Bedrock Brief 25 Feb 2026

Pleri AI

Pleri AI

— 4 min read
Bedrock Brief 25 Feb 2026

Welcome to this week's Bedrock Brief, where AWS is serving up AI with a side of southern hospitality. Amazon just announced plans to invest a whopping $12 billion in new AI data centers in Louisiana. Looks like the Bayou State is trading gumbo for GPUs. This massive investment is part of Amazon's $200 billion capital expenditure plan for 2026, with most of that cash earmarked for AI initiatives. Wall Street's not exactly doing cartwheels over the spending spree, but hey, you can't make an AI omelet without breaking a few eggs (and bank accounts).

Meanwhile, AWS is bringing vertical video creation into the 21st century with their new AI-powered Elemental Inference tool. Gone are the days of manually cropping footage like some kind of digital barbarian. Now, broadcasters can churn out TikTok-ready clips faster than you can say "vertical integration." NBC and Fox are already on board, presumably eager to make their sports highlights as addictive as cat videos.

In less celebratory news, AWS reported that over 600 Fortinet firewalls fell victim to an AI-assisted hacking campaign. The attackers used commercial language models to plan their assault, proving that even cybercriminals are hopping on the AI bandwagon. It's a stark reminder that while AI can help build better defenses, it can also be weaponized by the bad guys. As the old saying goes, "With great processing power comes great responsibility" ... or something like that.

Fresh Cut

  • Amazon Bedrock simplifies AI tool integration by allowing developers to connect and execute server-side tools directly through API calls, reducing complexity and speeding up AI-powered applications. Read announcement →
  • Developers can now troubleshoot AWS infrastructure and application issues faster using AI-assisted workflows in their IDE, with the new AWS Observability Kiro power providing instant access to CloudWatch, Application Signals, CloudTrail, and AWS documentation. Read announcement →
  • AWS Elemental Inference automatically generates vertical content and highlight clips from live video streams, helping broadcasters reach mobile audiences without requiring AI expertise or extra staff. Read announcement →
  • Amazon RDS users in AWS GovCloud (US) can now export database snapshots to S3 in Apache Parquet format, enabling easier data analysis and machine learning without impacting database performance. Read announcement →
  • Amazon's new EC2 C8i and C8i-flex instances, powered by custom Intel Xeon 6 processors, offer up to 60% faster performance for web applications and are now available in Malaysia and Brazil, providing developers with improved compute options for memory-intensive workloads. Read announcement →
  • Amazon Q Developer now visualizes AWS resource and cost data in tables and charts, making it easier for developers to understand their cloud infrastructure through natural language queries. Read announcement →
  • Developers can now generate AWS IAM policies directly within their IDE using the IAM Policy Autopilot tool, simplifying the process of creating secure access controls for AWS applications. Read announcement →
  • AWS's Automated Reasoning policies now link rules to source documents, making it easier for developers to review and refine AI-generated content compliance checks with up to 99% accuracy in detecting hallucinations. Read announcement →
  • AWS IAM Identity Center, a free service for managing workforce access across AWS applications with single sign-on, is available in the Asia Pacific (New Zealand) Region, expanding its reach to 38 AWS Regions globally. Read announcement →
  • EC2 G7e instances with NVIDIA RTX PRO 6000 GPUs are available in Tokyo, offering 2.3x faster inference for large language models and AI workloads, enabling developers to boost performance for cutting-edge AI applications in the Asia Pacific region. Read announcement →

The Quarry

Evaluating AI agents: Real-world lessons from building agentic systems at Amazon

Amazon's new framework for evaluating AI agents is like a Swiss Army knife for measuring agentic AI performance, tackling the thorny challenge of standardizing assessments across wildly different agent implementations. At its core, it combines a flexible workflow that can adapt to various agent types with a robust evaluation library that dishes out metrics faster than a caffeinated barista during the morning rush. For the tech-savvy, the framework's secret sauce lies in its ability to seamlessly integrate Amazon Bedrock AgentCore Evaluations with custom, use case-specific metrics, giving developers a 360-degree view of their agent's capabilities and quirks. Read blog →

More posts:

Core Sample

How Hawk Revolutionizes Financial Crime Detection with Explainable AI

Hawk's AI-native platform revolutionizes financial crime detection by leveraging Amazon SageMaker for machine learning models and AWS Lambda for real-time transaction monitoring. Their explainable AI approach empowers compliance teams to efficiently identify and investigate risks, offering unprecedented accuracy in detecting money laundering and preventing fraud. By partnering with banks and payment providers, Hawk strengthens financial crime and compliance programs, delivering transparent insights that transform how institutions tackle complex regulatory challenges. Watch video →

More videos:

Read more