in the AWS US West (N. California) Region. This task includes 10,000 repeated shots of the same circuit design. The cost to run this task includes a per-task charge of $0.30, plus 10,000 shots at a per-shot price of $0.00035.
The cost to run this algorithm:
Task charges: 1 task x $0.30 / task = $0.30
Shots charges: 10,000 shots x $0.00035 / shot = $3.50
Total charges: $3.80
Example 4: Hybrid job with on-demand simulator
A researcher designs a hybrid quantum-classical algorithm and runs it using the Amazon Braket Hybrid Jobs feature and the SV1 simulator ($0.075 per minute) in the AWS US East (N. Virginia) Region. The hybrid job is set up to run for 30 iterations each with 10 tasks. Each task takes 5 seconds to run on the simulator and the total duration of the job is 1500 seconds. The researcher uses an ml.m5.xlarge job instance ($0.23 per hour) for classical compute.
The cost to run this job:
SV1 simulator charges, if the researcher has exceeded the AWS Free Tier: ($0.075 / minute) x (1500/60 minutes) = $1.875
ml.m5.xlarge job instance usage charges: ($0.23 / 60 minutes) x (1500/60 minutes) = $0.096
Total charges: $1.97
Example 5: Hybrid job with a QPU
A scientist runs a quantum machine learning algorithm on the Rigetti quantum computer using Amazon Braket Hybrid Jobs in the AWS US West (N. California) Region. This job includes 50 iterations, each with 2 tasks that have 100 shots. The cost to run this job includes a per-task charge of $0.30, plus each run of the 100 shots at a per-shot price of $0.00035, for each iteration. The scientist also uses an ml.p3.8xlarge job instance ($14.688 per hour) for the entire duration of the job, which lasts 26 minutes.
The cost to run this job:
Task charges: 50 iterations x 2 tasks x $0.30 / task = $30.00
Shots charges: 50 iterations x 2 tasks x 100 shots x $0.00035 / shot = $3.50
ml.p3.8xlarge usage charges: ($14.688 / 60 minutes) x 26 minutes = $6.36
Total charges: $39.86
Example 6: Hybrid job with embedded simulator
A researcher designs a hybrid quantum-classical algorithm on 24 qubits and runs it on the embedded lightning.gpu simulator using Amazon Braket Hybrid Jobs, in the AWS US East (N. Virginia) Region. The hybrid job is set up to run for 10 iterations with zero shots each. The researcher uses a single ml.p3.8xlarge job instance ($0.2448 per-minute) for classical compute. The job takes 2 minutes to complete.
The cost to run this job:
ml.p3.8xlarge job instance usage charges: ($0.2448 per-minute) x (2 minutes) = $0.49
Total charges: $0.49