Think globally, compute locally

Welcome to Runtime! Today: why location will really matter over the next decade of cloud computing, the Snowflake/Databricks rivalry boils over, and this week in enterprise tech funding.

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

A cloud in every port

Cloud providers and data-center operators have been opening regions around the world for a very long time, but thanks to increasing regulation and the use of edge computing to power AI research and development, they're going to have to be more precise than ever when making decisions about capacity expansion.

Announcements Tuesday from Microsoft and TikTok (imagine if Microsoft actually bought them three years ago) underscore that regional approaches may no longer work as we approach the third decade of cloud computing. For a variety of reasons, in the future data will need to be stored much closer to the end user and delivered in more creative ways.

Microsoft, for example, announced plans to expand in the United Arab Emirates through an existing partnership with G42, which is run by the country's national security adviser.

• Microsoft is rather pointedly not building a "region" in the UAE, which would theoretically be available to anyone, like Azure capacity it plans to open next door in Saudi Arabia.
• It will, however, build "sovereign cloud" capacity in the UAE, a term used to describe cloud computing resources that keep all the data inside a country's physical borders.
• Sovereign clouds have grown increasingly popular thanks to data residency laws coming on the books around the world, as regulators hope to preserve some local control over an industry utterly dominated by U.S. companies.

TikTok's story is, of course, a little more complicated. After several delays TikTok opened its first European data center Tuesday in Ireland, and has begun the process of moving customer data into the facility.

• TikTok has been forced to localize its customer data thanks to concerns that its owner, China's ByteDance, could obtain a lot of data about TikTok users they could also get from a data broker because U.S. data laws are kind of silly.
• The company has acknowledged, however, that ByteDance employees have been able to access customer data from China, and it has agreed to a series of privacy-oriented measures and data localization plans.
• NCC Group, a U.K. cybersecurity company, was chosen Tuesday to audit TikTok's Irish operations and make sure European customer data can't be accessed by ByteDance.

Working around data-residency or sovereign cloud regulations is annoying but doable, and the generative AI boom is showing that those efforts could have substantial upside.

• A lot of companies working on AI-enhanced applications want to train and process data at the edge of their networks, rather than sending back and forth between the end user and a central data center.
• There are several reasons for this shift, including performance and cost, and while centralized cloud computing isn't going anywhere there's now a lot more demand for smarter edge-computing installations.
• As companies build computing infrastructure in more countries thanks to those localization laws, they'll also be able to deliver AI services to customers in those countries much more efficiently.

The downside of all this localization will be the usual culprit: Increased complexity.

• Scaling an application across borders will be more difficult, and in some cases will no longer be possible.
• But if there's one thing enterprise tech buyers can count on, it's the emergence of a parade of vendors to solve the complexity they created yesterday, today.

The data nerds are fighting

Thanks to Instacart, the Snowflake vs. Databricks rivalry has officially become the dumbest yet most reliably entertaining beef in enterprise tech.

As noted last week in Runtime, Instacart's S-1 appeared to reveal a sharp drop in its annual spending on Snowflake's software, and according to CNBC the Databricks fanboys wasted no time trash talking their rival across the fragmented social media landscape we now occupy. Snowflake clarified that one of the figures cited in that trash talk only covered the first half of 2023, but not before its supporters fired their own social-media barrage at Databricks employees.

"Many of the posts on Reddit, LinkedIn and X, the site formerly known as Twitter, have since been deleted," CNBC wryly noted Saturday. It's understandable why two companies that compete head to head for such a lucrative and growing sector of tech don't like each other, and this animosity has been going on for a while: in 2021 they got into a dispute over benchmark results and last year argued on social media over competing open-source software projects.

Enterprise funding

AI21 Labs raised $155 million at a $1.4 billion valuation, as its large-language models increasingly compete with some of the biggest companies in tech.

Mujin landed $85 million in Series C funding to meet growing demand for software that can control industrial robots.

Upwind raised $50 million at a $300 million valuation from a variety of investors including Steph Curry of the Golden State Warriors.

Context.ai launched with $3.5 million in seed funding from Google Ventures and Tomaz Tunguz's Theory Ventures, with plans to develop an analytics tool for AI applications.

The Runtime roundup

Zoom introduced several new AI features and rebranded Zoom IQ to AI Companion one month after it was at the center of a controversy around how it does, or doesn't, use customer data to train its AI services.

At the same time, Zoom CEO Eric Yuan, a MIcrosoft competitor, suggested that U.S. regulators should take a look at Microsoft's strategy of bundling Teams with Office after it voluntarily made changes in hopes of appeasing European regulators.

GitHub went down for about half an hour Tuesday morning West Coast time after its automatic failover procedures failed to respond to "an outage in one of our primary databases," the company said.

Rubrik is preparing the final stages of an IPO filing that could come this year, Bloomberg reported.

Thanks for reading — see you Thursday!