AWS vows to "push the limits" of AI infrastructure

Welcome to Runtime! Today: AWS introduces new infrastructure tools and services during the first big day of re:Invent 2025, ServiceNow gets deeper into the security business, and the latest funding rounds in enterprise tech.

(Please forward this email to a friend or colleague! And if it was forwarded to you, sign up here to get Runtime each week.)

Table stakes

LAS VEGAS — Three years ago at AWS's annual re:Invent conference, which took place days after the 2022 launch of ChatGPT, generative AI was not on the agenda. After making up for lost time by cramming AI into every single part of the 2023 re:Invent experience, the cloud leader found a little more balance last year under new CEO Matt Garman, and this year Garman used his opening keynote to argue that the main reason for AWS's success — the powerful and resilient infrastructure hardware and software services it has built over the years — make it the best home for a new class of enterprise applications.

Like everyone else in enterprise tech, Garman and AWS believe that despite their lack of progress to date, AI agents are going to change the way enterprise software is built, delivered, and consumed. "Getting to a future of billions of agents, where every organization is getting real-world value and results from AI, is going to require us to push the limits of what's possible with the infrastructure," he said Tuesday.

Trainium 3, AWS's custom AI chip, is now generally available and "offers the industry's best [ratio of] price/performance for large-scale AI training and inference," Garman said.
Technically AWS offers two AI chips, Trainium and the inference-oriented Inferentia, but it appears to have forgotten about Inferentia; AWS executives dodged questions about the future of the chip during a press conference on Monday.
Trainium can't yet match the performance of Nvidia's top-of-the-line GPUs, but is cheaper and easier to procure.
And development of Trainium 4 is already underway, with plans to add Nvidia's NVLink technology into the design when it arrives.

Google Cloud has also been working on custom AI chips for a very long time, and Microsoft also recently got into that game in hopes of denting at least some of Nvidia's hold on that market. But delivering applications built around AI goes way beyond the chip itself to include the servers, networking, and storage systems that are the foundation of any cloud infrastructure service, and that's where AWS thinks it still has an advantage over its faster-growing rivals.

Customers will now be able to run AWS hardware and software inside their own data centers with the launch of a service called AWS AI Factories, which builds upon similar services from previous years like Local Zones and Outposts.
"AWS AI Factories operate like a private AWS region, letting customers leverage their own data center space and power capacity that they've already acquired," Garman said.
Customers will also be able to run Nvidia GPUs on this data-center-in-a-box service, which was designed with government agencies and large organizations like utilities in mind, said Nishant Mehta, vice president of product management for EC2, in an interview with Runtime ahead of re:Invent 2025.

AWS took a little pressure off Tuesday's keynote after releasing its third-quarter earnings results in October, which showed it had returned to a growth rate it hadn't seen since the launch of ChatGPT. While it's clear that Microsoft and Google enjoyed a great deal of success at its expense over the last few years thanks to earlier and more aggressive investments in generative AI technology, AWS is still the center of enterprise tech.

The hype around generative AI and agents is starting to give way to reality; these new tools are very powerful in certain applications and systems, but are far more difficult to implement than initially thought and unreliable for a wide variety of enterprise tasks.
That means companies will adopt them at a more measured pace than a lot of people thought in 2023, which has given AWS time to figure out how to adapt to this shift.
And it seems to have concluded that the best way to adapt to enterprise generative AI is to do what it has always done: provide foundational services that help companies adopt and exploit a new technology.
Stay tuned for a deeper look at AWS's software ambitions in Thursday's Runtime.

If you value independent enterprise tech journalism from big events like AWS re:Invent, please consider becoming a monthly supporter of Runtime. For $10 a month, you'll help us continue our mission to bring reliable and actionable coverage of this vital sector of the economy and gain access to supporter-only features currently in the works, such as an exclusive discussion and events forum.

Become a Runtime Supporter

ID, please

Identity management — ensuring the people and devices on your network are who they say they are — is one of the most critical parts of corporate cybersecurity, and it is rapidly changing as companies start to roll out agents that will need different types of verification. ServiceNow has been wandering into markets beyond its core IT service management software for several years, and announced Tuesday that it will make a bigger splash in the security market with the purchase of Veza.

Veza's Access Graph provides security administrators with a visual representation of activity across their infrastructure and has been growing strongly since its 2020 launch, according to CRN. The plan is to integrate that graph into ServiceNow's AI Control Tower, which allows companies to visualize their AI assets, according to a press release.

The deal is just another sign that the "best of breed" era of enterprise software — in which businesses preferred to buy a focused software tool for a specific use case, like identity security — is ending as budget-conscious buyers prefer vendors who can provide a more comprehensive package of enterprise software, just like in the good (or bad, depending on your perspective) old days. It always was a terrible phrase.

Enterprise funding

X-energy raised $700 million in Series D funding to build out its small nuclear reactors, which a lot of data-center operators think could help solve the energy crunch.

Black Forest Labs landed $300 million in Series B funding for its image generation frontier models.

Eon scored $300 million in Series D funding as it builds a cloud data backup system that promises to "[turn] that dormant data from both IaaS and PaaS services into a live, browsable, searchable and queryable foundation for AI training and analytics."

Ricursive Intelligence launched with $35 million in seed funding to use AI tools to automate chip design, led by the two researchers who developed the AlphaChip model at Google DeepMind.

Momentic raised $15 million in Series A funding for its QA software, which allows customers to use natural-language prompts to write tests.

CoPlane landed $14 million in seed funding as it builds out an AI platform for ERP software.