Apple contains all the things; AMD's new GPUs

Welcome to Runtime! Today on Product Saturday: Apple introduces a new way to build containers on Macs, AMD renews its pursuit of Nvidia's GPU lead, and the quote of the week.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Ship it

Return of the Mac: Linux containers are the backbone of countless enterprise applications used throughout the world, and lots of developers build those containers on Apple's Macs before deploying them to production. Docker Desktop has long been a popular way of constructing those containers on laptops, but this week Apple introduced a new way for developers to build those containers on Macs using its M series chips.

Apple Containerization Framework was "built on an open-source framework optimized for Apple silicon and provides secure isolation between container images," Apple said in a press release. Alex Zenla, founder and CTO of Edera, called it "one of the most significant infrastructure announcements Apple has made in years," noting that Apple's approach to isolate containers from the hypervisor — a desktop version of what Edera is building on the server side — is far more secure than previous alternatives.

Come together: Last year at the Databricks Data & AI Summit CEO Ali Ghodsi vowed to unify the two leading open table formats — Delta Lake and Iceberg — after acquiring Tabular, and while full compatibility has yet to arrive the communities developing those formats made progress over the past year. This year, Databricks announced that users of its Unity Catalog will be able to write and read data in either format through new support for the Iceberg REST Catalog APIs.

Most data already stored in open formats was developed for Databricks' Delta Lake, but the data industry has embraced Apache Iceberg and most future tools will likely be designed around that approach. As work on merging the formats continues, both Databricks and Snowflake are addressing compatibility with their catalogs, which help data query tools understand where and how the data they need is stored.

A major development: After watching Intel dominate the cloud computing era with overwhelming market share, enterprise tech buyers would love to see someone make a dent in Nvidia's overwhelming market share for AI chips this time around. AMD might be in the best position to pull that off, and this week it introduced a new GPU designed to take Nvidia's latest Blackwell chips head-on.

The Instinct MI350 series will deliver a 4X improvement in performance compared to AMD's previous-generation GPUs, and the most powerful chip in the lineup "delivers significant price-performance gains, generating up to 40% more tokens-per-dollar compared to competing solutions," the company said in a press release. Meta, Microsoft, OpenAI, and Oracle all announced plans to adopt the new chips, and if the real-world benchmarks live up to those showcased during the presentation, expect others to follow.

Agents like to network: As generative AI changes the way software is built, it's also having a huge impact on security and IT operations teams by helping them make sense of an ever-expanding number of alerts and incidents. Cisco introduced several new AI services this week across a platform it is calling AgenticOps, which was "designed to transform real-time telemetry, automation, and deep domain expertise into intelligent, end-to-end actions," it said in a blog post.

Cisco AI Canvas is a new dashboard that in response to an incident, "pulls in the right data, connects the dots, and surfaces a live picture of what matters—before anyone even asks," it said. Like almost every agentic AI platform introduced over the last year, Cisco is hoping to convince current customers to allow it to control their AI agents before another vendor pulls them in.

Platform play: The pivot away from the so-called "best of breed" enterprise software stack has been underway for several years, but tighter IT budgets and the flexibility offered by generative AI tools is bringing the integrated suite back in a big way. Companies like ServiceNow that built a business around a particular niche are expanding into other sectors of enterprise software, and this week observability giant Datadog signaled its intentions to go after ServiceNow.

This week Datadog unveiled an Internal Developer Portal, which was designed to help "developers to build, test, deploy and monitor software with self-service actions and built-in delivery guardrails, while providing platform engineers with scorecards to help them meet reliability, security and monitoring standards," the company said in a press release. "All of these platform organizations are already overlapping, and right now they seem to be working and playing nice together, but at some point, markets will push one to try to squeeze the others out," Forrester Research's Carlos Casanova told TechTarget.

Stat of the week

You might have heard that folks are excited about the possibility that generative AI could transform their businesses (if you haven't, please share your strategy for blocking it out), but just how much are they spending to make that happen? According to a new report from Rackspace, the average business surveyed expects to spend $8.7 million on AI this year, a 250% increase compared to last year that is "evenly divided between the development of in-house solutions and the adoption of third-party products."

Quote of the week

"This is an over $100 billion [total addressable market] for us, so it is a big opportunity. I think it's going to be a journey, but I think it's going to be the most important marathon that Databricks runs over the next five to ten years." — Databricks CEO Ali Ghodsi, discussing the company's decision to acquire Neon for $1 billion and evangelize the "lakebase" category.

The Runtime roundup

Google released a report about what went wrong during Thursday's massive outage, attributing its fundamental role in the debacle to an update added in late May that lacked "appropriate error handling nor was it feature flag protected," and a change rolled out Thursday morning caused a crash loop that could have been prevented with proper error handling.

Meta will invest $14.5 billion in ScaleAI while poaching CEO Alexandr Wang for a senior AI role, according to Bloomberg, which could present a problem for ScaleAI's attempts to work with other model developers on data labeling.

Thanks for reading — see you Tuesday!