(Amazon AWS)

Amazon rolls out updated Trainium chip, new AI models at re:Invent conference

A flurry of product announcements from the developer conference in Las Vegas present a road map for Amazon’s AI computing dominance.

Jon Keegan

12/2/25 3:04PM

Today in Las Vegas, at its annual re:Invent developer conference, Amazon AWS made a bunch of product announcements that reveal its plan to continue capitalizing on the AI boom.

AWS currently enjoys a substantial lead in the cloud computing market, serving up AI infrastructure in the cloud to customers large and small using the models of their choice.

Trainium3

Probably the most consequential announcement was about the company’s new Trainium3 custom AI chip. Three years out from ChatGPT’s disruptive debut, AI companies are diversifying their computing resources away from GPU juggernaut Nvidia and starting to sample the price and performance benefits of alternative custom chips, such as Trainium.

Today, AWS announced the Trainium3 UltraServer, powered by the new Trainium3 chips. The company said the Trainium3 chips are 4x as fast and can train models for half the cost of the previous generation. AWS CEO Matt Garman acknowledged the misleading name of the chip, which excels not only at training AI models but also running AI models, known as inference.

“People often give us a little bit of a hard time about product naming in AWS. No, no, it’s true. Well, it turns out Trainium is no exception. We named it Trainium because it’s designed to be an awesome chip for Al training, and it is, but as it turns out, Trainium2 is actually the best system in the world currently for inference,” Garman said.

The company said work on Trainium4 is well underway, and the chips would be compatible with Nvidia systems.

Nova 2 models

Amazon announced updates to its own frontier AI model family called Nova:

Nova Lite (fast and cheap for everyday tasks),
Nova Pro (for complex reasoning workloads),
Nova Sonic (a new speech-to-speech model), and
Nova Omni (an “all-in-one model for multimodal reasoning and image generation”).

The new models may help the company offer cheaper, more efficient AI computing for its customers versus running competing frontier models on popular Nvidia GPUs.

Nova Forge

Complementing the new Nova 2 models is a product called “Nova Forge,” which makes it easier to train customized models using the Nova models as a base. Customers can bring custom data into the training process to easily create specialized expert models using the tool.

AI Factories

Amazon is also rolling out a way for customers to run AWS services inside their own data centers, keeping their data under their control. Amazon AI Factories works with both Amazon Trainium chips as well as Nvidia GPUs. Amazon said an AI factory is like deploying your own private AWS region.

Agents, agents, agents

AI agents were mentioned quite a bit during the keynote, and the company had lots of new offerings to help customers deploy autonomous AI helpers.

Nova Act: A new AWS service that lets customers spin up “fleets of reliable AI agents for automating production UI workflows.”

DevOps Agent: A new “frontier agent” that autonomously helps solve software failures on production systems when human engineers aren’t online. Amazon describes the new agent as “our always-on, autonomous on-call engineer.”

Jon Keegan

1/16/26

OpenAI set to start rolling out ads to free and ChatGPT Go users

OpenAI says user conversations will be kept private from advertisers, and answers will not be influenced by ads.

Jon Keegan1/16/26

EPA: xAI’s Colossus data center illegally used gas turbines without permits

The Environmental Protection Agency has ruled that xAI violated the law when it used dozens of portable gas generators for its Colossus 1 data center without air quality permits.

When xAI set out to build Colossus 1 in Memphis, Tennessee, CEO Elon Musk wanted to move with unprecedented speed, avoiding all of the red tape that could slow such a big project down.

To power the 1-gigawatt data center, Musk took advantage of a local loophole that allowed portable gas generators to be used without any permits, as long as they did not spend more than 364 days in the same spot. That allowed xAI to bring in dozens of truck-sized gas generators to quickly supply the massive amount of power the data center needed to train xAI’s Grok model.

The new EPA rule says the use of such portable generators falls under federal regulation, and the company did need air quality permits to operate the turbines. xAI is also using dozens of such generators to power its Colossus 2 data center just over the border in Alabama.

Elon Musk’s xAI datacenter generating extra electricity illegally, regulator rules

Rani Molla

AI candy

1/16/26

Google has a leg up over Apple in the race to roll out personalized AI

Google and Apple are going for the holy grail of AI chatbot integration, but it looks like Google is getting there first.

Rani Molla1/16/26

Trump to push Big Tech to fund new power plants as AI drives up electricity costs

President Donald Trump is expected to announce a plan Friday morning that would require Big Tech companies to bid on 15-year contracts for new electricity generation capacity. The move would effectively force companies to help fund new power plants in the PJM region as soaring demand from AI data centers pushes up electricity costs across the US power grid.

Earlier this week, Trump called on tech giants to “pay their own way,” arguing that households and small businesses should not bear the cost of power infrastructure needed to support energy-hungry data centers.

Microsoft quickly responded, saying it would “pay utility rates that are high enough to cover our electricity costs,” along with committing to other changes aimed at easing pressure on the grid. Other major tech companies are expected to follow suit, though Wedbush Securities analyst Dan Ives warned the added costs could slow the pace of data center build-outs.

As we’ve noted, forcing tech companies to shoulder higher electricity costs is likely to hit some firms harder than others. Companies like Microsoft, Google, and Amazon can pass at least some of those costs on to customers by selling data center capacity downstream. Meta, in contrast, does not have a cloud business, meaning its AI ambitions lack a direct revenue stream to offset rising power costs.

So far tech stocks don’t appear to be affected much in premarket trading. However utility companies most levered to the AI boom certainly are, with Vistra, Constellation Energy, and Talen Energy deep in the red ahead of the open as analysts at Jefferies warn that these firms face risks from this plan.

Trump Moves to Make Tech Giants Pay for Surging Power Costs

Jon Keegan1/15/26

OpenAI working to build a US supply chain for its hardware plans, including robots

When OpenAI purchased Jony Ive’s I/O, it entered the hardware business. The company is currently ramping up to produce a mysterious AI-powered gadget.

But OpenAI plans on making more than just consumer gadgets — it also plans on making data center hardware, and even robots.

Bloomberg reports that OpenAI has been on the hunt for US-based suppliers for silicon and motors for robotics, as well as cooling systems for data centers.

AI companies are looking toward robots as a logical next step for finding applications for their models.

OpenAI told Bloomberg that US companies building the AI brains of robots might have an edge against the Chinese hardware manufacturers that are currently making some impressive humanoid robots.

OpenAI Seeks US-Based Suppliers for Planned Robotics, AI Device Push