At #CES, Nvidia CEO Jensen Huang said Vera Rubin, the company's upcoming AI chip series, will deliver a 5X performance improvement over its previous GPUs. Get details about that — and why Nvidia signed a $20B licensing deal with startup Groq — in this edition of the Runtime newsletter.

As we reflected on the biggest stories of 2025, one throughline emerged. They all illustrate how enterprises grappled with a central problem: how to get the generative AI tools and agents that promise a new era of business productivity to actually work. www.runtime.news/five-importa...

Reporting from AWS re:Invent in Las Vegas, Tom Krazit sums up all you need to know after the first big day of the event, including AWS's custom AI chip, called Trainium 3 and now generally available.

After sifting through the fallout from a massive outage, it's clear that both AWS and its customers are still putting too many of their eggs into one very old and very worn basket. See why us-east-1 is the biggest problem in cloud computing.

F5 says a nation-state hacker gained long-term access to some of its servers. The hack comes as attacks on American software companies appear to be ramping up and as the Trump administration pulls staffers away from the Cybersecurity and Infrastructure Security Agency to focus on deportations.

If you want to make sure your business is able to deliver the high-performance throughput and low latency AI applications require, you'd better make sure your network is ready. Read why in this post, sponsored by @cisco.com www.runtime.news/why-future-p...

A hybrid cloud strategy solves a lot of problems, especially for companies in heavily regulated industries or that have big investments in data centers. But it also comes with its own set of challenges. We asked members of the Runtime Roundtable how they think about hybrid cloud app deployment.

New on @runtime.news: The latest in our How We Built It series, this time featuring Mastercard's George Maddaloni. We talked about how Mastercard onboarded generative AI tools across its employees and its current approach to AI agents, where MCP is spurring a lot of internal agent building.

Workday is one of many companies trying to convince enterprises to use its software as the foundation for their agentic AI strategies, and at Workday Rising it unveiled new tools that it hopes will close the deal. "The era of one-size-fits-all enterprise software is over," said CTO Peter Bailis.

Broadcom CEO Hock Tan says the future of the enterprise is private. Cloud provider earnings reports tell a different story. But for tech organizations going the hybrid cloud route, VMware unveiled several new ways it gives customers cloud-like performance capabilities inside their own data centers.

Voice-phishing (or "vishing") attacks are on the rise, with Workday being the latest company to disclose that it had been targeted by hackers. You can bet we'll all see more about this in our next corporate security training sessions.

Generative AI apps could help software designers break out of a decades-old rut when it comes to how we use their tools. At the very least, they need to rethink their approach to avoid frustrating users. Read about that, plus the latest moves by enterprise tech execs, in the Runtime newsletter.

Times have changed. Microsoft executives used to celebrate GitHub's independence within the larger company ecosystem. But that era is over, and today GitHub feels like just another Microsoft product.

OpenAI is living up to its name again. It hadn't shipped any major LLMs that could claim a degree of openness since 2019's GPT-2. But that just changed with the release of two new open-weight models the company says are as powerful as some of its leading closed models.

It's been a choppy few months for Windsurf. First, news leaked that OpenAI intended to acquire the company for $3 billion. Then Google came along with a billion-dollar acquihire, agreeing to pay $2.4 billion for a non-exclusive license to Windsurf's technology and hire about 40 of its employees.

Grok 4 could be a tempting choice for developers and enterprises looking for competitive but lower-cost alternatives to OpenAI and Anthropic. But it's impossible to understand why any serious business would put Grok at the heart of its AI strategy. Grok is unsafe at any speed for business use.

CoreWeave says it plans to acquire Core Scientific for $9 billion. The deal would allow CoreWeave to take direct control of infrastructure it's already renting from Core Scientific and will provide the additional capacity it needs to fulfill nearly $16 billion in orders from OpenAI.

This week, a pair of legends in the software engineering space shared their insights on the "Software 3.0 era" and laid out just how much disruption is coming for workers in the software mines. Plus, we bring you the latest in the Microsoft-OpenAI negotiations and two big executive moves.

The conscious uncoupling of Microsoft and OpenAI is sticky. We unpack where things stand now, and how some big factors — like OpenAI's pending acquisition of Windsurf and a slate of other powerful LLMs now on the market — are complicating things.

In this one-on-one interview, Amazon CSO Steve Schmidt tells us about: *Increasing nation-state attacks on tech companies *How generative AI is — and isn't — changing cybersecurity, on both the offensive and defensive sides *How Amazon works to prevent unforced errors in code

Last year, Databricks CEO Ali Ghodsi pledged to bring Delta Lake and Iceberg together in a "USB-C format" for data. The communities behind both formats have made progress toward that goal, but the last mile is tricky, and new tools could let end users worry about other problems.

From our roundup of the week's biggest product releases and updates: -A new open source project from Red Hat -A catalog of container images from Docker -Google Cloud's Gemini for Code Assist and for GitHub -Messaging directly with AI agents in Slack -Glean's Agents tool

Anthropic unveiled its latest Claude models, saying Claude Opus 4 "is our most powerful model yet and the best coding model in the world." The big breakthrough seems to be an ability to complete tasks unsupervised over long periods of time, compared to earlier versions that glitched out pretty fast.

At the Build conference, Microsoft said it will release the GitHub Copilot code extension inside Visual Studio Code under an open-source license and add a new coding agent to GitHub Copilot. Also, Qualcomm wants in on the lucrative server processor market. And see where two big funding rounds went.

Don't miss our regular roundup of the week's most important new enterprise tech product releases and updates, including: -Coding agents from OpenAI and Google -AWS's Transform service -Updates to Celonis's Process Intelligence API -Boomi AgentStudio

Jira is the bug-tracking system software developers love to hate. We sat down with the CEO of Linear, which has raised over $50 million to build a modern alternative, to talk about product strategy, generative AI, and AI agents. Also: Good news and bad news for CoreWeave, and four big exec moves

SBOMs are a great idea in theory. But are they really helping companies? We talk to Chainguard CEO Dan Lorenc about one key problem with them, as well as other challenges in the era of AI code. Plus: Project Stargate is behind schedule, and we clock six big rounds of enterprise tech funding.

We sat down with Freshworks CEO Dennis Woodside after a strong earnings report. For years, Freshworks has provided help desk and customer-service software to small and medium-size businesses, but Woodside is trying to take on bigger players like ServiceNow by moving into the enterprise market.

9 executives give their insight into how companies rushing to deploy GenAI apps can ensure what they're shipping is secure while still innovating quickly.

After a week of back-and-forth between Synadia and the CNCF, it appears the company will relinquish the NATS trademark after all. Here's how the unusual conflict over an open-source project went down.

Looking at early earnings reports from the biggest enterprise tech companies, it's clear that some are better positioned than others to weather the chaos of Trump's tariffs and Elon Musk's federal government cuts. Also: Synadia wants to take back ownership of a project it donated to the CNCF.

A year ago, Microsoft vowed to make big changes in response to a breach of its cloud services that enabled hackers to access federal government accounts. See what it's done. Meanwhile, in another sign that the AI boom is waning, AWS is taking its foot off the gas with data-center capacity planning.

Storage services are one of the most fundamental parts of any cloud architecture, and they're changing in response to changing AI workloads. Also: Due to federal funding cuts, the nonprofit that maintains the widely used CVE system for tracking cybersecurity issues says it can't do that anymore. 😐

Dispatch from Las Vegas, where Google Cloud is reintroducing itself in the enterprise AI space

If agents are the future of enterprise software, they will forever change the way business software is bought and sold. Zendesk CEO Tom Eggemeier tells us why he thinks customers should only pay for agentic activity that actually gets something done.

The most vocal proponents of the AI boom have said for years that the tipping point for enterprise AI will arrive when the cost of inference comes down, which is a logical progression that other tech breakthroughs have followed. The Trump tariffs just threw a wrench into that process.

Neo4j CEO Emil Eifrem says he sees a recurring pattern in how customers use the database. Due to frustration with SaaS spending paired with increased productivity of in-house engineering teams, he says, "We're moving into—in this classic build-versus-buy (dilemma)—a build era for the enterprise."

Tines, co-founded and led by Eoin Hinchy, raised $125 million in new funding last month. Tines began when Hinchey, a former security team leader at DocuSign, wanted to automate some of the most repetitive daily tasks his group encountered and started asking friends about their own workflows.

There is no shortage of hype around agentic AI. But a new protocol — Anthropic's Model Context Protocol, introduced in November — is gaining momentum as a vendor-neutral way to make AI agents much easier to implement.

Google Cloud's $32 billion deal for Wiz is its biggest acquisition bid ever. The cloud security platform, designed to work across all major cloud providers, will test the company's commitment to working with competitors, and you can bet Wiz customers running on AWS and Azure are watching closely.

When it comes to AI, don't let perfection be the enemy of the good. The key is to launch, get feedback, and iterate, the same way software developers have always improved their apps. That was the message from a panel of experts at the #HumanX conference this week.

Apps built for or around generative AI have different hardware and software requirements than their older cousins. At the #HumanX conference, we talked to four tech leaders who gave their advice about what to do and what not to do when drawing up infrastructure plans for AI.

CoreWeave, one of the biggest winners of the generative AI frenzy, says it's planning to go public. Here's what we learned from the company's S-1. Plus, the six biggest rounds of enterprise tech funding from the past week. www.runtime.news/so-you-wanna...

New on @runtime.news: Pinecone just announced an overhaul of its serverless managed vector database infrastructure that focuses on balancing speed, performance, and freshness without making users fiddle with a bunch of dials. My interview with founder and CEO Edo Liberty:

• Making sense of AWS's earnings • The code and training data for a new LLM—which researchers at Stanford, UW, and AI2 say rivals OpenAI and DeepSeek—is up on GitHub • Plus, your skimmable roundup of the biggest C-level moves and other enterprise tech news of the last few days

Exclusive: Vercel's new Fluid Compute is designed to make it cheaper for customers to run AI apps. "This is the natural evolution of serverless computing."

It's our weekly roundup of the product news you need to know: new AI models meant to compete with DeepSeek; Microsoft DocumentDB is back-ish; developers have new inference computing options on Hugging Face; and Puppet users delivered OpenVox, a proper open-source fork of old-school Puppet.

This has been a hell of a week in the already frenetic AI industry. You can be forgiven if you missed the whiplash-inducing news cycle about Chinese startup DeepSeek sending Nvidia stock into a historic spiral, thanks to some novel engineering techniques. We broke down what happened...

A new AI coding assistant from JetBrains, a major update from TigerEye, updates to the Qdrant vector database that can tap into the power of GPUs, the launch of the new Sonar API service, etc.: Get those details and more news from the week in our weekend product roundup.

Yesterday OpenAI introduced Operator, a computer-use agent that functions as a website users can visit and enter a prompt to have it execute a task, like filling out a form or ordering groceries. We're curious: What fields or use cases would you want to see tools like this applied to?