In Season 5, Episode 15, Karl and Jon are joined by Damien Jones, an AWS Community Builder, to discuss AWS Interconnect, now generally available for multi-cloud connectivity with Google Cloud Platform, with Azure and Oracle Cloud coming later; database migration acceleration using Kiro and Amazon Bedrock Agent Core to speed up migrations to Amazon Aurora DSQL; Project Glasswing, Anthropic’s restricted-preview model for detecting AI-driven cyberattacks and identifying vulnerabilities; Amazon’s AI revenue, with the CEO revealing $15 billion in annualized AI services revenue, roughly 10% of AWS’s run rate; and Project Houdini, AWS’s initiative using prefabricated modular data centers to accelerate construction timelines. And, of course, the guys got excited about the prospect of a Lidl cloud platform... 07:40 - AWS Interconnect - Multi-Cloud Connectivity AWS has announced the general availability of AWS Interconnect, a dedicated service for connecting AWS with other cloud providers more reliably and efficiently than VPNs. It currently supports Google Cloud, with Azure and Oracle Cloud expected by late 2026. Pricing depends on capacity and distance, starting around $90,000 per month for 10 Gbps between nearby regions and rising to nearly $400,000 for longer cross-region links. AWS has also open-sourced the specification on GitHub to encourage broader adoption. The service removes unpredictable internet egress fees and guarantees capacity, making it most relevant for large enterprises with hybrid or multi-cloud environments. Still, it is a premium solution for moving data between clouds, not for reducing multi-cloud complexity itself. 17:22 - Accelerating Database Migration with Kiro and Bedrock Agent Corp AWS shared a technical guide showing how Kiro and Amazon Bedrock Agent Core can speed up schema analysis for migrations to Amazon Aurora DSQL. The approach helps identify schema mapping needs and compatibility issues early, reducing the need for deep migration expertise during planning. But the discussion raised concerns about production readiness: it depends on persistent Kiro CLI sessions that lose in-memory analysis if interrupted, forcing a restart, and it lacks the real-time observability of native AWS DMS tools. While useful for proof-of-concept work and easing upfront analysis, the panelists were cautious about recommending it for production migrations without stronger persistence and observability. More broadly, they noted that AI-driven “faster” database migration tooling is part of a familiar cycle, while the core migration challenges remain largely the same. 27:06 - Project Glasswing - AI-Driven Cybersecurity Tool Anthropic launched Project Glasswing, a restricted-preview model aimed at detecting and preventing AI-driven cyberattacks by finding software vulnerabilities. It reportedly uncovered thousands of critical bugs in core internet infrastructure, including projects like FFMPEG and OpenSSL, often maintained by very small teams. Access is limited to about 40 organizations, which sparked debate over publicly promoting a powerful tool that few can use. The discussion raised concerns about a two-tier security landscape, possible future “Glasswing scan” requirements in cyber insurance, and broader AI safety issues as models grow more capable. While restricting access to dangerous tools may be sensible, the panelists argued that the public hype creates perverse incentives and could let a small group of firms charge premium prices for exclusive access. 38:29 - Amazon AI Revenue and Investment Strategy Amazon CEO Andy Jassy said in the annual shareholder letter that AWS AI services are now generating more than $15 billion in annualized revenue, about 10% of AWS’s total run rate. Amazon has committed $200 billion in capital spending for data centers and AI chips, reflecting strong demand for specialized infrastructure. Jassy also noted that two major AWS customers asked for exclusive access to all Graviton capacity in 2026, a request Amazon declined to avoid limiting other customers. The letter underscored the strategic value of AWS’s AI and chip business, with discussion pointing to a more disciplined approach: Amazon is now securing customer demand before building capacity, with production already committed into 2027 and 2028. While the ROI horizon is still long given the scale of spending, demand and adoption appear to be accelerating. 45:16 - Project Houdini - Prefabricated Data Center Construction AWS announced Project Houdini, which uses prefabricated modular data center units, or “skids,” to speed up data center construction and AI infrastructure deployment. While the idea of prefabrication is not new, AWS is standardizing it at scale to cut build times. The panelists noted the bigger constraint is power, not construction: aging UK and European grids are already under strain and often cannot support modern data center demand. That has pushed companies like Amazon, Google, and Microsoft to tackle infrastructure problems typically handled by governments. Amazon is exploring options such as small modular nuclear reactors, but broader questions remain about whether private companies should be solving national power challenges. One advantage of the modular approach is portability, allowing deployment in areas where power is available.
