[Update] All Upfront / Partial Upfront options have been added to Amazon Redshift RA3 instance reservations

[Update] All Upfront / Partial Upfront options have been added to Amazon Redshift RA3 instance reservations

Amazon Redshift reserved instances for RG instances now offer two new payment options: "All Upfront" and "Partial Upfront." This update expands cost optimization choices by allowing you to adjust upfront payment amounts. Here we introduce the characteristics of the three payment methods and how to choose between them.
2026.06.26

This page has been translated by machine translation. View original

This is Ishikawa from the Cloud Business Division. Amazon Redshift Reserved Instances for RG instances have newly added "All Upfront" and "Partial Upfront" payment options. Combined with the existing "No Upfront" option, you can now choose from three types of payment methods.

https://aws.amazon.com/jp/about-aws/whats-new/2026/06/amazon-redshift-ri-upfront-pricing-rg-instances/

This is an update that expands the options for further cost optimization by adjusting upfront payments for those who are continuously running RG instances.

What Are Amazon Redshift RG Instances and Reserved Instances?

Amazon Redshift is a fully managed cloud data warehouse service. It operates in units of "clusters" that bundle compute nodes, and nodes are billed at on-demand or reserved pricing.

RG instances are the latest generation node type based on AWS Graviton, made generally available in May 2026. Compared to the conventional RA3, they are up to 2.2x faster for data warehouse workloads and up to 2.4x faster for data lake (Apache Iceberg) workloads, with a 30% lower price per vCPU. Additionally, they feature an integrated data lake query engine that runs on the cluster's own compute, enabling direct queries against Apache Iceberg / Parquet (see the GA blog post for details). RG instances come in two sizes: rg.xlarge and rg.4xlarge.

https://dev.classmethod.jp/articles/amazon-redshift-rg-instances-graviton/

Reserved Instances (also called "Reserved Nodes" in Amazon Redshift) are a billing model where you commit to 1 or 3 years of usage in advance, receiving significant discounts compared to on-demand pricing. They are suitable for stable, continuously running production workloads. Note that reservations are purely a billing concept — purchasing a reservation does not automatically create nodes.

Update Details

With this update, the following payment options are now available for RG instance Reserved Instances.

The main changes are as follows:

  • All Upfront: A new option to pay the full amount for the reservation period upfront at the start. This offers the highest discount rate of the three.
  • Partial Upfront: A new option to pay a portion upfront and the remainder monthly. This balances discount and initial outlay.
  • These are offered in addition to the existing No Upfront option.
  • All options support 1-year and 3-year reservation terms.

This allows RG instances, like existing node types such as RA3, to choose from three payment options based on financial circumstances.

Overview of the Three Payment Options

The characteristics of each payment option are as follows. The same logic applies as with Amazon EC2 or RA3 reservations: the larger the upfront payment, and the longer the reservation term (3 years), the higher the discount rate.

Payment Option Upfront Monthly Discount Tendency
No Upfront None Yes Lowest
Partial Upfront Partial Yes Middle
All Upfront Full None Highest

The following diagram illustrates how to think about choosing a payment option.

Sample Pricing for the Tokyo Region (1-Year Term)

The following are the RG instance Reserved Instance prices for the Tokyo Region (ap-northeast-1) (1-year term, compared to on-demand). The discount rate for a 1-year term is approximately 30–34%. Choosing a 3-year term applies an even larger discount.

No Upfront

Instance Upfront Monthly Effective Hourly Rate Effective Annual Price per TB vs. On-Demand On-Demand Rate
rg.4xlarge $0.00 $1,829.52 $2.506 $171.52 30% $3.5803
rg.xlarge $0.00 $457.14 $0.626 $171.43 30% $0.8946

Partial Upfront

Instance Upfront Monthly Effective Hourly Rate Effective Annual Price per TB vs. On-Demand On-Demand Rate
rg.4xlarge $10,506.53 $875.55 $2.399 $164.16 33% $3.5803
rg.xlarge $2,625.35 $218.78 $0.599 $164.09 33% $0.8946

All Upfront

Instance Upfront Monthly Effective Hourly Rate Effective Annual Price per TB vs. On-Demand On-Demand Rate
rg.4xlarge $20,699.47 $0.00 $2.363 $161.71 34% $3.5803
rg.xlarge $5,172.30 $0.00 $0.590 $161.63 34% $0.8946

* Monthly fee Effective hourly rate = (Upfront + Monthly × 12) ÷ 8,760 hours Effective annual price per TB including managed storage. Definitions of each metric conform to the Amazon Redshift pricing page. Prices are subject to change, so always verify the latest values on the pricing page before purchasing.

Supported Regions

The All Upfront / Partial Upfront payment options are available in 28 AWS Regions, including Tokyo and Osaka. They cover major areas including the United States, Canada, South America, Europe, Africa, Asia Pacific, and Mexico.

Impact on Pricing

There are no additional charges from this update itself. It simply adds more payment method choices to existing Reserved Instances. Specific discount rates and pricing vary by region, node type, and term, so check the Amazon Redshift pricing page for the latest figures.

Note that Reserved Nodes are billed throughout the reservation period regardless of whether they are in use. Upfront and monthly payments occur even if the target nodes are not actually running.

Thinking About the Break-Even Point

Reserved Instances are a fixed-cost billing model where charges occur regardless of usage once purchased. On-demand, on the other hand, is pay-as-you-go billing only for the hours used. Therefore, the two key factors in the adoption decision are how much you will run the instance per year (utilization rate) and how much upfront payment you can tolerate.

Break-Even Point vs. On-Demand (Utilization Rate)

Using the annual on-demand cost for continuous operation (24 hours, 365 days) as a baseline, the break-even utilization rate at which reservations become advantageous is calculated as "1 − discount rate." For the Tokyo Region, 1-year term, rg.4xlarge, the figures are as follows (rg.xlarge has the same discount rate, so the break-even utilization rate is the same).

Payment Option Annual Reserved Cost Break-Even Utilization Rate Estimated Annual Hours
No Upfront $21,954.24 Approx. 70% Approx. 6,132 hours (approx. 8.4 months)
Partial Upfront $21,013.13 Approx. 67% Approx. 5,869 hours
All Upfront $20,699.47 Approx. 66% Approx. 5,782 hours

※ Annual on-demand cost (rg.4xlarge) = $3.5803 × 8,760 hours = approx. $31,363

If your annual utilization rate exceeds these values, Reserved Instances will be cheaper. For production clusters running continuously, reservations are definitively advantageous, and among them, All Upfront is the least expensive. Conversely, if you frequently use pause / resume and your utilization rate falls below 66–70%, on-demand or Amazon Redshift Serverless may be cheaper.

Break-Even Point Between Payment Options (Recovering the Upfront Cost)

The larger the upfront payment, the lower the total cost, but it requires an initial cash outlay. Comparing the annual total cost for 1-year, continuous operation (rg.4xlarge) is as follows.

Payment Option Annual Total Difference vs. No Upfront
No Upfront $21,954.24 Baseline
Partial Upfront $21,013.13 −$941 (approx. −4.3%)
All Upfront $20,699.47 −$1,255 (approx. −5.7%)

Choosing All Upfront eliminates the monthly payment of $1,829.52 (rg.4xlarge) that would otherwise be paid each month under No Upfront. The time to recover the upfront payment of $20,699.47 through this monthly savings is approximately 11.3 months, meaning recovery completes near the end of the 1-year contract. With a 3-year term, the discount rate is even larger, so the upfront cost is recovered more quickly and the total savings benefit expands further.

Which Payment Option Should You Choose?

The payment option is chosen based on the trade-off between "discount rate" and "cash flow (initial burden)."

  • To maximize cost savings: All Upfront with full upfront payment offers the highest discount rate. It is suited for production workloads with confirmed long-term stable operation.
  • To reduce initial burden while still getting a discount: Partial Upfront is an intermediate option. A fixed upfront payment reduces the monthly fee.
  • To avoid upfront payment or smooth out the budget: No Upfront allows monthly-only payments. The discount rate is modest, but the impact on cash flow is minimized.

For RG instances with stable operation, 3-year All Upfront offers the highest discount. On the other hand, if usage is uncertain or you prioritize flexibility, it is worth considering other options such as Amazon Redshift Serverless.

Closing

All Upfront and Partial Upfront payment options have been added to Amazon Redshift Reserved Instances for RG instances. Combined with the existing No Upfront option, you can now choose from three options, expanding the range of cost optimization to suit your financial situation. Available in 28 regions including Tokyo and Osaka. In the Tokyo Region, a 1-year term offers approximately 30–34% discount, making reservations advantageous for steady workloads with an annual utilization rate exceeding approximately 66–70%.

If you are continuously running RG instances, consider switching to the optimal payment option by weighing your company's cash flow against the discount rate, as well as the break-even utilization rate.


コスト最適化、打ちっぱなしで元通りになっていませんか

タグ付けも不要リソースの棚卸しも、施策は打てる。でも続ける仕組みがなければ、コストは数か月でじわじわ戻る。一度きりで終わらせず、FinOpsを組織に定着させる=CCoEの役割。最適化を回し続ける進め方を、無料資料にまとめました。

CCoE総合支援

FinOpsを定着させる資料をもらう

Share this article

AWSのお困り事はクラスメソッドへ