At re:Invent, AWS strikes back in AI with new chips, models – and a story – News

This week was all about AWS re:Invent, the biggest cloud conference of the year — which also makes it by definition, these days, one of the biggest AI conferences of the year.

The headlines from our 20 or so stories plus more interviews from theCUBE tell the story. But the gist is that if CEO Matt Garman (pictured) didn’t knock everyone’s socks off, he made a pretty good case that Amazon’s in solid shape for seizing the AI and especially agent opportunity in multiple ways — especially when it comes to who’s eventually going to spend the biggest bucks: enterprises.

The massive crowds at the Venetian suggest a lot of customers are game. Or at least that they’re looking to cash in on everything AI and they figure AWS is a prime place to do that.

And Amazon isn’t the only empire to strike back on the AI front. A surging Google apparently has put a scare into OpenAI CEO Sam Altman, who called a “code red” to try to keep his company in the AI lead.

So Netflix finally did it: It’s acquiring Warner Bros. for $82.7 billion, wresting the deal away from David Ellison and Paramount. If exclusive deal talks work out and it passes antitrust muster, no sure thing with Trump in the White House, that’s a huge deal with huge implications for all of media.

Databricks is in talks to raise a few billion dollars more, in what feels like, what, a Series X round? Why is CEO Ali Ghodsi so allergic to going public? Anthropic CEO Dario Amodei certainly doesn’t seem to be, as it reportedly hired a lawyer to get ready.

In a mixed bag for enterprise earnings, Salesforce’s growth slowed but its AI efforts gained some steam, while Snowflake slipped on slowing sales as well. HPE fell as customer delays buying AI servers (!) also slowed sales. But UiPath gained momentum on its agentic AI management push and MongoDB rallied big on sales momentum following several new AI initiatives.

Next week Oracle and Broadcom earnings will offer more insight into the state of enterprise software and hardware.

Here’s all the enterprise and emerging tech news from News and beyond:

Special Report: News and analysis from AWS re:Invent

Deeper dives:

Exclusive: AWS CEO Matt Garman declares a new era: Agents are the new cloud AWS is prepping a systematic rearchitecture of the cloud for a world where billions of agents operate across industries.

‘Why not?’ At re:Invent, AWS answers with big step into frontier AI model reasoning and agentic services

Build without limits: AWS outlines an easier path for agentic AI deployment

Agentic infrastructure takes center stage in AWS’ evolving cloud strategy: theCUBE keynote analysis from re:Invent

Five thoughts from CEO Matt Garman’s keynote at AWS re:Invent AWS preps itself for the new age of work in which people and agents work together.

Amazon CTO Werner Vogels foresees rise of the ‘renaissance developer’ in his final keynote at AWS re:Invent

AWS showcases AI factories, models, chips and agents to drive its core infrastructure strategy

AWS broadens its approach to agentic development

A few thoughts from five keynotes and a dozen interviews with AWS executives, engineers, scientists and customers:

* Garman made a strong case that it can help “industrialize” AI and agents in enterprises as well as or better than anyone. The emphasis I heard over and over from him and other execs is how AWS’ big advantage is owning the whole infrastructure stack for AI, so it can optimize it in multiple ways. “We’re really well-positioned to do any AI workloads,” Rahul Kulkarni, GM of product management for AWS compute and AI infrastructure, told me. “We own the stack from the custom silicon layer to the rack. We have fine-grained control of the entire stack.” Of course, Google could say something similar, but AWS remains the largest cloud with the tools and massive customer signal, so it’s not a stretch when Garman claims it’s the best place to do AI.

* At the base layer, the chips, new and forthcoming Trainium AI chips and Graviton processors, as well as its Nitro chips for offloading networking and security from those processors, remain a core focus. They’re popular with AWS customers, even if there’s no AWS rockstar like Nvidia CEO Jensen Huang to whip up outside enthusiasm, but it’s also true that they lack some of the deep software libraries that make Nvidia GPUs so dominant.

* Even if AWS’ Nova models, several new versions of which it debuted at re:Invent, aren’t yet threatening OpenAI or Anthropic, researchers are doing interesting work to make AI models better and more reliable. Byron Cook, a VP and distinguished scientist at AWS, told me about several ways it has incorporated strategies such as automated reasoning and neurosymbolic AI into AWS products to reduce not only hallucinations but all manner of erroneous model behavior by using mathematical proofs. Hallucinations are great for surfacing new ideas, he noted, so they’re something of a feature as much as a bug. “If it’s a poem, they’re great,” he said. “But if it’s ‘will this building stand or not stand,’” not so much.

* Nova Forge, the new service that allows companies to build their own custom models, is more important than the relatively scant coverage indicates. AWS is betting, sensibly, that enterprises will want their own custom AI models, fed by their own proprietary data. There won’t be one model to rule them all, and the more there are, the more there are to run on AWS and the more customers need cloud services such as Amazon’s vast array to create, manage and run them.

* Has AWS proved it’s “back” in AI? Not yet. Nova models still aren’t going to sway many people from OpenAI and Anthropic anytime soon. Some products such as Bedrock and Trainium, and especially its Nova models, have not yet proved they have caught up to Google’s and OpenAI’s services and models in usage or performance. 

* Agents, with the need for many tools and orchestration software to make them work beyond proofs-of-concept, may provide a new opening for AWS. As Garman put it: “This is where you’re starting to see material returns in your investments. It’s turning from a technical wonder to something that’s producing real business.”

* To their credit, execs admit that it’s not certain how, and how fast agents will change the way software, and therefore work, is done — because customers aren’t yet sure either. “The reality is it’s not going to happen as fast as people want it to,” Jeff Hammond, a former Forrester analyst who is head of independent software vendor product management transformation, told me. “We’re all figuring this out along the way.” Or as Peter DeSantis, tk, put it in an echo of Amazon founder Jeff Bezos: “It’s still day one in AI.”

* But the bottom line is that it’s clear Amazon is fully engaged on the AI opportunity now, working hard on upgrading everything from chips to models to developer tools. And as Google recently has shown, there’s room and time to catch up to the leaders, especially for the leading cloud provider where a large chunk of customers’ data and applications already resides.

All the big news:

AWS introduces Nova Forge for training bespoke ‘Novella’ frontier models Enterprises would love and perhaps require custom AI models for all kinds of processes, based on their own data and workflows. Although AWS folks admit there’s a steep learning curve here, it seems like a smart bet.

AWS expands Nova foundation models, adds multimodal support

AWS simplifies AI agent customization with automated reinforcement learning

AWS puts Kiro and other AI agents to work on truly autonomous software development

Amazon introduces Kiro powers for cleaner, context-aware AI-driven development

AWS brings sovereign AI on-prem with new AI Factories alongside Trainium3 and Nvidia GB300 launches

AWS debuts Graviton5, a powerful and efficient custom CPU for cloud workloads

Amazon rolls out Nova Act for AI browser agents, combined with 1Password credential security

Nvidia and AWS expand partnership with specialized AI hardware and software

AWS and Google Cloud partner on faster multicloud connectivity

AWS unveils EKS capabilities to reinvent Kubernetes operations as AI workloads surge

AWS rolls out Security Agent and strengthens GuardDuty and Security Hub at re:Invent 2025

Salt Security launches ‘Ask Pepper AI’ to simplify API risk analysis

Skyhawk Security adds agentic AI to validate cloud defenses in real time

And our many unsponsored editorial interviews at the event on theCUBE:

New AWS releases show how agentic agents are entering real operations

Nova Forge aims to close AI’s domain knowledge gap with customizable frontier models

AWS bets on S3 Vectors as AI agents reshape the demands of storage

Sovereign cloud emerges as a defining force in modern AI adoption, says Greylock’s Jerry Chen

Beyond bug detection: Inside the rise of self-fixing developer tools

Real-time AI emerges as artificial intelligence hits a new performance threshold

More stores are ditching checkout lines as AWS’ Just Walk Out tech gains steam

AI agents reshape expectations for data connectivity as Salesforce expands its ecosystem strategy

AWS and SentinelOne double down on security as AI agents widen the attack surface

For AI-native applications, scalability — not just latency — is the true enterprise challenge

Data-centric architectures emerge as the new battleground in intelligent automation

From lift-and-shift to one-and-done: Inside AWS’ agentic cloud modernization push

AWS bets on agentic AI to power the next-gen contact center

The AI boom is pushing companies toward deeper multicloud integration

AI complexity fuels new demand for deeper observability frameworks

More to come…

AI and data: Suddenly AI looks wide open again

Analysis and food for thought

OpenAI CEO Sam Altman reportedly declares ‘code red’ after rivals launch new AI models

Microsoft lowers AI software sales quota as customers resist new products, The Information reports Hmm. But Microsoft pushed back.

Studies find AI chatbots have the persuasive power to change political opinions

New York Times escalates battle against Perplexity with new lawsuit

Google sees 3.2 times more pages on the internet than OpenAI, 4.6 times more than Microsoft, and 4.8 times more than Anthropic or Meta does, says Cloudflare CEO Matthew Prince (per Wired)

Money matters

Databricks reportedly in talks to raise $5B at $134B valuation

OpenAI acquires AI tooling provider Neptune to enhance its model training workflows

Anthropic reportedly preparing for one of the largest IPOs ever in race

Meta acquires Sam Altman-backed AI wearable startup Limitless

Open-source image generator startup Black Forest Labs raises $300M

Anthropic makes first acquisition with purchase of Bun to accelerate Claude Code

AI-focused legal startup Harvey raises $160M to expand platform capabilities

Audio language model startup Gradium raises $70M to create more realistic voice AI systems

Vinci nabs $46M to speed up chip simulations with AI

Ricursive Intelligence launches frontier AI lab with $35M funding to transform semiconductor design

Lemurian Labs raises $28M for its efficiency-optimized AI model compiler

Pine secures $25M in Series A funding to free consumers of digital chores

Retail operations automation startup Duvo raises $15M

AI startup Vambe raises $14M to automate conversational commerce

Unlimited Industries raises $12M to transform infrastructure construction

Curvestone AI raises $4M to bring reliable agentic automation to regulated industries

Soxton emerges with $2.5M to build lawyer-in-the-loop AI for startup legal work

New models and services

Runway, DeepSeek release new foundation models

Google launches Workspace Studio to let users build AI agents for everyday work

Pinecone scales its vector database to support more demanding workloads

Stack Overflow launches AI Assist chatbot for developers

TurinTech launches preview of Artemis, a structured AI software engineering platform 

Salesforce targets nonprofits with its agentic enterprise vision

Identity verification startup Incode launches Deepsight to detect deepfake video calls

Prophecy accelerates data pipeline construction with quick-start AI agents

Prior Labs debuts tabular AI foundation model that scales to 10 million rows

Fortanix joins HPE and Nvidia to advance confidential enterprise AI

How Atlassian’s System of Work has become the engine for innovation at Williams Racing

Around the enterprise: Chipping away

Analysis

TPUv7: Google takes a swing at the king SemiAnalysis’ take on how Nvidia has yet another contender to the throne besides AMD. And that’s not even counting Amazon’s increasingly capable Trainiums and Gravitons.

Money matters

Marvell bets big on optical interconnects, buying Celestial AI for up to $5.5B

Intel scraps plan to spin off NEX networking chip business

Nvidia invests $2B in Synopsys as part of multiyear AI collaboration

Cloud backup startup Eon raises $300M at $4B valuation

Axiado raises $100 million for chip to save space, power in AI data centers

David Sacks, Silicon Valley’s man in the White House, is benefiting himself and his friends (per the New York Times) Grifters all.

Akamai acquires WebAssembly function-as-a-service startup Fermyon

Earnings:

Salesforce’s AI bets gain momentum, but overall growth remains sluggish

Snowflake’s stock slips on slowing sales as it expands Anthropic deal

UiPath beats expectations as it doubles down on agentic AI orchestration

MongoDB rallies on better-than-expected earnings and raised full-year outlook

HPE misses revenue targets as AI server sales slide on customer delays

Shares of Pure Storage sink after earnings results fail to impress

Cyber providers CrowdStrike and Okta post solid quarters but fail to impress investors

Rubrik shares surge on strong earnings and outlook as SentinelOne slides on weak guidance

Box’s Q3 revenue rises 9% while earnings fall 31%, lifting stock

Disappointing guidance tanks GitLab shares

Asana shares rise after strong Q3 earnings and upgraded forecast

C3 AI’s earnings beat expectations, stock rises

Samsara beats earnings expectations with 29% revenue growth, shares slip

Docusign stock slides as earnings beat fails to impress

Domo shares tumble as earnings results fail to impress

New products and services

HPE broadly expands AI-native networking, hybrid cloud and storage offerings

Aruba meets Juniper Mist: At Discover, HPE unveils Its unified AI-native network brain

Artificial Superintelligence Alliance members launch first cloud GPU cluster to power decentralized AI

Cyber beat: ServiceNow moves deeper into security

Money matters

ServiceNow to acquire Veza in move to expand identity security portfolio

7AI raises $130M to automate cybersecurity investigations with AI agents

Zafran Security nabs $60M for its vulnerability management platform

Imper.ai launches with $28M to stop AI-powered impersonation attacks

Lumia Security raises $18M to bring governance and control to enterprise AI agents

Helmet Security lands $9M to protect enterprises from MCP-based AI security risks

New services

Box launches Shield Pro with agentic AI for automated content security

Netskope adds MCP security controls to protect enterprise AI agents

Sumo Logic expands Dojo AI with new agentic tools for modern security operations

SandboxAQ launches new service to tackle shadow AI security risks

Hack The Box debuts HTB AI Range to test AI and human cyber defense side by side

Movius launches CoreLine to embed secure communications directly into enterprise SIM devices

Elsewhere in tech: Netflix’s blockbuster Warner Bros. deal

Netflix to acquire key Warner Bros. assets in blockbuster $72B deal

SpaceX reportedly seeking $800B valuation ahead of 2026 IPO

Security device startup Verkada raises funding at $5.8B valuation

Tutor Intelligence raises $34M to scale fleet of AI-powered warehouse robots

Comings and goings: Musical chairs at Apple

Meta reportedly could let go 30% of its metaverse unit’s workforce

Apple named former Microsoft, Google executive Amar Subramanya to replace current AI chief John Giannandrea.

Apple’s longtime design head Alan Dye joined Meta to lead new creative studio.

Chief revenue officers seem to be in high demand, as Smartsheet tapped Scott Torrey and OutSystems hired Kim Seabrook.

Chief marketing officers too: Joyce Kim headed to Proofpoint and Ryan Shopp landed at Camunda.

Oh, and chief financial officers: Mobile security firm Zimperium appointed Alistaire Davidson CFO and GitLab named Jessica Ross to the same job.

Orca Security named former Check Point Software VP Gera Dorfman chief product officer.

What’s next

Earnings

Wednesday, Dec. 10: Oracle, Adobe, Synopsys

Thursday, Dec. 11: Broadcom

Photos: Robert Hof/ News