AWS Updates 2026 | April 30-January 2 | Cloud Provider News
April 30, 2026
One-click CSV exports from the Cost Optimization Hub for quick offline analysis
The new option exports the hub’s cost-optimization recommendations, respecting the console’s filters and grouping, into a CSV file you can share or analyze offline.
Plus, this complements existing automated S3 exports and makes it simpler to circulate prioritized recommendations to stakeholders or load them into spreadsheets and reporting tools.
Also, that speeds collaboration between FinOps and engineering teams when you need to review or act on optimization suggestions quickly.
Compute Optimizer now understands the newest EC2 and RDS types so recommendations match current price/perf
This expands the optimizer’s visibility into recently launched families and DB classes, letting it recommend migration targets or downsizes using the current-generation price/performance landscape.
Meanwhile, FinOps teams can use those updated recommendations to identify modernization opportunities and potential savings when planning instance refreshes.
Also, that helps reduce surprise costs from blind spots where older tooling didn’t account for new, cheaper or higher-performing instance options.
Redshift Serverless turns on AI-driven scaling by default for new workgroups
The default uses ML to predict compute needs and automatically adjust resources for Serverless workgroups, improving price/performance without manual tuning.
Plus, new workgroups benefit immediately from automated right-sizing, which can reduce over-provisioned capacity and lower cost for unpredictable workloads.
Additionally, supporting scaling down to 8 RPU gives FinOps teams a smaller minimum footprint to tune against when estimating and controlling spend.
C8gn (Graviton4, network-optimized) expanded to more regions for net-new cost/perf picks
These instances are tuned for ultra-high networking and are suitable for network appliances and inference workloads where throughput matters.
Plus, broader regional availability gives teams more options for regional price/perf tradeoffs and capacity planning.
Additionally, this can help lower costs by using Graviton4 where architecture changes are viable.
AWS has streamlined its EC2 portfolio with 6th-gen Intel Xeon instances
Massive Throughput (R8in/ib & M8in/ib): New memory-optimized and general-purpose instances offering up to 600 Gbps network and 300 Gbps EBS bandwidth to eliminate I/O bottlenecks.
Performance Gains (C8ine & M8ine): Now generally available, providing up to 43% better performance than prior generations, ideal for reducing instance count and TCO.
Wider Availability (C8i-flex): Regional expansion to Ireland, London, and New Zealand, offering more flexible high-performance compute options globally.
April 23, 2026
Deliver cost reports cross-account with AWS Data Exports
You can now export Cost and Usage Report (CUR 2.0), FOCUS, Cost Optimization Recommendations, and Carbon Emissions reports directly to S3 buckets in other AWS accounts, eliminating duplicate storage.
Additionally, centralizing those reports into a single account removes replication costs and simplifies centralized analytics for FinOps teams running cross-account reporting.
Hide managed resources from EC2 inventory views to reduce noise
This setting lets you exclude managed resources (EKS, ECS, Lambda, WorkSpaces, etc.) from default console lists and describe API outputs so your inventory focuses on self-managed, billable resources.
Also, reducing inventory noise makes cost reports and governance checks more accurate and lowers the effort to reconcile billable resources for FinOps teams.
Aurora Serverless Platform v4: better performance and zero scale
Platform v4 provides up to 30% improvement in performance while retaining zero-capacity scaling, so bursty and intermittent workloads can be faster and cost less when idle.
Also, the smarter scaling behavior reduces billed capacity during idle periods, giving FinOps teams a clearer lever to lower database costs for variable workloads.
SageMaker HyperPod adds flexible instance groups for capacity resilience
Flexible groups let you specify an ordered list of instance types and multiple subnets so jobs can fall back to alternate shapes instead of failing when capacity is limited.
Furthermore, ordered fallback reduces the need to manage many separate instance groups and cuts operational overhead and cost from manual capacity juggling.
S3 Express One Zone now supports S3 Inventory for large-scale audits
With S3 Inventory support you can schedule object listings and metadata exports (CSV/ORC/Parquet) for large buckets instead of relying on synchronous List API calls that can be expensive at scale.
Also, having scheduled inventory helps validation workflows and audits run more efficiently and reduces runtime costs tied to heavy, on-demand listing operations.
AWS Marketplace streamlines VAT handling for EMEA deemed supply
Marketplace sellers in the covered jurisdictions can now submit VAT invoices via a self-service flow and rely on automated VAT disbursements, reducing manual tax administration.
Additionally, automating VAT workflows lowers administrative overhead and compliance cost for sellers operating in those EMEA markets.
April 16, 2026
Amazon Bedrock now supports cost allocation by IAM user and role for precise chargeback
This lets you use IAM tags and cost allocation tags to show who’s driving Bedrock spend, improving chargeback and FinOps accountability.
Consequently, teams can produce clearer showback/chargeback reports and enforce budgets or guardrails on model usage.
EC2 Capacity Manager adds tag-based dimensions so capacity maps to cost centers
That means capacity metrics and exports can align with cost-center or team tags, improving capacity-to-cost mapping accuracy.
So FinOps teams can more reliably correlate capacity needs with budget owners and make reservation or rightsizing decisions tied to organizational tagging.
OpenSearch Serverless adds Zstandard (zstd) index compression to shrink index size
Choosing zstd can lower storage costs for index-heavy workloads while you balance indexing throughput and compression level.
As a result, FinOps teams get a concrete knob to reduce managed-storage spend on search and analytics indices.
Top‑K optimization in Amazon Redshift reduces scanned data and query cost
The optimization is automatic (no query rewrite needed) and speeds common top-K patterns such as most-recent queries.
Consequently, query cost and latency drop for many analytics workloads that rely on top-K results.
Unified observability in OpenSearch Service with Managed Prometheus & agent tracing
Consolidating telemetry into a single workspace cuts tool sprawl and reduces duplication and storage overhead.
Therefore, SRE and FinOps teams can lower observability costs while keeping integrated metric/log/trace workflows.
AWS Interconnect, multicloud reaches GA with single-fee pricing for smarter egress planning
This new pricing model is based on bandwidth and scope and can materially change multicloud egress and interconnect cost structures.
Consequently, customers should evaluate the new single-fee approach when redesigning multicloud data flows to reduce repetitive egress charges.
Billing & Cost Management dashboards: scheduled email delivery for automatic reports
Automated distribution reduces manual report generation and helps finance and engineering stakeholders get timely cost insights.
As a result, teams can improve cost communication workflows and reduce ad‑hoc reporting overhead.
S3 Lifecycle now pauses expiration/transition for objects with failed replication
Once replication is fixed and completes, lifecycle actions proceed as normal, protecting against accidental data loss.
Therefore, FinOps and data-protection teams can avoid unexpected recovery costs caused by premature deletions.
CloudWatch Pipelines: compliance and governance controls (keep original)
Keeping originals improves auditability but increases storage, so retention policies need review to balance governance and storage spend.
Consequently, FinOps teams should evaluate retention and lifecycle for raw logs when enabling keep original to avoid unexpected cost increases.
EC2 P6‑B300 instances for huge‑model training with more GPU memory and networking
These instances deliver higher GPU memory and networking to shorten training time and potentially improve price/performance for huge-model workloads.
Therefore, ML teams should benchmark P6‑B300 against existing P6 variants to understand cost-efficiency tradeoffs for specific models.
Increased EBS-optimized performance for C8gn, M8gn, R8gn 48xlarge and metal sizes at no extra cost
This boosts I/O-bound workload performance and can reduce runtime without higher instance fees, potentially lowering overall TCO.
Consequently, customers can stop/start running instances to pick up the upgrade where those sizes are offered and realize performance gains.
April 9, 2026
Cost Explorer gets natural-language queries powered by Amazon Q, faster answers for cost teams
This adds Amazon Q–driven natural-language query capability to Cost Explorer so teams can ask questions in plain English and get visualized answers without building complex queries.
Plus, the visualizations update automatically from those queries, which speeds up cost analysis and makes cost conversations easier for non-experts.
Amazon S3 Files (GA) exposes S3 as a high-performance file system, cut duplication and egress
S3 Files removes the need to duplicate data between object and file storage by letting teams access buckets as a file system; it also caches hot data to reduce latency.
On top of that, avoiding separate file systems can reduce storage and egress costs tied to maintaining duplicate datasets.
OpenSearch Service adds Graviton4‑based i8ge storage instances, better price/perf for search clusters
Amazon OpenSearch Service now supports Graviton4‑based i8ge storage-optimized instances.
i8ge instances deliver significantly better compute and NVMe storage performance versus prior generations for storage-intensive search and analytics workloads.
Plus, improved price-performance can reduce cost-per-query and lower cluster latency for large deployments.
EKS managed node groups can use EC2 Auto Scaling warm pools, trade cost for faster scale
Warm pools can be kept Stopped for lower cost or Running for faster availability, reducing cold-start provisioning delays.
Also, this helps avoid expensive cold-starts during bursts while giving you levers to tune the cost vs. availability trade-off.
Lambda response streaming expands to all commercial Regions, faster responses, watch streaming charges
Response streaming can reduce time-to-first-byte for latency-sensitive apps by progressively sending outputs from functions.
However, AWS notes streaming can incur additional network transfer charges based on bytes streamed over thresholds.
Partner Revenue Measurement adds User Agent string support, better attribution for partner-driven consumption
Partners can use the User Agent string to quantify service consumption driven by their solutions and attribute usage more accurately.
Also, this complements tagging and metering strategies to help track API-driven workload consumption.
Partner Revenue Measurement integrates with Marketplace Metering, auto-attribute AMI/ML downstream usage
This integration automatically attributes downstream compute usage to partner solutions without extra partner implementation.
Plus, it improves visibility into product-driven cloud consumption for both revenue and usage attribution.
April 2, 2026
AWS launches a Sustainability console to make emissions tracking exportable and API-ready
AWS released a standalone Sustainability console with regional and service-level carbon estimates.
The console expands on the Customer Carbon Footprint Tool and provides market- and location-based carbon estimates by AWS Region, service, and emissions scope, plus CSV exports and API/SDK access.
Additionally, you can pull emissions data without needing billing permissions, which makes it easier to integrate into reporting workflows and internal chargebacks.
CloudWatch centralization discovers and selects logs by data source name/type to simplify cost-aware aggregation
CloudWatch Logs centralization now supports selection by discovered data source name and type.
You can centralize logs across accounts and Regions by data source (for example, VPC Flow Logs or CloudTrail) instead of maintaining long lists of log groups.
Athena adds Capacity Reservations in more Regions so you can isolate query capacity and plan spend
Amazon Athena expanded Capacity Reservations to additional Regions.
Capacity Reservations give dedicated serverless capacity to isolate important workloads and control concurrency for mission-critical queries.
S3 Vectors expands to 17 more Regions to help localize vector storage and manage egress/latency costs
Amazon S3 Vectors expanded to 17 additional AWS Regions.
S3 Vectors is a cost-optimized native vector object storage for AI workloads (RAG/embeddings), and broader regional availability reduces cross-region egress and latency.
ECS Managed Instances now support EC2 instance store volumes so you can avoid EBS costs for local data
Amazon ECS Managed Instances added support for EC2 instance store as data volumes.
You can use local NVMe instance store volumes for container data on ECS Managed Instances instead of provisioning EBS, improving I/O performance for latency-sensitive workloads.
CloudWatch Logs centralization and selection improvements cut the need for long log lists and custom mapping
CloudWatch Logs centralization enhancements let you select logs by discovered data source name/type.
This enables centralized security and ops pipelines without maintaining long lists of log groups, improving consistency and reducing manual maintenance.
Consequently, it’s simpler to enforce centralized pipelines only for the sources you need and keep others local to control costs.
Also, this reduces fragile automation and the overhead of constant list updates.
AWS Marketplace sellers can self-serve refunds and cancel agreements to simplify billing reconciliation
Sellers can issue refunds and cancel agreements with automated visibility on buyer charge summaries, cutting down support tickets and manual billing adjustments.
That streamlines reconciliation and can reduce operational billing overhead for both buyers and sellers.
Plus, faster refund flows make chargeback and invoice cleanup less painful for FinOps teams.
AWS Organizations now returns full account/OU paths in one API response to simplify automation
AWS Organizations APIs now include the complete account/OU path in single responses.
API responses now contain the full hierarchy like o-{org}/r-{root}/ou-…/account, eliminating extra calls to reconstruct paths.
EC2 new instance family availability gives more price-performance choices (R8gd, I8ge, M8a)
AWS expanded availability of EC2 families: R8gd, I8ge, and M8a in additional Regions,
R8gd (Graviton4 with local NVMe), I8ge (storage-optimized Graviton4), and M8a (5th Gen AMD EPYC) are available in more Regions, broadening choices for I/O-heavy and general-purpose workloads.
March 27, 2026
Extended support cost projection adds tagging and tag-aggregation views for RDS, EKS, OpenSearch, and ElastiCache
This update adds explicit support for tag-based analysis across RDS, EKS, OpenSearch, and ElastiCache, plus new tag-aggregation views to roll up costs by tagging taxonomy.
Also, that means you can filter and group cost projections directly by tags you already use for chargeback and allocation.
CUDOS Dashboard v5.7.3 improves billing grouping and tagging for more accurate FinOps reporting
This release surfaces product/service code and service category in the dashboard, adjusts billing grouping to place discounts first, and adds tag-based filters and fixes for charge types.
Those changes mean billing summaries reflect discounts more accurately and let you slice cost data by tags for faster analysis.
EC2 Fleet can now consume interruptible Capacity Reservations to use spare on-demand capacity
You can now request shared, temporary Capacity Reservations that are interruptible and use them across launch templates in one Fleet operation.
Also, this enables accounts to use spare On‑Demand reservation capacity when it’s available, improving utilization of excess reservation investments.
AWS Batch adds quota shares and job preemption for SageMaker training to improve GPU utilization
AWS Batch introduced quota shares and job preemption for SageMaker training jobs.
This feature lets you partition GPU capacity with quota shares and automatically preempt lower-priority jobs to free resources for higher-priority experiments.
Additionally, that gives you a straightforward way to prioritize critical workloads and reduce wasted GPU time.
Amazon Aurora PostgreSQL serverless added to the AWS Free Tier to lower trial/onboarding costs
The Free Tier listing provides free credits and the ability to create and test Aurora serverless clusters without immediate charge.
Also, that makes it cheaper to evaluate managed PostgreSQL for dev/test and early prototypes.
EC2 C8gn (Graviton4) instances expanded to more regions for better price-performance options
Broader regional availability means more places to choose Arm‑based instances for network-intensive workloads.
Additionally, that gives FinOps teams more opportunity to optimize price-performance by moving suitable workloads to Graviton4.
EC2 I7ie storage-optimized instances now in more regions with higher I/O and compute gains
Those improvements can boost price-performance for large I/O workloads that need faster storage throughput.
Also, wider availability gives teams flexibility to pick regions where these instance benefits reduce overall cost.
March 19, 2026
AWS CDK Mixins are GA — enforce cost and compliance patterns in your IaC with .with()
Mixins let teams attach behaviors such as auto-delete, encryption, and lifecycle policies to L1/L2 or custom constructs using a composable API, so guardrails live in code instead of manual configs.
Moreover, that makes enforcement of cost‑control patterns part of the deployment pipeline, reducing drift and accumulating FinOps debt.
SageMaker Training Plans can extend reserved GPU capacity without reconfiguring workloads
This feature helps teams keep long training workflows running under existing commitments, improving predictability and avoiding disruption.
Additionally, it supports better FinOps capacity planning because you can lengthen commitment windows to match training timelines rather than redeploy or rebook capacity.
SageMaker HyperPod idle resource sharing boosts utilization across teams
The capability increases GPU utilization by letting workloads use idle capacity, which lowers the effective cost of shared generative AI infrastructure.
Moreover, borrowing is controlled by admin limits, so platform teams keep governance while improving utilization for tenant workloads.
OpenSearch 3.5 on Amazon OpenSearch Service trims LLM token costs with context management
Those capabilities reduce the size of inputs sent to models — directly lowering token‑based inference costs — and include improved search relevance tools and observability.
Moreover, if you’re running retrieval‑augmented generation or chat interfaces, the automatic summarization and truncation can cut backend model spend.
Amazon Redshift speeds up first‑run queries (up to 7x) with no extra cost
The optimization improves dashboard and ETL responsiveness with no configuration or extra charge.
Additionally, faster first‑run queries can let teams consider cluster rightsizing because workloads that were previously provisioned for latency might now run acceptably on smaller configurations.
CloudWatch can auto‑enable EC2 detailed monitoring across your AWS Organization
This helps standardize finer‑grained metrics collection for consistent autoscaling and observability.
However, additionally, teams should note this increases monitoring billable metrics, so FinOps needs to account for the added monitoring costs when rolling this out.
CloudWatch Application Signals adds SLO recommendations and calendar‑aligned reporting
The new features give data‑driven suggestions for realistic SLO thresholds and produce reports aligned to calendars for easier stakeholder review.
Moreover, this helps FinOps and reliability teams agree on targets that balance cost and reliability, reducing alert fatigue and unnecessary operational expense.
New EC2 M6in and M6idn instances expanded, better network and EBS bandwidth for network‑heavy workloads
Those instances are suited to network‑heavy workloads and give FinOps teams more options to right‑size based on network and storage throughput needs.
Additionally, because they’re available across purchase options including Savings Plans and Spot, you can mix purchase models to lower unit costs.
March 12, 2026
Database Savings Plans now cover Amazon OpenSearch Service and Amazon Neptune Analytics, up to ~35% savings
With this change, you can commit to a consistent $/hour for a one-year term (no upfront) and receive up to ~35% savings on eligible usage across serverless and provisioned modes.
Also, the plan automatically applies across supported engines, instance families, sizes, deployment options, and regions (except China), so you can change instance types or scale without losing the discount.
Amazon Bedrock: TimeToFirstToken and EstimatedTPMQuotaUsage metrics, inferencing SLA and quota visibility
Amazon Bedrock now emits TimeToFirstToken and EstimatedTPMQuotaUsage CloudWatch metrics.
These metrics track inference latency (TTFT) and estimated tokens-per-minute quota consumption so you can monitor SLA impact and anticipate quota throttling.
Additionally, using these metrics with alarms helps you avoid unexpected throttling or quota-related slowdowns that could drive unplanned retries and cost spikes.
Graviton4 instance expansion (C8gd, M8gd), better price-performance options
AWS expanded availability of Graviton4-powered C8gd and M8gd instances.
Graviton4 instances can deliver roughly ~30% better performance versus Graviton3 and provide flexible bandwidth weighting, offering stronger price-performance for compatible workloads.
Moreover, these instances are available across On‑Demand, Savings Plans, and Spot, so you can choose purchase options that fit cost and availability needs.
Amazon CloudWatch Logs: higher Logs Insights concurrency and API limits, more parallel analysis
You can now run many more simultaneous Logs Insights queries and dashboards, improving analysis throughput for operations and SRE teams.
However, greater concurrency can increase query-related spend if usage isn’t governed, so FinOps teams should pair this with query cost controls and governance.
OpenSearch Service: capacity‑optimized blue/green deployments, smaller spare capacity during upgrades
When full spare capacity isn’t available, the service can use incremental batch updates instead of reserving large extra capacity, which lowers the extra instances required during domain updates.
Also, this reduces both the cost and timing friction of upgrades for large clusters, making maintenance less expensive and disruptive.
OpenSearch Service removes 3 TiB in-place increase limit, scale storage without blue/green for many cases
This lets you scale volumes in place past the old 3 TiB limit for many domains, reducing the need for disruptive blue/green migrations.
Additionally, avoiding full data migrations lowers operational overhead and migration costs while speeding capacity scaling.
March 5, 2026
Plan for VPC Encryption Controls billing — per-VPC hourly charges are coming
AWS announced that VPC Encryption Controls is moving from free preview to paid pricing.
The change means accounts with Encryption Controls enabled will incur an hourly fixed rate per non-empty VPC with Encryption Controls turned on, and there are special considerations for Transit Gateway attachments.
This is a direct new line item you’ll see on bills, so finance and FinOps teams need to fold it into regional cost forecasts and tagging/chargeback models.
AgentCore Policy is generally available, stop runaway agents from burning compute
The feature centralizes control of agent-tool interactions and input validation outside agent code, converting human-readable policies into Cedar under the hood to enforce rules consistently.
Bedrock Projects API (Mantle) adds per-project isolation, IAM and tagging for cleaner chargeback
Amazon Bedrock’s Projects API (Mantle) provides per-project isolation with IAM controls and tagging.
That means teams can assign tags and isolate usage per project, improving access control and making it easier to segregate billing for FinOps reporting and chargeback.
AWS Marketplace supports concurrent agreements, simplify multi-team SaaS procurement
This lets different teams manage expansions or negotiated terms independently without needing separate accounts for the same product.
What’s more, that procurement flexibility has direct FinOps implications for managing commitments, discounts, and internal chargeback across teams.
ECS Managed Instances can use EC2 Capacity Reservations for predictable capacity
You can choose reservation preferences such as reservations-only or reservations-first to guarantee capacity for mission-critical workloads.
Centralize CloudWatch logs with custom destination group names for easier cost attribution
That makes it simpler to organize and attribute centralized logs for retention policies and cost allocation across accounts.
OpenSearch Cluster Insights adds overload and sharding recommendations to avoid wasted resources
These insights identify resource pressure and shard imbalances and provide actionable advice on scaling or re-sharding to avoid throttling and inefficient resource spend.
Following those recommendations helps prevent costly performance degradation and overprovisioning.
AWS Config expands to cover 30 new resource types for better governance
Broader coverage improves automated inventory and governance, making it easier to track and enforce rules on resources that drive cost and compliance.
EC2 M8i and M8i-flex expand region availability — more instance choices for price-performance
AWS expanded availability of M8i and M8i-flex instances to additional regions.
These instances (custom Intel Xeon 6) give options for improved price-performance and memory bandwidth for general-purpose workloads.
Also, having more regions with these instance types helps FinOps teams pick the best regional price-performance trade-offs.
EC2 I8g.metal-48xl GA — Graviton4 storage-optimized instance for I/O-heavy workloads
They offer improved compute and real-time local NVMe storage performance for I/O-intensive databases and analytics workloads.
These instances may change the price-performance calculus for workloads that benefit from storage locality.
February 27, 2026
Save up to ~45% on analytics with 3‑year Redshift Serverless reservations
AWS added 3‑year Serverless Reservations for Amazon Redshift Serverless.
Amazon Redshift Serverless now offers 3‑year Serverless Reservations that let you commit RPUs for a three‑year term. This gives teams predictable savings (up to 45%) and steadier budgeting for analytics workloads.
And because reservations are shareable at the payer account level, you can allocate committed capacity across multiple accounts to get better utilization and lower overall analytics compute spend.
Trusted Advisor’s improved NAT Gateway check spots true cost savings faster
AWS Trusted Advisor improved its unused NAT Gateway detection.
Trusted Advisor now uses Compute Optimizer signals plus a 32‑day lookback to reduce false positives when flagging idle NAT Gateways. That means fewer noisy alerts and clearer, more reliable cleanup targets.
Also, Trusted Advisor surfaces estimated monthly cost savings with findings, so FinOps teams can prioritize which idle NAT Gateways will yield the biggest impact.
S3 access logs now show request source region to cut cross‑region costs
Amazon S3 server access logs now include the source AWS Region.
S3 server access logs will now record the AWS Region where requests originated, helping you detect cross‑region access patterns that drive egress and latency costs.
And since this is available in all Regions at no additional cost, you can start using the logs to inform data placement and reduce unnecessary cross‑region traffic.
Visualize cost and resource data faster with Amazon Q Developer artifacts
Amazon Q Developer added generative‑AI artifacts that automatically produce tables and charts from your resource inventory and billing data inside the AWS Console. That helps teams spot trends and anomalies without manual charting.
Plus, having these visuals in‑console speeds cost reviews and makes it easier to explain billing patterns to stakeholders.
Compute Optimizer tags snapshots it creates for better governance
AWS Compute Optimizer now applies tags to EBS snapshots created during automation.
When Compute Optimizer snapshots and deletes unattached EBS volumes, it automatically adds the tag aws:compute-optimizer:automation-event-id to the snapshots it creates. This improves traceability for automation‑created snapshots.
And that tagging makes it much easier to audit, filter, and potentially reclaim snapshot costs tied to automated optimization runs across supported Regions.
Storage‑heavy search gets faster and more cost‑effective with i7i instances
Amazon OpenSearch Service added support for Intel i7i storage‑optimized instances.
OpenSearch now supports i7i storage‑optimized instances that deliver higher I/O performance and lower latency than prior generations, which helps big‑index workloads run faster.
Consequently, you can get better throughput and potentially lower operational costs for large indexes and heavy query volumes.
Deadline Cloud task chunking reduces overhead and render costs
AWS Deadline Cloud added task chunking to group short tasks into single execution chunks.
Deadline Cloud can now combine multiple short tasks into a single execution chunk, lowering startup overhead for high‑throughput render jobs and cutting overall runtimes.
And because it reduces wasted startup time, chunking directly reduces compute minutes and the cost of large batch render workloads; it’s available in all Regions where Deadline Cloud is supported.
February 19, 2026
OpenSearch Serverless Collection Groups let collections share OCUs while keeping encryption boundaries
See how Collection Groups let multiple collections share OCUs and set minimum allocations.
Amazon OpenSearch Serverless added Collection Groups so multiple collections can share OCUs while retaining per-collection encryption keys and having minimum OCU guarantees.
And this shared-compute model enables cost optimization across multi-tenant or mixed workloads while avoiding cold-start variability.
Amazon Bedrock Project Mantle and six open-weight models — serverless inference improvements and cost wins
Read about Bedrock adding six fully-managed open-weight models and Project Mantle.
Amazon Bedrock added six fully-managed open-weight models and introduced Project Mantle, a distributed inference engine that improves serverless inference performance, quota management, and capacity pooling.
And these changes can lower operational inference costs and simplify capacity planning for model deployments.
AWS Backup: single-action cross-Region snapshot copy into logically air-gapped vaults
Read about single-action cross-Region snapshot copies to logically air-gapped vaults for databases.
AWS Backup now supports single-action cross-Region copies of database snapshots directly into logically air-gapped vaults for Aurora, Neptune, and DocumentDB.
And this removes intermediate steps, reducing complexity and cost for ransomware-resilient backups and improving RPO timelines.
CloudWatch Alarm Mute Rules to cut alert fatigue and unnecessary on-call costs
See how CloudWatch Alarm Mute Rules let you silence alarms temporarily or on a schedule.
Amazon CloudWatch introduced Alarm Mute Rules so teams can temporarily silence alarm notifications (one-time or recurring) while preserving alarm state and automatic reactivation after expiration.
And that reduces operational noise and the overhead or cost of unnecessary on-call responses during planned windows.
AWS Batch adds job queue snapshots and fair-share utilization metrics
Check the new AWS Batch job queue and fair-share utilization visibility features.
AWS Batch added job queue snapshots and fair-share utilization metrics to show capacity consumption by queues and allocations.
And that visibility helps teams optimize scheduling, balance resource distribution, and make cost-aware decisions for batch workloads.
OpenSearch Service: Graviton4 instance support and i7i storage-optimized instances
Read about OpenSearch Service adding Graviton4-based families and i7i storage-optimized support.
Amazon OpenSearch Service expanded support for Graviton4-based instance families (c8g, m8g, r8g/r8gd) delivering up to ~30% better compute versus Graviton3 to improve price-performance.
And additionally it added storage-optimized i7i instances powered by 5th-gen Intel Xeon, offering up to ~23% better compute and materially better SSD I/O performance versus prior generation.
EC2 instance updates: new M8azn, HPC Hpc8a, expanded C8gn, and U7i high-memory availability
See the new and expanded EC2 families for different workload needs and regions.
AWS announced several EC2 updates: new M8azn general purpose instances (5th-gen AMD EPYC, up to 5GHz, up to 2x compute vs prior M5zn), new HPC Hpc8a instances (5th-gen AMD EPYC “Turin” with up to ~40% higher performance and ~25% better price-performance), expanded availability of C8gn (Graviton4) instances to many regions (up to ~30% better compute vs Graviton3), and broader availability of U7i high-memory instances (6–16 TiB variants on custom 4th-gen Intel Xeon).
And these choices give teams more avenues to optimize price/performance for latency-sensitive, HPC, CPU-bound, and large in-memory workloads.
February 13, 2026
AWS Network Firewall price reductions reduce multi‑VPC and TLS inspection costs
AWS announced price reductions for Network Firewall, extending NAT Gateway hourly and data processing discounts to service‑chained secondary endpoints and removing additional data processing charges for Advanced Inspection (TLS inspection). AWS lowered costs for Network Firewall in architectures that chain services across VPCs and use TLS inspection by applying existing discounts to secondary endpoints and removing extra Advanced Inspection charges.
Also, the changes specifically extend NAT Gateway discounts and eliminate additional data processing fees for TLS inspection.
Amazon Athena adds 1‑minute capacity reservations to save on bursty queries
Athena introduced 1‑minute Capacity Reservations and lowered the minimum reservation to 4 DPUs, enabling fine‑grained, short‑duration capacity control and cost savings up to 95% for short workloads. Athena’s 1‑minute reservations let you reserve the exact capacity for short, bursty query runs and set a lower minimum of 4 DPUs for reservations.
Plus, AWS highlights potential cost savings up to 95% for short workloads when you match reservations tightly to query bursts.
EC2 capacity blocks for ML can be shared across accounts to improve utilization
AWS announced GA for cross-account sharing of EC2 Capacity Blocks for ML using AWS RAM, enabling organizations to pool reserved GPU capacity across accounts. The GA release enables sharing EC2 Capacity Blocks for ML across accounts through AWS Resource Access Manager (AWS RAM), so reserved GPU capacity can be pooled.
Additionally, pooling helps improve utilization of reserved GPU capacity and reduces per-account underutilization.
OpenSearch Serverless Collection Groups to share OCUs and reduce compute waste
Amazon OpenSearch Serverless introduced Collection Groups to share OCUs across collections while allowing distinct KMS keys, and to set minimum OCU allocations for predictable performance. Collection Groups let you pool OCU capacity across multiple collections to improve utilization and lower compute cost, along with the ability to enforce a minimum OCU for steady performance.
Also, the feature reduces cold‑start overhead by keeping a baseline allocation and supports distinct encryption keys per collection.
AWS Config adds 30 new resource types, including a Cost Category resource for governance
AWS Config expanded coverage with 30 new resource types, including AWS::CE::CostCategory, to improve governance coverage. The expansion brings broader resource tracking into AWS Config and explicitly includes a CostCategory resource type to help manage cost-related constructs.
Additionally, more resource coverage helps teams enforce policies and detect drift across a wider set of services.
Amazon Redshift can allocate extra compute for automatic optimizations with observability
Redshift added the ability to allocate extra compute for autonomics (ATO, ATS, Auto Vacuum, Auto Analyze) and exposes SYS_AUTOMATIC_OPTIMIZATION observability while providing cost-control limits. You can now let Redshift autonomic operations use extra compute during peak activity so optimizations run reliably, and you get SYS_AUTOMATIC_OPTIMIZATION visibility into what’s happening.
Plus, AWS included cost-control limits so these optimizations can’t grow compute without bounds.
Amazon EKS Auto Mode can send managed capability logs as CloudWatch Vended Logs
EKS Auto Mode can now configure its managed capabilities as CloudWatch Vended Logs sources for reliable, lower‑cost log delivery of autoscaling, storage, and networking observability. EKS Auto Mode’s managed capabilities can be configured as CloudWatch Vended Logs sources, improving delivery reliability and potentially reducing log handling overhead.
Also, this helps centralize autoscaling, storage, and network logs for debugging and cost analysis of managed Kubernetes operations.
February 6, 2026
New partner revenue measurement for service consumption attribution
This provides clearer attribution of how partner solutions generate AWS service usage and revenue.
Consequently, partners and customers can improve FinOps reporting and understand service consumption tied to third‑party offerings.
AWS Marketplace supports localized billing for Professional Services in EMEA
This reduces procurement friction and simplifies billing and chargeback for EMEA customers and partners.
New EC2 C8id / M8id / R8id instances, better compute and memory price‑performance
AWS states these families offer up to 43% higher compute performance and up to 3.3x memory bandwidth versus prior generations, plus flexible Instance Bandwidth Configuration.
As a result, you can choose instance types that improve price‑performance for compute, balanced, or memory‑heavy workloads without wholesale architecture changes.
G7e GPU instances for faster inference, higher throughput per dollar
AWS claims up to 2.3x inference performance versus the prior generation and supports On‑Demand, Spot, and Savings Plans.
Redshift autonomics now work across multi‑cluster setups
Quick read: Amazon Redshift extended its autonomic features to multi‑cluster environments.
Amazon applied ATO, ATS, Auto Vacuum, and Auto Analyze across consumer clusters so optimizations consider query patterns across the whole deployment.
EventBridge payloads grow to 1 MB
This lets you put richer payloads—like larger LLM prompts or telemetry—into single events instead of using external storage or chunking.
ECS publishes container health as a CloudWatch metric
Heads up: Amazon ECS now emits UnHealthyContainerHealthStatus to CloudWatch Container Insights.
ECS will report container health status so you can set alarms and automate remediation when containers become unhealthy.
Lambda gets better observability for Kafka event source mappings, debug and control retry costs
Lambda now provides log level options and grouped metrics for Kafka polling and processing behavior.
As a result, you get clearer visibility into retries, backoff, and processing bottlenecks that can drive extra compute and invocation costs.
And that helps teams tune throughput and error handling to keep function-related spend under control.
DynamoDB global tables now support multi‑account replication, cleaner cost allocation and governance
Note: Amazon DynamoDB global tables can now replicate across AWS accounts and Regions. This enables organizational isolation while using global tables pricing, so teams can keep accounts separate for governance or billing.
GameLift Servers scale to and from zero, avoid idle instance charges
Quick: Amazon GameLift Servers added automatic scaling that can scale down to zero and back up.
Lightsail adds memory‑optimized instance bundles up to 512 GB
These bundles offer simpler, predictable pricing paths for memory‑heavy workloads that previously needed larger EC2 setups.
Amazon Keyspaces pre‑warming (WarmThroughput) for tables
You can pre‑warm tables in both provisioned and on‑demand modes with a one‑time charge to reduce throttling during launches or campaigns.
January 30, 2026
Dashboard fixes in the AWS CUDOS FinOps framework improve reporting accuracy
CUDOS Dashboard v5.7.1 fixes week number calculations and Shield Advanced visuals.
The release corrects week number calculations and restores accurate visuals for Shield Advanced paid subscription costs, so trend lines and executive reports reflect the right time slices.
Also, those fixes reduce reporting noise and improve the reliability of FinOps insights, helping teams trust dashboards when making cost decisions.
CORA Dashboard visual fix makes multi-period savings easier to spot
CORA Dashboard v0.0.11 fixes the “Top Potential Savings Over Time” visual.
This patch ensures the chart shows all dataset rows instead of dropping some periods, improving visibility into potential savings that span multiple reporting windows.
Moreover, that means FinOps teams can more reliably identify multi‑period opportunities and present consistent savings narratives to stakeholders.
Longer Bedrock prompt caches for Anthropic models can cut inference spend
Amazon Bedrock now supports a 1-hour prompt caching TTL for select Anthropic models.
The feature adds a 1‑hour cache option (in addition to the default 5‑minute cache) for select Anthropic Claude models, letting identical prompts reuse cached context for longer.
Also, the 1‑hour cache is billed at a different rate than the default, so you can trade off cache duration and cost depending on workload patterns.
Faster RDS Blue/Green switchovers cut deployment risk and downtime
Amazon RDS Blue/Green Deployments reduces single‑Region switchover downtime to about five seconds.
The improvement brings typical switchover time for writer instance changes down to ~5 seconds for direct endpoint clients during engine upgrades, maintenance, or scaling.
Plus, that enables safer deployment patterns with minimal user impact, which reduces operational risk and the cost of prolonged maintenance windows.
More regions for C8i instances and C8i‑flex for better migration choices
Amazon EC2 C8i and C8i‑flex instances are available in Asia Pacific (Sydney) and Europe (Frankfurt) and C8i is also now in Europe (London).
These Intel® Xeon 6‑based compute‑optimized instances offer up to ~15% better price‑performance and higher memory bandwidth versus previous Intel generations.
Also, you can buy them On‑Demand, Spot, or via Savings Plans to optimize cost based on workload predictability.
C8gn Graviton4 instances and Graviton4 DB families expand region coverage
C8gn offers up to ~30% better compute performance and high network bandwidth (up to 600 Gbps), while Graviton4 DB instances report 29–40% price/performance gains versus prior generations.
Moreover, broader availability means you can migrate distributed workloads and databases regionally to realize Graviton4 price‑performance benefits.
January 23, 2026
Instance Scheduler adds enhanced scaling, event‑driven automation, and reliability features
New features let the scheduler react to tag events, attempt alternate instance types on capacity shortfalls, and provide tags to help teams troubleshoot.
That improves reliability of scheduled start/stop automation and reduces idle instance waste.
Amazon EBS now lets you modify a volume up to four times in 24 hours, respond faster to storage needs
AWS just made it easier to adjust your cloud storage on the fly.
Amazon Elastic Block Store (EBS) now supports up to four Elastic Volumes modifications per volume within a rolling 24‑hour window, up from the prior cadence. You can change size, volume type, and performance (IOPS/throughput) and start a new modification as soon as the previous one completes, provided you’ve done fewer than four in the past 24 hours.
This works without detaching volumes or restarting instances, so applications keep running during changes. Also, the capability is automatically enabled and available in all commercial, GovCloud, and China Regions.
Amazon ECR adds cross‑repository layer sharing (blob mounting), reduce image storage and push times
Shared layers remove duplicate storage for common base layers and speed up pushes by mounting blobs instead of pushing identical layers repeatedly.
January 16, 2026
Faster, clearer billing with the enhanced Transactions view in the Billing Console
The updated Transactions view now loads pages in milliseconds instead of minutes and was rolled out to all customers on January 12, 2026 across all AWS commercial regions. It gives teams access to complete transaction histories without timeouts, even for accounts with tens of thousands of transactions.
Additionally, the view adds consolidated balance tracking, clear +/- indicators for amounts owed versus funds available, improved status indicators, and advanced filters including a Usage Consolidation Account column for organizations using Billing Transfer which reduces reconciliation time and confusion.
Get exact AI chargeback: AWS Data Exports reports granular Amazon Bedrock operation types
Cost and Usage Reports (CUR), CUR 2.0, and Data Exports for FOCUS now include granular Amazon Bedrock operation names (for example InvokeModelInference and InvokeModelStreamingInference), making it possible to attribute model spend by operation and provider.
This helps FinOps teams do precise chargeback and billing analysis for foundation-model usage, enabling more accurate cost allocation and targeted optimization across Bedrock models.
Attribute transient analytics costs: EMR Serverless adds job run–level cost allocation
Amazon EMR Serverless can now attribute costs to individual job runs for finer chargeback.
EMR Serverless supports job run–level tagging and cost allocation so individual runs show up in Cost Explorer and in Cost and Usage Reports (CUR). This lets teams see exact costs per job or domain instead of only application- or cluster-level aggregates.
Control runaway analytics costs with Redshift Serverless queue-based query management
Redshift Serverless adds queue-based query resource management to limit expensive queries.
You can create dedicated query queues with custom monitoring rules that can auto-abort long-running or expensive queries and assign queues by user role or query group. This gives engineering and FinOps teams an automated lever to enforce cost-aware SLAs and prevent single queries from blowing analytics budgets.
Lower TCO option: Amazon Neptune adds Graviton3 R7g and Graviton4 R8g instances (priced ~16% lower)
Neptune supports R7g (Graviton3) and R8g (Graviton4) instances on engine v1.4.5+; the announcement notes R7g/R8g are priced about 16% lower versus R6g.
New memory-optimized EC2 family (X8i) now GA
AWS launched the X8i memory-optimized instances generally available for heavy-memory workloads.
The X8i family (Intel Xeon 6 custom) offers up to 1.5x memory capacity up to 6 TB and is available via On-Demand, Savings Plans, and Spot.
S3 Storage Lens organization-wide metrics available in AWS GovCloud (US)
S3 Storage Lens org-wide metrics are now available in GovCloud (US) regions to spot storage waste.
Organization-wide object usage and activity metrics, including identification of non-current versions, incomplete multipart uploads, and cross‑Region transfer hotspots are now supported in GovCloud; advanced metrics also offer longer retention for trend analysis.
This helps FinOps teams find storage waste and prioritize lifecycle or replication changes in regulated or government environments.
January 9, 2026
EC2 Capacity Manager adds Spot interruption metrics
EC2 Capacity Manager now reports Spot interruption metrics across accounts and regions. EC2 Capacity Manager added Spot Total Count, Spot Total Interruptions, and Spot Interruption Rate metrics so you can quantify Spot reliability.
That gives FinOps and platform teams the visibility to tune Spot strategies, weigh cost vs. interruption risk, and optimize fleet composition for predictable spend. Moreover, having interruption metrics across accounts and regions helps you identify which zones or instance types yield the best Spot stability for your workloads.
AWS Marketplace Seller Reporting shows collections visibility — better seller cashflow insight
AWS Marketplace Seller Reporting now provides collections visibility in revenue dashboards.
That helps sellers and finance teams forecast cashflow and reconcile revenue timing more accurately, improving financial planning and ops. And for buyers, clearer collection status in feeds can aid reconciliation with vendor invoices.
EC2 instance announcements and regional expansions
AWS rolled out Graviton4 families (M8g, M8gn, M8gb, C8g, C8i) and expanded R8i, I7ie, and others into additional regions, plus GA for M8gn/M8gb.
EC2 Spot reliability and warm buffers for GameLift Streams autoscaling
AWS updated GameLift Streams with real‑time metrics and Gen6 stream classes with enhanced autoscaling. GameLift Streams added per‑session CPU/GPU/VRAM and memory metrics and clearer session error reasons, plus Gen6‑based stream classes and autoscaling with warm buffers.
That helps you choose the right instance classes and tune autoscaling (min/max/target‑idle) to control streaming costs and improve user experience. And real‑time performance stats let you avoid oversized instances and reduce wasted GPU/CPU spend.
January 2nd, 2026
Simplified import of CloudTrail Lake data into CloudWatch
Importing historical CloudTrail Lake events into CloudWatch is straightforward and the import operation itself is free. Also, imported data is billed under CloudWatch custom logs pricing, so you’ll want to account for ingestion and storage in observability budgets.
