AWS and Google Are... Partners Now?

Welcome back to The Cloud Cover, your essential guide to navigating the dynamic world of cloud for Solutions Architects, engineers, and IT leaders. This week, the center of gravity was undeniably Las Vegas. AWS re:Invent 2025 wraps up today, and the message from Matt Garman and team was clear: the era of simply renting virtual machines is over. The focus has shifted aggressively toward "agentic" workflows, custom silicon, and a surprising pivot regarding how AWS plays with its neighbors. Let’s dive in.

For years, the AI narrative has been dominated by who has the "smartest" model. At re:Invent, AWS continued their pursuit of a slightly different strategy: they aren't just trying to win the model war; they are trying to own the entire factory floor, from the silicon up to the application layer.

The headliner here is the sheer scale of the infrastructure investment. AWS unveiled Trainium3, a 3nm chip that reportedly delivers a 4.4x performance increase over its predecessor. These chips will help the buildout of "Project Rainier," a massive compute cluster comprised of over 1 million Trainium chips. This is a clear signal to the market (and specifically Nvidia) that AWS is building "frontier-capable" silicon at a scale designed to break the "Nvidia tax" on training workloads.

On top of this silicon, AWS is hoping to train a genuinely competitive set of models with the second generation of their Amazon Nova family. While early benchmarks show the Nova 2 Pro slightly trailing GPT-4o in absolute accuracy, it claims to be 65% more cost-effective and 97% faster. AWS is betting that for enterprise agents, speed and price ("intelligence margins") matter more than having the absolute highest IQ in the room.

Perhaps the most intriguing strategic move is Nova Forge. Addressing the "catastrophic forgetting" that plagues traditional fine-tuning, this service allows enterprises to build "Novellas"—custom models injected with proprietary data alongside the original training data. It’s a compelling pitch: owning a model that deeply understands your business logic without sacrificing general reasoning capabilities.

Will these moves be enough to help Amazon leapfrog their AI competition? Only time will tell, but it is interesting to see them playing a different strategic game than some other providers.

For a massive chunk of its history, AWS operated with a "bring your data to us" mentality. Interoperability was possible, but rarely prioritized. That changed this week with two announcements that signal a major shift in how the cloud giant views the hybrid landscape.

First, the launch of AWS AI Factories brings full-stack AI infrastructure (including Trainium chips and Nvidia Blackwell GPUs) directly into customer data centers. This acknowledges a reality that "data gravity" and strict sovereignty laws (like the EU AI Act) often make moving petabytes to US East 1 impossible. AWS is effectively productizing sovereignty, bringing the cloud to the data rather than forcing the reverse.

Second, and perhaps more surprising, was the preview of AWS Interconnect - Multicloud. In a rare collaborative move, AWS and Google Cloud launched a service to provision private, high-bandwidth links between the two clouds "in minutes." For Solutions Architects, this is a massive quality-of-life improvement. It suggests that AWS is accepting a multi-cloud reality where customers might run analytics in BigQuery while training models on Trainium, and they are finally paving the roads to make that architecture viable.

We are wrapping up the busy conference season, having just finished Microsoft Ignite and AWS re:Invent. Catch an interesting virtual session or two over the holiday season.

It is easy to get lost in the buzzword soup of that can come out of Vegas this time of year. However, if you look past the marketing, the tooling is undeniably maturing. We are moving from "experimental" AI chatbots to "industrial" AI factories.

The interesting test for 2026 will be adoption. Will enterprises actually trust a "Nova" model to run their business logic just because it's cheaper and integrated into AWS, or will the "intelligence premium" of OpenAI and Anthropic keep them on top?

Until next week, keep building.

Do you enjoy these emails? Your friends and colleagues might, too! Help us grow the cloud community by sharing the newsletter with others.