Why hasn't LinkedIn moved to Azure?

Today: turns out cloud migrations are hard, even when you work for the cloud company, Intel tries again to compete in the AI chip market, and the latest moves in enterprise tech.

Why hasn't LinkedIn moved to Azure?
Photo by Greg Bulla / Unsplash

Welcome to Runtime! Today: turns out cloud migrations are hard, even when you work for the cloud company, Intel tries again to compete in the AI chip market, and the latest moves in enterprise tech.

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


Microsoft has done a remarkable job over the last decade transforming itself from an old-guard enterprise IT provider to a modern cloud company, a journey that few of its 40-something-year-old enterprise tech peers have pulled off. So why is one of its crown jewels still doing things the old-fashioned way?

That's just one question raised by CNBC's report Thursday that LinkedIn's migration to Microsoft Azure, which was first announced in 2019, has been indefinitely suspended, just like Draymond Green. LinkedIn is even building a new data center it plans to operate to handle the increasing demand for its services, according to the report.

  • "With the incredible member and business growth we’re seeing, we’ve decided to begin a multi-year migration of all LinkedIn workloads to the public cloud," LinkedIn said in a 2019 blog post.
  • However, last year it began telling employees that it no longer planned to move everything to Azure, although it sounds like a fair amount of workloads did wind up in its dad's cloud.
  • "With the incredible demand Azure is seeing and the growth of our platform, we’ve decided to pause our planned migration of LinkedIn to allocate resources to external Azure customers," a LinkedIn executive said in an email viewed by CNBC.
  • That is an interesting statement about Microsoft and the cloud in general, for reasons we'll shortly get into.

LinkedIn isn't exactly an infrastructure neophyte. People working there prior to its acquisition by Microsoft developed a number of now widely used enterprise technologies.

  • The most famous example is Kafka, the streaming-data technology that proved very popular across enterprise tech and launched Confluent, which is worth about $8 billion as of the close of business Thursday.
  • But as my former colleague Derrick Harris unpacked in a 2015 interview with Confluent co-founder Jay Kreps, LinkedIn has a proud history as an enterprise tech innovator across several different infrastructure disciplines.
  • Unique architectures can be more difficult to move to the cloud, which basically requires that customers rewrite their existing applications with cloud services in mind, as Deloitte's Nishita Henry said in our recent interview.
  • And as the old saying goes, if it ain't broke, don't fix it.

But LinkedIn's decision to hug its servers is pretty remarkable, given its role within a conglomerate that has seen pretty much all of its growth from selling cloud services.

  • "Dogfooding" is a long-time Microsoft maxim, the notion that you should use the products you make on a regular basis to really understand how they work.
  • Things work somewhat differently at Microsoft subsidiaries, as our interview with GitHub COO Kyle Daigle showed, but it's always clear who is in charge.
  • And the idea that LinkedIn paused its migration in part because Azure customers needed that capacity deserves more explanation; were Microsoft's early-pandemic capacity problems more widespread and persistent than we thought?

As you make your plans for 2024, please consider sponsoring Runtime and getting your message in front of the more than 20,000 enterprise tech industry leaders and decision makers that receive this newsletter each week. We also plan to roll out several new products next year, including special reports, sponsored content, and events, both virtual and live. If you're interested in learning more, contact us here.

CUDA shoulda

After an incredible year for Nvidia's GPU business, Intel CEO Pat Gelsinger said the quiet part out loud Thursday, as reported by Tom's Hardware: "the entire industry is motivated to eliminate the CUDA market."

Gelsinger was referring to Nvidia's AI programming model for its GPUs, which has proven to be a friendly walled garden for AI developers confronted with the possibility of changing their GPU strategy. Earlier this year Intel launched the Unified Acceleration Foundation, which counts a sizable number of enterprise tech companies among its members, to develop a new way to build software optimized for the unique characteristics of GPUs.

Nvidia has a long history in AI programming and it's hard to see how that advantage will be erased quickly. But Gelsinger's comments underscore his company's precarious position as the chip money shifts to AI workloads, and it's more than a little surreal to hear the CEO of a company that has dominated tech standards for generations adopt the stance of an underdog.

Enterprise moves

Sarah Franklin is the new CEO of Lattice, replacing Jack Altman, who will become executive chairman of the HR software company.

Ryan Polk is the new chief product officer at StackOverflow, which is facing an existential crisis thanks to AI coding assistants.

Yossi Dagan is the new CFO of Noname Security, which has raised $220 million in venture-capital funding.

The Runtime roundup

HashiCorp co-founder Mitchell Hashimoto is leaving the company, two years after stepping down from the leadership team. I interviewed Hashimoto, co-founder Armon Dadgar, and CEO Dave McJannet for a different fledgling publication.

Google's GitHub Copilot competitor is now generally available, and it will start using its latest Gemini AI model "in the coming weeks" according to Techcrunch, which raises the question of why Google would launch a non-Gemini version in the wee hours of 2023.

Microsoft took down a global cybercrime organization this week, after being granted permission by a U.S. federal court to seize the assets of a group that "plays a significant role in the highly specialized cybercrime-as-a-service ecosystem."

RedMonk's semi-annual ranking of popular programming languages has been delayed after a sharp drop in activity on both Stack Overflow and GitHub, one of which makes sense and one of which does not at all.

Thanks for reading — see you Tuesday!

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to Runtime.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.