ChipsThe story, in brief

AWS Replaces Fat-Tree Data Center Networks with Random Graph Theory, Cutting Routers by 69%

69% fewer routers. AWS just rewrote the rules on data center efficiency—and that changes everything about AI infrastructure costs.

Paper-cut illustration of an amber microchip with circuit paths extending into a row of data-center cabinets.
The infrastructure powering AI.AI illustration by KeyNews
The KeyNews take

Why it matters

AWS's shift to random graph theory networking architecture directly reduces the capex and opex burden of AI data center buildouts. For founders and investors, this signals a major efficiency unlock that could reshape competitive dynamics in compute-intensive AI workloads.

The key facts

7 to know
  1. AWS replaces fat-tree hierarchies with Resilient Network Graphs (RNG) based on quasi-random graph theory

  2. 69% reduction in routers required

  3. 33% throughput improvement

  4. 40% reduction in network power consumption

  5. Uses passive optical ShuffleBoxes for ToR-to-ToR mesh connections

  6. Now default for most new AWS data center builds

  7. Published June 2026

Go to the source

InfoQ AI/MLinfoq.com

Publisher excerpt: AWS disclosed that Resilient Network Graphs, a flat network architecture based on quasi-random graph theory, is now the default for most new data center builds. The design replaces fat-tree hierarchies with direct ToR-to-ToR mesh connections using passive optical ShuffleBoxes, cutting routers by…
Read original report
Back to today's editionMore chips news

Keep reading

Related stories

More from Chips