All The Biggest Reveals From AWS re:Invent 2025

Amazon’s AWS re:Invent 2025 brings together cloud experts from around the world to witness the company’s latest reveals surrounding new products, services, and announcements regarding agentic AI, cloud infrastructure, and foundation models.

This year’s iteration was held from December 1 to 5 across multiple venues on the Las Vegas Strip, spanning Mandalay Bay, MGM Grand, The Venetian, Caesars Forum, and Encore, among others.

AWS announced AWS Interconnect - multicloud, which is a new feature that brings users simplified multicloud connectivity between AWS and other major cloud service providers. The idea is to remove complicated processes observed across traditional multicloud networking by allowing them to quickly configure connections with other cloud providers.

The company also spoke about Amazon Connect, a complete AI-powered solution delivering automated voice experiences using neural text-to-speech in over 30 languages and automated speech recognition in over 25 languages, as well as Deepgram’s native integration with Amazon SageMaker AI.

Ahead of this, AWS announced a wide range of innovative products and services added to its portfolio:

AWS introduced Graviton5 processors, its most advanced custom chip that can handle a broad set of cloud workloads. It can deliver up to 25% higher performance than its predecessor, sporting 192 cores per chip and a 5x larger cache. Graviton brings customers faster performance, lower costs, and helps them meet sustainability commitments, and is optimized for demanding workloads such as real-time gaming, high-performance databases, big data analytics, application servers, and Electronic Design Automation (EDA).


Amazon’s Nova portfolio finds an expansion with four new models.

Nova 2 Lite spans a fast, cost-effective reasoning model for everyday workloads that processes text, images, and videos to generate text, and is best suited for customer service chatbots, document processing, and business automation.

Nova 2 Pro does the same as Amazon's most intelligent reasoning model, but is more advanced and highly accurate, best suited for complex tasks such as agentic coding, long-range planning, and sophisticated problem-solving.

Both deliver industry-leading price-performance across reasoning, multimodal processing, conversational AI, code generation, and agentic tasks.

Nova 2 Sonic, which is a speech-to-speech model that unifies text and speech understanding and generation for real-time, human-like outputs, features expanded multilingual support with expressive voices, higher accuracy, and a one-million token context window for sustained interactions.

Nova 2 Omni comes as an industry first—a unified multimodal reasoning and generation model that can process text, images, video, and speech inputs to generate text and images. When it comes to size, it handles up to 750,000 words, hours of audio, long videos, and hundred-page documents simultaneously.

Another first-of-its-kind service comes in the form of Nova Forge, which allows organizations to build customized and optimized versions of Nova (dubbed Novellas) by blending proprietary data with Nova’s frontier capabilities using “open training” through pre-trained, mid-trained, and post-trained Nova model checkpoints.

Nova Act, powered by Nova Lite 2, allows developers to build and deploy highly reliable AI agents that can take actions in web browsers.

AWS also launched 3 frontier agents, which act as extensions of an organization’s software development team.

It includes Kiro—which works as the team’s virtual developer that maintains context and learns over time—AWS Security Agent—which is an advanced security consultant for app design, code reviews, and penetration testing, helping build secure applications—and AWS DevOps Agent—which acts as an on-call operational team that helps resolve and proactively prevent incidents with continuous improvement.

As per Amazon, frontier agents are a new class of AI agents that are autonomous, scalable, and work for hours or days without constant intervention.

AWS also revealed Amazon EC2 Trn3 UltraServers.

These are powered by AWS's first 3nm AI chip. It can pack up to 144 Trainium3 chips into a single integrated system, which delivers up to 4.4x more compute performance and 4x greater energy efficiency than its predecessor. It also offers 3x higher throughput per chip and 4x faster response times, which help reduce training times from months to weeks. Moreover, Amazon Bedrock is already serving production workloads on Trainium3.

The innovation comes as AWS recognized “as AI models grow in size and complexity, training cutting-edge models requires infrastructure investments that only a handful of organizations can afford.”

Partnering with industry-leading GPU maker NVIDIA, AWS also unveiled AWS AI Factories, which transforms customers’ existing infrastructure into high-performance AI environments by deploying dedicated AWS AI infrastructure in their own data centers.


However, the coup de grace is that it combines cutting-edge NVIDIA GPUs, Trainium chips, AWS high-speed, low-latency networking, high-performance storage and databases, security, and energy-efficient infrastructure, blended with AWS AI services such as Amazon Bedrock and SageMaker AI.

“AWS AI Factories operate like a private AWS Region that gives secure, low-latency access to compute, storage, database, and AI services,” said AWS.

What do you think about the latest AWS announcements?

Let us know in the comments below!