Bedrock Brief 18 Feb 2026
Welcome back, AI aficionados and cloud curious! This week in the world of AWS, it's all about the Benjamins—$200 billion of them, to be precise. Amazon CEO Andy Jassy is betting big on AI, announcing plans to funnel a mountain of cash into AWS infrastructure. It's like watching your normally frugal friend suddenly decide to buy a yacht. Sure, it might be fun, but is it really necessary?
The market's reaction? Let's just say investors aren't exactly popping champagne. Amazon's stock took a nosedive, shedding a cool $450 billion in value during a 9-day losing streak. Ouch. It's as if AWS decided to throw a lavish AI party, but forgot to send out invitations explaining why everyone should be excited. The silver lining? The stock finally snapped its losing streak on Tuesday, proving that even in the cloud, what goes down must eventually come up.
But here's the kicker—Amazon might not have much choice in the matter. AWS has been hemorrhaging market share to Microsoft and Google, with its slice of the cloud pie shrinking to a multiyear low of 28%. It's a classic case of "damned if you do, damned if you don't." Spend big on AI and watch investors sweat, or pinch pennies and risk becoming the Blockbuster of cloud computing. One thing's for sure: the AI arms race is heating up, and AWS isn't about to bring a butter knife to a lightsaber fight.
Fresh Cut
- Amazon Bedrock now allows developers to fine-tune popular open-weight AI models using reinforcement learning, making it easier to improve model accuracy without extensive machine learning expertise or large datasets. Read announcement →
- Amazon Bedrock now offers Claude Sonnet 4.6, Anthropic's latest AI model that excels at coding and professional tasks, enabling developers to create more efficient and reliable browser-based automation tools for businesses. Read announcement →
- Amazon EC2 M7i instances, powered by custom Intel processors offering 15% better performance than competitors, are now available in Israel, providing enhanced capabilities for gaming servers, ML, and video-streaming workloads. Read announcement →
- Amazon Connect's AI now summarizes Tasks and suggests next steps, helping agents process customer requests like refunds more efficiently by providing context and guidance. Read announcement →
- AWS Outposts now offers C8i, M8i, and R8i instances with 20% better performance and 2.5x more memory bandwidth, allowing developers to run more powerful on-premises workloads like larger databases and advanced analytics within the same rack space and power consumption. Read announcement →
- Amazon Bedrock expands its AI model offerings in Sydney, giving developers easier access to powerful tools from DeepSeek, Google, Mistral, and others for building AI applications. Read announcement →
- Amazon Bedrock's expanded AWS PrivateLink support enables secure, private access to OpenAI API-compatible endpoints, allowing developers to build AI applications with enhanced performance and reliability. Read announcement →
- Amazon Bedrock boosts Claude Sonnet 4.5's capacity in GovCloud, allowing developers to process 25 times more AI requests in regulated environments without extra steps. Read announcement →
- AWS Support's AI troubleshooting is now available in seven additional languages, helping developers quickly resolve issues in their preferred language without language barriers. Read announcement →
- AWS expands its Data Transfer Terminal service to six new cities worldwide, enabling developers to rapidly upload massive datasets to the cloud using secure, high-speed physical locations. Read announcement →
The Quarry
Build long-running MCP servers on Amazon Bedrock AgentCore with Strands Agents integration
Tired of your AI agents throwing in the towel on marathon tasks? This post dishes out a clever combo of context message strategies and asynchronous task management to keep your Amazon Bedrock AgentCore agents chugging along like the Energizer Bunny. By integrating these techniques with Strands Agents, you'll craft AI workhorses capable of tackling complex, time-intensive operations without breaking a sweat—or more importantly, without blocking other critical processes. Read blog →
More posts:
- Customize AI agent browsing with proxies, profiles, and extensions in Amazon Bedrock AgentCore Browser
- AI meets HR: Transforming talent acquisition with Amazon Bedrock
- Build long-running MCP servers on Amazon Bedrock AgentCore with Strands Agents integration
- NVIDIA Nemotron 3 Nano 30B MoE model is now available in Amazon SageMaker JumpStart
- Mastering Amazon Bedrock throttling and service availability: A comprehensive guide
- Swann provides Generative AI to millions of IoT Devices using Amazon Bedrock
- How LinqAlpha assesses investment theses using Devil’s Advocate on Amazon Bedrock
Core Sample
Secure Agent Authorization with OAuth 2.0 | Amazon Bedrock AgentCore
Amazon Bedrock AgentCore Identity leverages OAuth 2.0 to lock down your AI agents, ensuring only authorized callers can invoke them. Developers simply configure authorization and deploy, while AgentCore handles the heavy lifting of JWT token validation, caller identity verification, and policy enforcement. The system's inbound request flow is particularly clever, using a multi-step process to scrutinize everything from token signatures to fine-grained permissions before granting access. Watch video →
More videos:
- Secure User Data Access for AI Agents | Amazon Bedrock AgentCore
- Secure External API Access for AI Agents | Bedrock AgentCore
- AWS Global Sales transforms insights delivery with Amazon Quick