Amazon Elastic Compute Cloud (EC2) C8id instances powered by custom Intel Xeon 6 processors feature up to 384 vCPUs, 768GiB of memory, and 22.8TB of NVMe SSD storage and deliver up to 43% higher performance and 3.3x more memory bandwidth compared to previous generation C6id instances. Starting today, C8id instances are available in Europe (Spain) region.
These instances deliver up to 46% higher performance for I/O intensive database workloads, and up to 30% faster query results for I/O intensive real-time data analytics than previous sixth-generation instances. Additionally, these instances support Instance Bandwidth Configuration, allowing 25% flexible allocation between network and EBS bandwidth, allocating resources optimally for each workload.
C8id instances are ideal for compute-intensive workloads such as high-performance web servers, batch processing, distributed analytics, ad serving, video encoding, and gaming servers.
C8id instances are available in US East (N. Virginia, Ohio), US West (Oregon), Europe (Frankfurt, Spain), and Asia Pacific (Tokyo) regions. Customers can purchase these instances via Savings Plans, On-Demand instances, and Spot instances. For more information visit the Amazon EC2 instance type page.
Amazon Elastic Compute Cloud (EC2) C8id instances powered by custom Intel Xeon 6 processors feature up to 384 vCPUs, 768GiB of memory, and 22.8TB of NVMe SSD storage and deliver up to 43% higher performance and 3.3x more memory bandwidth compared to previous generation C6id instances. Starting today, C8id instances are available in Europe (Spain) region. These instances deliver up to 46% higher performance for I/O intensive database workloads, and up to 30% faster query results for I/O intensive real-time data analytics than previous sixth-generation instances. Additionally, these instances support Instance Bandwidth Configuration, allowing 25% flexible allocation between network and EBS bandwidth, allocating resources optimally for each workload. C8id instances are ideal for compute-intensive workloads such as high-performance web servers, batch processing, distributed analytics, ad serving, video encoding, and gaming servers. C8id instances are available in US East (N. Virginia, Ohio), US West (Oregon), Europe (Frankfurt, Spain), and Asia Pacific (Tokyo) regions. Customers can purchase these instances via Savings Plans, On-Demand instances, and Spot instances. For more information visit the Amazon EC2 instance type page.
Amazon Elastic Compute Cloud (Amazon EC2) C8gd and M8gd instances with up to 11.4 TB of local NVMe-based SSD block-level storage are now available in additional regions. C8gd instances are now available in South America (Sao Paulo). M8gd instances are now available in Europe (Ireland). These instances are powered by AWS Graviton4 processors, delivering up to 30% better performance over Graviton3-based instances. They have up to 40% higher performance for I/O intensive database workloads, and up to 20% faster query results for I/O intensive real-time data analytics than comparable AWS Graviton3-based instances. These instances are built on the AWS Nitro System and are a great fit for applications that need access to high-speed, low latency local storage.
Each instance is available in 12 different sizes. They provide up to 50 Gbps of network bandwidth and up to 40 Gbps of bandwidth to the Amazon Elastic Block Store (Amazon EBS). Additionally, customers can now adjust the network and Amazon EBS bandwidth on these instances by 25% using EC2 instance bandwidth weighting configuration, providing greater flexibility with the allocation of bandwidth resources to better optimize workloads. These instances offer Elastic Fabric Adapter (EFA) networking on 24xlarge, 48xlarge, metal-24xl, and metal-48xl sizes.
Amazon Elastic Compute Cloud (Amazon EC2) C8gd and M8gd instances with up to 11.4 TB of local NVMe-based SSD block-level storage are now available in additional regions. C8gd instances are now available in South America (Sao Paulo). M8gd instances are now available in Europe (Ireland). These instances are powered by AWS Graviton4 processors, delivering up to 30% better performance over Graviton3-based instances. They have up to 40% higher performance for I/O intensive database workloads, and up to 20% faster query results for I/O intensive real-time data analytics than comparable AWS Graviton3-based instances. These instances are built on the AWS Nitro System and are a great fit for applications that need access to high-speed, low latency local storage. Each instance is available in 12 different sizes. They provide up to 50 Gbps of network bandwidth and up to 40 Gbps of bandwidth to the Amazon Elastic Block Store (Amazon EBS). Additionally, customers can now adjust the network and Amazon EBS bandwidth on these instances by 25% using EC2 instance bandwidth weighting configuration, providing greater flexibility with the allocation of bandwidth resources to better optimize workloads. These instances offer Elastic Fabric Adapter (EFA) networking on 24xlarge, 48xlarge, metal-24xl, and metal-48xl sizes. To learn more, see Amazon C8gd Instances and Amazon M8gd Instances. To explore how to migrate your workloads to Graviton-based instances, see AWS Graviton Fast Start program and Porting Advisor for Graviton. To get started, see the AWS Management Console.
Starting today, Amazon Elastic Compute Cloud (Amazon EC2) R7gd instances with up to 3.8 TB of local NVMe-based SSD block-level storage are available in South America (Sao Paulo) Region.
R7gd are powered by AWS Graviton3 processors with DDR5 memory are built on the AWS Nitro System. They are ideal for memory-intensive workloads such as open-source databases, in-memory caches, and real-time big data analytics and are a great fit for applications that need access to high-speed, low latency local storage, including those that need temporary storage of data for scratch space, temporary files, and caches.
Starting today, Amazon Elastic Compute Cloud (Amazon EC2) R7gd instances with up to 3.8 TB of local NVMe-based SSD block-level storage are available in South America (Sao Paulo) Region. R7gd are powered by AWS Graviton3 processors with DDR5 memory are built on the AWS Nitro System. They are ideal for memory-intensive workloads such as open-source databases, in-memory caches, and real-time big data analytics and are a great fit for applications that need access to high-speed, low latency local storage, including those that need temporary storage of data for scratch space, temporary files, and caches. To learn more, see Amazon R7gd Instances. To get started, see the AWS Management Console.
Amazon EC2 High Memory U7i instances with 8TB of memory (u7i-8tb.112xlarge) are now available in AWS Asia Pacific (Hyderabad), and U7i instances with 12TB of memory (u7i-12tb.224xlarge) are now available in AWS Europe (Spain). U7i instances are part of AWS 7th generation and are powered by custom fourth generation Intel Xeon Scalable Processors (Sapphire Rapids). U7i-8tb instances offer 8TiB of DDR5 memory, U7i-12tb instances offer 12TiB of DDR5 memory, enabling customers to scale transaction processing throughput in a fast-growing data environment.
U7i-8tb instances offer 448 vCPUs and U7i-12tb instances offer 896 vCPUs. Both instance types support up to 100 Gbps of Amazon Elastic Block Store (Amazon EBS) bandwidth for faster data loading and backups, up to 100 Gbps of network bandwidth, and ENA Express.
U7i instances are ideal for customers using mission-critical in-memory databases like SAP HANA, Oracle, and SQL Server.
Amazon EC2 High Memory U7i instances with 8TB of memory (u7i-8tb.112xlarge) are now available in AWS Asia Pacific (Hyderabad), and U7i instances with 12TB of memory (u7i-12tb.224xlarge) are now available in AWS Europe (Spain). U7i instances are part of AWS 7th generation and are powered by custom fourth generation Intel Xeon Scalable Processors (Sapphire Rapids). U7i-8tb instances offer 8TiB of DDR5 memory, U7i-12tb instances offer 12TiB of DDR5 memory, enabling customers to scale transaction processing throughput in a fast-growing data environment.
U7i-8tb instances offer 448 vCPUs and U7i-12tb instances offer 896 vCPUs. Both instance types support up to 100 Gbps of Amazon Elastic Block Store (Amazon EBS) bandwidth for faster data loading and backups, up to 100 Gbps of network bandwidth, and ENA Express.
U7i instances are ideal for customers using mission-critical in-memory databases like SAP HANA, Oracle, and SQL Server.
To learn more about U7i instances, visit the High Memory instances page.
Amazon CloudWatch Database Insights expands the availability of its on-demand analysis experience to AWS GovCloud (US-East) and AWS GovCloud (US-West). CloudWatch Database Insights is a monitoring and diagnostics solution that helps database administrators and developers optimize database performance by providing comprehensive visibility into database metrics, query analysis, and resource utilization patterns. This feature uses machine learning models to help identify performance bottlenecks during the selected time period, and gives advice on what to do next.
Previously, database administrators had to manually analyze performance data, correlate metrics, and investigate root cause. This process is time-consuming and requires deep database expertise. With this launch, you can now analyze database performance monitoring data for any time period with automated intelligence. The feature automatically compares your selected time period against normal baseline performance, identifies anomalies, and provides specific remediation advice. Through intuitive visualizations and clear explanations, you can quickly identify performance issues and receive step-by-step guidance for resolution. This automated analysis and recommendation system reduces mean-time-to-diagnosis from hours to minutes.
You can get started with this feature by enabling the Advanced mode of CloudWatch Database Insights on your Amazon Aurora and Amazon RDS databases using the RDS service console, AWS APIs, the AWS SDK, or AWS CloudFormation. Please refer to Aurora documentation or RDS documentation to get started.
Amazon CloudWatch Database Insights expands the availability of its on-demand analysis experience to AWS GovCloud (US-East) and AWS GovCloud (US-West). CloudWatch Database Insights is a monitoring and diagnostics solution that helps database administrators and developers optimize database performance by providing comprehensive visibility into database metrics, query analysis, and resource utilization patterns. This feature uses machine learning models to help identify performance bottlenecks during the selected time period, and gives advice on what to do next. Previously, database administrators had to manually analyze performance data, correlate metrics, and investigate root cause. This process is time-consuming and requires deep database expertise. With this launch, you can now analyze database performance monitoring data for any time period with automated intelligence. The feature automatically compares your selected time period against normal baseline performance, identifies anomalies, and provides specific remediation advice. Through intuitive visualizations and clear explanations, you can quickly identify performance issues and receive step-by-step guidance for resolution. This automated analysis and recommendation system reduces mean-time-to-diagnosis from hours to minutes. You can get started with this feature by enabling the Advanced mode of CloudWatch Database Insights on your Amazon Aurora and Amazon RDS databases using the RDS service console, AWS APIs, the AWS SDK, or AWS CloudFormation. Please refer to Aurora documentation or RDS documentation to get started.
Amazon Connect now enables you to choose the «From» email address when replying to inbound emails or sending new outbound messages, helping contact centers ensure the correct brand or business identity is used for every customer interaction. Administrators can configure multiple sender addresses per queue, allowing agents to search and select the appropriate email address based on the queue they are working in. This capability is especially useful for contact centers that support multiple brands or lines of business from a single Amazon Connect instance.
Amazon Connect email is available in the US East (N. Virginia), US West (Oregon), Africa (Cape Town), Asia Pacific (Seoul), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Canada (Central), Europe (Frankfurt), and Europe (London) regions. To learn more and get started, please refer to the help documentation or visit the Amazon Connect website.
Amazon Connect now enables you to choose the «From» email address when replying to inbound emails or sending new outbound messages, helping contact centers ensure the correct brand or business identity is used for every customer interaction. Administrators can configure multiple sender addresses per queue, allowing agents to search and select the appropriate email address based on the queue they are working in. This capability is especially useful for contact centers that support multiple brands or lines of business from a single Amazon Connect instance. Amazon Connect email is available in the US East (N. Virginia), US West (Oregon), Africa (Cape Town), Asia Pacific (Seoul), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Canada (Central), Europe (Frankfurt), and Europe (London) regions. To learn more and get started, please refer to the help documentation or visit the Amazon Connect website.
Amazon Connect now delivers integrated agent coaching workflows that enable contact center managers to provide timely, targeted feedback directly within the Connect UI. When managers identify improvement opportunities through evaluation scorecards, they can immediately create coaching plans with specific customer interaction examples. For example, a manager can share interactions with an agent where they excelled at problem-solving but could show more customer empathy, with examples of empathetic language to use going forward. After coaching sessions, agents acknowledge feedback and add notes to confirm understanding of expectations and next steps. Both managers and agents access all coaching history on a single page, enabling systematic progress tracking and improved coaching effectiveness. This integrated approach eliminates coaching delays and creates accountability throughout the agent development process, accelerating performance improvement across contact center operations.
This feature is available in all regions where Amazon Connect is offered. To learn more, please visit our documentation and our webpage.
Amazon Connect now delivers integrated agent coaching workflows that enable contact center managers to provide timely, targeted feedback directly within the Connect UI. When managers identify improvement opportunities through evaluation scorecards, they can immediately create coaching plans with specific customer interaction examples. For example, a manager can share interactions with an agent where they excelled at problem-solving but could show more customer empathy, with examples of empathetic language to use going forward. After coaching sessions, agents acknowledge feedback and add notes to confirm understanding of expectations and next steps. Both managers and agents access all coaching history on a single page, enabling systematic progress tracking and improved coaching effectiveness. This integrated approach eliminates coaching delays and creates accountability throughout the agent development process, accelerating performance improvement across contact center operations. This feature is available in all regions where Amazon Connect is offered. To learn more, please visit our documentation and our webpage.
Copilot Cowork: Una nueva forma de realizar el trabajo
Por: Charles Lamanna, presidente de aplicaciones empresariales y agentes.
Si han usado Copilot, han visto lo rápido que puede ayudarlos a encontrar una respuesta o redactar un correo electrónico. El siguiente paso es igual de importante: convertir esa intención en acciones reales en Microsoft 365.
Durante el último año, hemos presionado a Copilot para que tome medidas. Eso significa completar tareas, ejecutar flujos de trabajo y trabajar en su nombre.
Copilot Cowork está diseñado para eso: ayuda a Copilot a actuar, no solo a chatear.
Cowork facilita delegar trabajo. Describan el resultado que desean y Cowork fundamenta en automático el trabajo en sus correos, reuniones, mensajes, archivos y datos. Impulsado por Work IQ, Cowork se basa en señales de Outlook, Teams, Excel y el resto de Microsoft 365 para poder actuar con la misma comprensión que ustedes aportas a su trabajo.
Cuando entregan una tarea a Cowork, su petición se convierte en un plan. El plan continúa en segundo plano, con puntos de control claros para que puedan confirmar el progreso, hacer cambios o pausar la ejecución en cualquier momento. Cowork pregunta si necesita aclaraciones. Ustedes pueden ver cualquier acción que recomiende y luego aprobar los cambios antes de que se apliquen. Copilot funciona de manera independiente sin que ustedes pierdan el control.
En las últimas semanas, lo que más ha destacado es lo natural que encaja Cowork en un día ajetreado. Es fácil tener una docena de tareas en vuelo a la vez, donde cada una avanza mientras ustedes se concentras en lo que solo ustedes puedes hacer. La era de la ejecución de Copilot ha llegado.
Ese bucle de planificación a acción es la diferencia entre obtener una respuesta y lograr que algo se haga. Aquí tienen cuatro ejemplos donde Cowork convierte la intención en acciones en Microsoft 365. Cada una comienza con una simple petición y termina con acciones que permanecen bajo su control. En cada escenario, Cowork no solo crea contenido, sino que coordina el trabajo que lo rodea.
1. Limpiar su calendario: Reprogramar reuniones y proteger el tiempo de concentración
La mayoría de las semanas empiezan con un calendario abarrotado y poco tiempo de concentración. Ahora pueden pasar ese triaje a Cowork para que revise su horario de Outlook, pregunte qué quieren priorizar y señala conflictos y reuniones de bajo valor. Luego propone cambios. Una vez que lo aprueban, aplica los cambios al aceptar, rechazar o reprogramar reuniones y añadir bloques de enfoque. Incluso puede enviar un documento de preparación para la reunión.
Ustedes obtienen una semana más limpia y más tiempo para el trabajo que importa, sin hacerlo de manera manual.
Una vez que su semana está controlada, la siguiente pregunta suele ser a qué se enfrentan y qué tan preparados estarán.
2. Elaborar el paquete de la reunión y alinear al equipo: Generar la presentación, documentar y hacer seguimiento
Preparar una reunión con clientes puede ocupar su tarde. Con Cowork, pueden repartir el esfuerzo de principio a fin. Cowork extrae entradas relevantes de correos electrónicos, reuniones y archivos, programa el tiempo de preparación en el calendario y luego produce un conjunto conectado de entregables: un documento informativo, análisis de apoyo y una presentación lista para el cliente. Todo se guarda en Microsoft 365 para que su equipo pueda pulirlo en conjunto.
El resultado: ustedes entran con una presentación compartible, un documento de presentación en el que su equipo puede alinearse, la hora de preparación programada ya en el calendario y un borrador de actualización por correo electrónico que recoge las decisiones clave y adjunta los archivos más recientes.
Y no son solo reuniones. El mismo enfoque funciona cuando la «preparación» es una investigación más profunda y necesitas algo en lo que puedas confiar.
3. Investigar con rapidez una empresa: extraer fuentes, compilar análisis y empaquetar resultados
La investigación profunda requiere tiempo y rigor. Con Cowork, pueden descargar la investigación de la empresa a través de la web y fuentes de trabajo. En este caso, Cowork recopila informes de resultados, presentaciones de la Comisión de Bolsa y Valores (SEC, por sus siglas en inglés), comentarios de analistas y noticias relevantes, con énfasis en los datos financieros primarios. Luego organiza los hallazgos con las citas. Ustedes reciben un resumen ejecutivo formateado para correo electrónico, un memorando de investigación estructurado con supuestos claros y análisis de apoyo, y un cuaderno de Excel con pestañas etiquetadas.
En lugar de pasar horas en ensamblar entradas, ustedes obtienen salidas que pueden usar de inmediato.
Por último, cuando el trabajo es transversal y sensible al tiempo, Cowork puede coordinar un flujo de trabajo que produzca tanto la narrativa como el plan.
4. Crear el plan de lanzamiento: Construir inteligencia competitiva y activos compartibles
Los lanzamientos de nuevos productos avanzan rápido, en especial cuando el panorama competitivo cambia a mitad de camino. Con Cowork, puedes delegar un flujo de trabajo de lanzamiento y pasar con rapidez de la intención a un enfoque completo. Cowork construye una comparación competitiva en Excel, destila la diferenciación en un documento de propuesta de valor y genera una presentación para clientes. También puede detallar hitos, propietarios y siguientes pasos. Esto no se limita a la estrategia. Se traduce en acciones coordinadas.
Ustedes obtienen una historia coherente con rapidez, además de los archivos que la respaldan, sin tener que unir versiones entre herramientas. A partir de ahí, su equipo puede distribuirlo, revisarlo y seguir con las mejoras a medida que avance el lanzamiento.
Construido para la empresa
Copilot Cowork funciona dentro de los límites de seguridad y gobernanza de Microsoft 365. Las políticas de identidad, permisos y cumplimiento se aplican por defecto, y las acciones y salidas son auditables. Cowork se ejecuta en un entorno de nube protegido dentro de un sandbox, para que las tareas puedan avanzar de forma segura a medida que ustedes se mueven entre dispositivos. Esto es lo que hace que la ejecución sea duradera a gran escala empresarial.
Con el trabajo cercano con Anthropic, hemos integrado la tecnología detrás de Claude Cowork en Microsoft 365 Copilot. Es esta ventaja de múltiples modelos lo que hace que Copilot sea diferente. Su trabajo no está limitado por una sola marca de modelos. Copilot aloja la mejor innovación de toda la industria y elige el modelo adecuado para el trabajo, sin importar quién lo haya construido. Este es un patrón de trabajo que solo se volverá más poderoso a medida que surjan nuevos modelos y formas de trabajar.
Prueben Copilot Cowork
Copilot Cowork está a prueba en la actualidad con un grupo limitado de clientes en Research Preview (Vista Previa de Investigación), y estará disponible de manera más amplia en el programa Frontier a finales de marzo de 2026.
Amazon Bedrock AgentCore Runtime now supports stateful Model Context Protocol (MCP) server features, enabling developers to build MCP servers that leverage elicitation, sampling, and progress notifications alongside existing support for resources, prompts, and tools. These capabilities allow MCP servers deployed to AgentCore Runtime to collect user input interactively during tool execution, request LLM-generated content from clients, and provide real-time progress updates for long-running operations.
With stateful MCP sessions, each user session runs in a dedicated microVM with isolated resources, and the server maintains session context across multiple interactions using an Mcp-Session-Id header. Elicitation enables server-initiated, multi-turn conversations to gather information such as user preferences. Sampling allows servers to request AI-powered text generation from the client for tasks like personalized recommendations. Progress notifications keep clients informed during operations such as searching for flights or processing bookings. These features work together to support complex, interactive agent workflows that go beyond simple request-response patterns.
Stateful MCP server features are supported in AgentCore Runtime across fourteen AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Mumbai), Canada (Central), Asia Pacific (Seoul), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Europe (Frankfurt), Europe (Ireland), Europe (London), Europe (Paris), and Europe (Stockholm).
Amazon Bedrock AgentCore Runtime now supports stateful Model Context Protocol (MCP) server features, enabling developers to build MCP servers that leverage elicitation, sampling, and progress notifications alongside existing support for resources, prompts, and tools. These capabilities allow MCP servers deployed to AgentCore Runtime to collect user input interactively during tool execution, request LLM-generated content from clients, and provide real-time progress updates for long-running operations. With stateful MCP sessions, each user session runs in a dedicated microVM with isolated resources, and the server maintains session context across multiple interactions using an Mcp-Session-Id header. Elicitation enables server-initiated, multi-turn conversations to gather information such as user preferences. Sampling allows servers to request AI-powered text generation from the client for tasks like personalized recommendations. Progress notifications keep clients informed during operations such as searching for flights or processing bookings. These features work together to support complex, interactive agent workflows that go beyond simple request-response patterns.
Stateful MCP server features are supported in AgentCore Runtime across fourteen AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Mumbai), Canada (Central), Asia Pacific (Seoul), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Europe (Frankfurt), Europe (Ireland), Europe (London), Europe (Paris), and Europe (Stockholm).
To learn more, see Stateful MCP server features in the Amazon Bedrock AgentCore documentation.
Amazon Bedrock is a fully managed service for building generative AI applications using high-performing foundation models from leading AI providers. It now supports two new CloudWatch metrics, TimeToFirstToken and EstimatedTPMQuotaUsage, giving you deeper visibility into inference performance and quota consumption.
TimeToFirstToken measures the latency from when a request is sent to when the first token is received, for streaming APIs (ConverseStream and InvokeModelWithResponseStream). You can use this metric to set CloudWatch alarms which monitor latency degradation and establish SLA baselines, without any client-side instrumentation. EstimatedTPMQuotaUsage tracks your estimated Tokens Per Minute (TPM) quota consumption, including cache write tokens and output burndown multipliers, across all inference APIs (Converse, InvokeModel, ConverseStream, and InvokeModelWithResponseStream). You can use this metric to set proactive alarms before reaching your quota limit, track your quota consumption across your models, and request further quota increases before usage is rate limited.
Both metrics are supported in all commercial Bedrock regions for models available via cross-region inference profiles and in-region inference, updated every minute for successfully completed requests. These are available in your CloudWatch out of the box; you pay only for the underlying model inference you consume, with no API changes or opt-in required.
Amazon Bedrock is a fully managed service for building generative AI applications using high-performing foundation models from leading AI providers. It now supports two new CloudWatch metrics, TimeToFirstToken and EstimatedTPMQuotaUsage, giving you deeper visibility into inference performance and quota consumption.
TimeToFirstToken measures the latency from when a request is sent to when the first token is received, for streaming APIs (ConverseStream and InvokeModelWithResponseStream). You can use this metric to set CloudWatch alarms which monitor latency degradation and establish SLA baselines, without any client-side instrumentation. EstimatedTPMQuotaUsage tracks your estimated Tokens Per Minute (TPM) quota consumption, including cache write tokens and output burndown multipliers, across all inference APIs (Converse, InvokeModel, ConverseStream, and InvokeModelWithResponseStream). You can use this metric to set proactive alarms before reaching your quota limit, track your quota consumption across your models, and request further quota increases before usage is rate limited.
Both metrics are supported in all commercial Bedrock regions for models available via cross-region inference profiles and in-region inference, updated every minute for successfully completed requests. These are available in your CloudWatch out of the box; you pay only for the underlying model inference you consume, with no API changes or opt-in required.
To learn more about TimeToFirstToken and EstimatedTPMQuotaUsage, see our documentation page on Monitoring Amazon Bedrock.
