Serverless Layer Cuts Streaming Data Compute Costs
Operators running streaming pipelines on Apache Iceberg can compare Eon Yukon's vendor-reported compute savings against their own ingestion bills.
Updated on Nov. 5, 2026 in Startups

Eon, a data infrastructure company, launched Eon Yukon, a fully managed data layer that continuously turns streaming data into query-ready tables for agents, analytics, and AI. For operators running data pipelines, it writes open Apache Iceberg tables to their own object storage on a serverless, zero-idle-cost model.
Why it matters
Keeping data fresh across a growing number of sources drives up compute costs, and agents can lack the context to interpret data even when it is current. Eon Yukon targets both problems at once by combining ingestion, processing, and serving in a single serverless foundation.
Eon reports 4-7x less compute than Apache Spark and Flink in ingestion benchmarks and 7x less computing power in data-loading tests against the same two engines. In a replay of one hour of Rise production traffic, the company cites 85% fewer machines than Flink and 88% fewer than Spark while holding a 60-second freshness target. The company has raised $500 million from investors including Sequoia Capital and Lightspeed Venture Partners.
The players
Eon
Eon is a data infrastructure company whose data engine and data protection products handle ingestion, processing, and serving for production analytics and AI workloads.
Rise
Rise is an advertising platform that uses the Eon Data Engine for analytics and production applications and supplied the production traffic used in Eon's replay test.
Sequoia Capital
Sequoia Capital is a venture capital firm that is among the investors that have backed Eon's $500 million in funding.
Lightspeed Venture Partners
Lightspeed Venture Partners is a venture capital firm that is among the investors that have backed Eon's $500 million in funding.
The details
Eon Yukon combines ingestion, processing, and serving in one serverless layer, so event-level tables and configured aggregates are updated in the same ingestion pass rather than through separate jobs. It writes open Apache Iceberg tables to customers' own object storage and runs serverless SQL with zero idle compute costs, so a business pays for queries rather than standing clusters. Context about what data means and how it changes is exposed to agents through Model Context Protocol (MCP), and the product can be adopted on its own or alongside Eon Data Protection.
Timeline
Eon launched Eon Yukon on November 5, 2026.
Market Landscape
Eon positions Yukon against Apache Spark and Flink, the streaming engines most often used to keep Apache Iceberg tables current as data sources multiply. The launch is a claim about compute efficiency in a segment where the open-table format is widely adopted and the cost of running ingestion engines is a major line item.
Operators running streaming pipelines should factor in whether compute spend on ingestion and table maintenance is rising faster than data volume, since Eon's benchmarks suggest serverless engines can change that cost curve. Teams that depend on Iceberg tables should evaluate how any new ingestion layer fits with their existing storage and query tools before committing.
The takeaway
The operating insight is that ingestion compute, not storage, is becoming the cost driver as sources grow, so benchmarks on compute per unit of freshness are worth requesting from any vendor. Finance and data leads should track their own compute-per-refresh figures over the next two quarters to test vendor claims against their own workloads.
Further reading
For more on how young companies are building and funding infrastructure like this, see our Startups coverage.
Source note: This article includes information reported by The Manila times.




