Snowflake makes open its new north star

Welcome to Runtime! Today: Snowflake targets another layer of the data stack after embracing open formats, why hackers targeted Snowflake customers and stole a massive amount of data, and the latest funding in enterprise tech.

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

Open for business

SAN FRANCISCO — Snowflake grew into a $45 billion company on the strength of its very sticky data analytics engine, which companies poured their data into over the last decade to improve their decision-making processes. However, as the value of that data skyrocketed, its customers started to realize they wanted more control over how it was used and where it was stored.

At its Data Cloud Summit this week, Snowflake unveiled a new open-source project called Polaris that will help users work with Apache Iceberg, one of several open table formats that have upended the data storage and analytics markets by giving end users the ability to store their data in the repository of their choice. It also extended its support for Iceberg, Snowflake's preferred open format for this new era of data management.

• Polaris is a catalog, a data tool that sits above data stored in Iceberg and ensures "data engineers and their pipelines can modify tables concurrently, and queries on these tables produce accurate results," Snowflake said in a blog post.
• It will allow companies working with Iceberg tables to more easily manage the process of applying query engines from several different vendors to data stored in Iceberg, even if Snowflake isn't one of those vendors.
• "Many of you, especially those at large enterprises, have been asking us for better ways to centralize security and access to all your data, not just the data that's stored in Snowflake," said CEO Sridhar Ramaswamy in a Monday afternoon keynote speech.
• Polaris will be released as an open-source project in the next 90 days; the company hasn't made a final decision about licensing, but it is leaning toward an Apache license given Iceberg's heritage, said executive vice president of product Christian Kleinerman in an interview Tuesday.

Snowflake also now fully supports Iceberg by allowing Snowflake customers to store Iceberg data in AWS, Microsoft Azure, and Google Cloud, Kleinerman announced in a keynote Tuesday.

• Until very recently, most Snowflake users were required to store a copy of their data with Snowflake in a proprietary format to get the benefits of its query engine, but that approach fell out of favor as a cost-cutting mindset set in among tech departments.
• Snowflake announced partial support for Iceberg in 2022, but didn't complete the link that would allow its customers to run Snowflake on externally stored data until now.
• The move will cost Snowflake some storage revenue, but the company was at risk of broader customer defections had it not embraced the open format trend.
• Not all customers will choose to move their data out of Snowflake, but the point is that they now have that choice if that's their preference, Kleinerman said in the interview.

However, in an announcement timed to hit precisely when Snowflake executives kicked off Tuesday morning's keynote, rival Databricks announced that it was also planning to double down on Iceberg.

• Databricks acquired Tabular, a 40-person startup founded by the creators of the Iceberg format, for "between $1 billion and $2 billion," according to the Wall Street Journal.
• "As one, we are going to lead the way with data compatibility so that you are no longer limited by which lakehouse format your data is in," Databricks said in a blog post.
• Tabular sold a "managed storage engine" that made it easier for customers that wanted to adopt Iceberg as their open format of choice link their data with any data query engine they preferred, including Snowflake, Databricks, and data products from cloud providers like Amazon Athena and Google Cloud's BigQuery.

Databricks was ahead of the open format trend with an architecture that never required data storage to sit alongside data compute, but Snowflake's decision to open up its platform means that competition is shifting back to the query engines themselves as AI puts new demands on enterprise data managers.

• "Go evaluate the best engine for the best workload," Kleinerman said in the interview. "If anyone gives you the best performance and the best price/performance, go for it. We're just confident that we have a really good engine."
• Databricks kicks off its own customer event next Tuesday in San Francisco.

Meanwhile…

Over the weekend before Snowflake's event, reports emerged of massive data breaches at Snowflake customers Ticketmaster and Santander. Snowflake confirmed that an incident took place, but insisted that the breach did not stem from any vulnerabilities in its data-warehousing software, which contains an enormous amount of confidential customer information.

According to Ars Technica, the attackers used infostealer malware to obtain legitimate login credentials for several Snowflake customers. Those accounts were not protected by multifactor authentication, which allowed the attackers to make off with the personal information, including some credit card data, of up to millions of customers who paid Ticketmaster exorbitant transaction fees for the privilege.

It's a little hard to understand how anyone responsible for managing a corporate database account containing a treasure trove of sensitive information in 2024 has not activated multifactor authentication on that account. But security researcher Kevin Beaumont criticized Snowflake's MFA implementation and suggested it could have done more to protect its customers.

Enterprise funding

Firefly raised $23 million in Series A funding to help companies apply security and governance to their code-managed infrastructure.

Stacklet raised $14.5 million in Series B funding for similar technology that helps customers detect and eliminate wasteful cloud spending.

The Runtime roundup

The FTC is asking Microsoft competitors for dirt on its bundling strategy as part of a potential antitrust investigation, according to NextGov/FCW.

CoreWeave offered $1 billion to acquire Core Scientific, a bitcoin mining company that could follow CoreWeave's path and become a GPU provider to AI companies, according to Bloomberg.

Sales of AI servers doubled during the last quarter at HPE, which handily beat Wall Street expectations and raised its guidance for the year.

Intel CEO Pat Gelsinger had the chutzpah to complain that U.S. government regulations on chip sales to China were hurting its business, months after receiving $8.5 billion in subsidies from the U.S. government to invest in domestic manufacturing capabilities as a hedge against the security threats posed by, well, China.

Thanks for reading — see you Thursday!