Systems of agents bring Service-as-Software to life READ MORE

DoubleZero: Scaling distributed systems through new internet rails

Ideas / News / DoubleZero: Scaling distributed systems through new internet rails

10.02.2025 | By: Rodolfo Gonzalez, Alejandra Martinez, Steve Vassallo

This map is an illustration of the DoubleZero network on day 0. Image courtesy of DoubleZero.

DoubleZero’s network launched today. As early investors in the company and internet nerds, we couldn’t be more excited for the team to actively deploy what we believe is a new way of operating how data moves around the world.

Over the past two decades, compute and storage have scaled exponentially. Moore’s Law and cloud-native architectures made it trivial to spin up vast compute clusters on demand. Storage costs have fallen to near zero.

But one part of the stack hasn’t kept pace: network performance. The public internet was designed for resiliency and cost efficiency, not for predictable low latency (how fast messages arrive) and high bandwidth (how much data moves at once).

Today, just 17 ISPs (Internet Service Providers) band together to control most of the routes and hold an oligopoly for internet traffic. The effect? Data packets take unpredictable routes. Congestion, jitter, and spam are inevitable. This is fine for serving web pages and processing a traditional bank payment. It breaks down when systems need real-time coordination at scale.

That’s why the most demanding workloads (live-streaming, global social networks, hyper-scale cloud, high-frequency trading) rely on private fiber backbones. Netflix, which our fund backed in 1999 when it was faster to ship bits via the US Postal service, now has its own proprietary network to get rid of the long buffering times of its early streaming days. Large tech companies like Microsoft and Meta have their own international fiber networks. Quant funds build direct fiber between exchanges to shave microseconds off trade latency.

Everyone else is stuck sharing bandwidth on the public internet. Owning a fiber network costs millions. The result: distributed systems are limited by latency and bandwidth ceilings, even when compute and storage aren’t the bottleneck.

Why this matters now?

We’re entering an era where latency-sensitive distributed systems aren’t niche – they’re everywhere:

  • Next-gen blockchains and real-time databases need to synchronize instantly
  • AI inference at scale requires models running across geographies to share state without lag
  • Content delivery and edge workloads demand flexible bandwidth that cloud providers can’t efficiently provision
  • Multiplayer gaming and AR/VR break immersion with even tiny bits of jitter

In all these domains, performance is collapsing to two physical constraints: bandwidth and latency. If you can’t improve both, you hit a hard ceiling.

DoubleZero founders: Andrew McConnell, Austin Federa, and Mateo Ward

It’s time to IBRL: Increase bandwidth, reduce latency

DoubleZero is building a new networking layer that makes bandwidth higher and latency lower, without requiring companies to build their own private networks.

It does this in two ways:

  1. Filtration. DoubleZero strips out junk traffic before it hits critical systems. By deduplicating and prioritizing important packets, it increases effective bandwidth and reduces wasted compute cycles.
  2. Optimized routing. Instead of sending packets through ISP cost-optimized paths, DoubleZero routes them over underutilized fiber links contributed by independent operators. These routes are verified on-chain for neutrality and accountability, but the effect is simple: packets move faster, with less jitter, along more direct paths.

The model is permissionless: anyone with spare fiber can contribute, get paid for it, and strengthen the network. Anyone needing high-performance networking can buy access. It’s a market-driven approach to bandwidth and latency, not a reliance on an ISP oligopoly or a hyperscaler.

In other words, this brings the performance of private networks to a growing set of companies needing dedicated bandwidth and low-latency global links.

Unlocking throughput in blockchains and beyond

Blockchains are the first testbed for DoubleZero because their scalability constraints are visible in real time. Every transaction requires thousands of machines to agree instantly. Today, this runs on the public internet, which introduces the jitter, spam, and latency issues we’ve highlighted. That’s why blockchains hit throughput limits even as validator hardware improves.

We invested in Solana back in 2018 with the goal of launching a permissionless “blockchain at the speed of NASDAQ”. The design goal of Solana is for the whole network to match the performance of a single node so that everyone in the world can access information simultaneously. In our previous thoughts about high-throughput chains, we’ve described how the real-time synchronization for Solana to work would come from optimizations at the hardware and software level. However, we were missing a constraint even further down the stack: consensus relies on public internet links. That means:

  • Validators spend cycles processing spam.
  • Votes and blocks propagate slowly and inconsistently.
  • Hardware requirements escalate to compensate, centralizing the network.

DoubleZero solves these issues at the root. By filtering traffic and routing it optimally, it allows blockchains to support higher throughput and increase the value of blockspace.

DoubleZero Network preventing a DDOS attack on a blockchain by filtering transactions before sending to the leader for execution

After deploying with high-throughput blockchains, the use cases for the DoubleZero network broaden naturally: gaming, CDN optimization, AI inference pipelines, enterprise distributed databases, and more. Any workload where milliseconds matter.

DoubleZero’s launch today marks the dawn of a new era for the internet

Rewriting how latency-sensitive traffic is routed across the internet is no easy feat. The DoubleZero team combines deep credibility in distributed systems with decades of experience in low-latency networking. Mateo Ward and Andrew McConnell spent decades building ultra-low-latency networking systems for quantitative trading: environments where a single millisecond determines billions in value. On the other side, Austin Federa helped scale one of the fastest growing, high-throughput, blockchains as the former Head of Strategy at the Solana Foundation.

DoubleZero’s mainnet launch today puts their work to the test. While other players continue to scale compute and reduce storage costs, DoubleZero will improve bandwidth and latency by aggregating underutilized fiber and routing traffic intelligently. In time, instant coordination across the internet will happen at the speed that physics allows.

The goal is simple: Increase bandwidth, reduce latency. And it’s working. In a few short months, over 20% of Solana stake was running on DoubleZero testnet and we’re seeing 14%-80% performance improvement vs public internet routes. This is just the beginning and we’re excited about what the team will accomplish now that the platform is live. 

At Foundation Capital, we back extraordinary founders with audacious ideas like Mateo, Andrew, and Austin. We typically invest in them early, before a product is out and sometimes even before a market is created. If that’s you, please don’t hesitate to reach out!

The DoubleZero team

Published on October 2, 2025
Written by Rodolfo Gonzalez, Alejandra Martinez, and Steve Vassallo

Related Stories