Snowflake: Your data needs AI agents, too

Welcome to Runtime! Today: Snowflake rolls out new tools for building agents that work with corporate data, CDN companies deal with the effects of streaming saturation, and the latest funding rounds in enterprise tech.

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

Driving your data

As 2024 draws to a close, nearly every company in enterprise software has rolled out a plan for helping customers build agentic AI into their applications. Snowflake joined that cohort Tuesday from the data analysis side of the world, introducing Snowflake Intelligence at its Build 2024 conference.

Snowflake Intelligence is a low-code platform for building agents that can analyze corporate data based on natural-language input, in hopes of finding those magical "business insights" in the data haystack that basically every data tool promises to do. It's more or less the same concept that's driving investment in agentic AI this year, in that applications built around newer AI models can now understand imperfect queries in order to figure out how to accomplish a task.

Business intelligence analysts have been able to build applications that pull interesting nuggets from data pools in the past, but usually they had to learn specific formula languages or query languages to get anything done.
And before large-language models took off, building applications that could work with unstructured corporate data was much more challenging.
"[AI] unlocks unstructured data," said Snowflake co-founder Benoît Dageville during a press conference last week. "Unstructured data was completely opaque to the analytical platform" when Snowflake was founded more than a decade ago, he said.
Snowflake's new development tool promises to allow users to build applications that can work with third-party data sources outside of Snowflake's domain, and it cited products such as Microsoft Sharepoint, Salesforce, and Google Workspace as sources that these data agents will be able to access.

Snowflake has been working overtime this year to convince the corporate data-analysis customers that it can adapt to the rise of AI, which has long been identified more closely with its fierce rival Databricks. It replaced former CEO Frank Slootman with longtime Google executive Sridhar Ramaswamy, embraced open storage formats after years of hoarding customer data, and spent significant time during its June customer conference touting its AI prowess.

But Snowflake was very cagey about how Snowflake Intelligence works and when it will actually be available for testing: It did not discuss the product during the press conference, and would only say that it would "soon" be available in "private preview."
And its stock is down more than 33% this year compared to a 39% rise in the Nasdaq index, suggesting that the AI pitch has yet to resonate.
However, Snowflake now has "a few hundred [customers], just in the last quarter, that are in production deployed with GenAI," said executive vice president of products Christian Kleinerman during the press conference.

The company also introduced several other products and services involving open formats and security, two of the other major themes that have surrounded Snowflake this year.

A Snowflake-managed version of the Apache Polaris Catalog, which it introduced in June and later released as an open-source project, is now generally available.
Polaris allows data teams to better organize and manage data stored in open table formats like Iceberg and Hudi, but like many open-source projects, some customers prefer to let a vendor operate it for them.
And its Horizon Catalog, which allows users to set data governance priorities, can now monitor underground hacking circles for leaked Snowflake credentials and automatically disable them if they have been compromised.
This week the Department of Justice filed charges against two hackers who used stolen Snowflake credentials earlier this year to steal data from accounts that weren't secured with multifactor authentication, including personal information belonging to nearly every AT&T customer.

The lonely road

Content-delivery networks saw enormous traffic growth over the past decade as streaming video took off and consumers demanded a less-choppy experience for their monthly fee. But that growth has stalled seemingly across the board, according to Light Reading, and it has led to tough times for companies like Akamai and Fastly.

CDN traffic "is growing very slowly, at rates that we haven't seen in the 25-plus years we've been in this business," Akamai CFO Ed McGowan said last week on an earnings conference call. Fastly echoed that sentiment during its own earnings call, and both companies have laid off employees in recent months.

After pandemic lockdowns inspired millions to sign up for new video services, "video streaming in the US is saturated, reaching 96% of, or 124M, US households, only growing 1% quarter-on-quarter," Kantar said in a recent report. Cloud providers have also expanded their CDN efforts as they court media companies that might already be using their compute and storage services.

Enterprise funding

Writer raised $200 million in Series C funding, valuing the AI app-development platform company at $1.9 billion.

11x scored $50 million in Series B funding for its sales automation service, which employs perky bots named Alice and Jordan to qualify sales leads and handle outreach to customers.

ScaleOps landed $48 million in Series B funding on the first day of KubeCon to expand its Kubernetes cost-management service.

Northflank raised $22 million in Series A funding for its application development platform, which aims to make it easier for software developers to deploy apps on Kubernetes.

PointFive secured $20 million in additional funding as part of an earlier Series A round, with plans to expand its cloud cost-management services across multiple cloud providers.

UnifyApps landed $20 million in Series A funding as it tries to help businesses manage application sprawl with AI agents.

The Runtime roundup

AWS is expected to release a new generation of its Trainium AI chips next month at re:Invent, according to the Financial Times.

Red Hat acquired Neural Magic, an expert in "inference performance engineering" as its enterprise customers — many of which employ hybrid cloud architectures — invest more heavily in generative AI applications.

Signal introduced support for video calling, which could give privacy-minded companies a new option for team meetings other than Zoom, Microsoft Teams, or Google Meet.

Thanks for reading — see you Thursday!