Amazon Just Broke a 40-Year-Old Networking Paradigm

Since the mid-1980s, data centers have been built on what's called a "fat-tree" topology — essentially a hierarchical stack of routers and switches that data moves up and down through. It works. It's also increasingly a bottleneck at modern scale.

Amazon says it solved that problem.

According to Wired, a team at Amazon Web Services has been quietly deploying a new networking architecture called RNG — Resilient Network Graphs — in real data centers since late 2024. The design is "quasi-random," borrowing from decades of academic research on random network graphs that nobody had successfully deployed at scale.

The new system also required Amazon to invent a new piece of hardware. They call it the ShuffleBox — a device that automatically handles the cable configurations required for quasi-random networking. Without it, the physical complexity would be unmanageable.

"By essentially flattening the network, we eliminated the bottlenecks that come with traditional networking designs," said Matt Rehder, Vice President of AWS Network Engineering, in an exclusive interview with Wired. "We think we're the only ones who have done this at scale."

Brighten Godfrey, a computer science professor at the University of Illinois at Urbana-Champaign and co-author of a seminal 2012 paper on random network graphs, called Amazon's deployment "remarkable" — and he wasn't involved in the project.

Amazon says RNG is not primarily an AI training play. Rehder told Wired that AI training workloads are "far more coordinated and centrally orchestrated" and don't benefit from random graph architecture. The focus instead is on making everyday cloud computing more efficient — faster speeds, lower energy use, for general workloads.

Project Titus: Building Faster for the AI Era

Separate from the networking breakthrough, Business Insider obtained internal AWS planning documents describing Project Titus — an initiative to dramatically overhaul how Amazon builds data centers from the ground up.

The goals are aggressive. AWS wants to cut the time from construction start to first operational server room to under 35 weeks — well below current industry norms. They're also pushing liquid cooling systems and more flexible power architectures to handle the brutal energy demands of next-generation AI chips.

"The objective of the Titus portfolio of program variants is to deliver the next AWS generational Data Center design," one internal document states, per Business Insider.

Amazon already announced a $200 billion capital expenditure plan for 2026, much of it tied to AI infrastructure. They have a separate modular data center initiative called "Houdini." And they recently promoted Prasad Kalyanaraman, their AWS infrastructure services VP, to Amazon's elite S-team leadership group — the clearest possible signal of how seriously the company is treating this buildout.

The Green Initiative: Real Commitment or Checkbox PR?

On May 27, 2026, a nonprofit called Elemental Impact announced the launch of the Data Center Innovation Initiative (DCII), backed by Amazon, Google, Meta, and Microsoft. The initiative will invest between $500,000 and $5 million per project in up to 10 startups working on cleaner cooling, energy storage, and low-carbon building materials — by 2027.

So the four biggest spenders in the history of data center construction are collectively contributing to a fund that will deploy a maximum of $50 million in startup funding over two years.

For context: Amazon alone is spending $200 billion on data centers this year.

Philanthropic backers include Breakthrough Energy Discovery, Builders Vision Philanthropy, and the Salesforce and Stolte Family foundations. Legal partner is Wilson Sonsini. Dawn Lippert, CEO and Founder of Elemental Impact, framed it as a way to "pull forward important innovations" — according to the official press release via PR Newswire.

The initiative addresses genuine problems — advanced cooling, energy storage, low-carbon materials. But calling this Amazon's "green commitment" while the company invests $200 billion in capex represents a small fraction of the company's total infrastructure spending.

The Coverage Gap: What Nobody Is Saying

Mainstream coverage of the green initiative treated it as a significant environmental step. According to Bisnow, current AI growth projections put future emissions at 24 to 44 million metric tons of CO2 annually — equivalent to adding 5 to 10 million cars to U.S. roadways.

Even Pope Leo XIV weighed in this month, publishing a papal encyclical warning about the energy and water demands of AI infrastructure and calling for "more sustainable technological solutions."

When the Vatican is issuing position papers on your industry's carbon footprint, a $50 million startup fund registers as a modest gesture at best.

The RNG networking breakthrough, by contrast, shows real potential on efficiency — faster speeds with lower energy use across existing general workloads. But nobody's publishing the energy reduction numbers, and Amazon isn't disclosing them either.

What This Means for Regular People

Your cloud bill, your streaming, your online storage — all of it runs through infrastructure like this. A more efficient Amazon network means AWS can handle more traffic for less power, which theoretically flows through to cost and reliability.

Project Titus means Amazon can build new capacity faster, which matters as AI demand outpaces supply.

The DCII may produce some useful technologies. The broader question is whether incremental improvements in data center sustainability can match the scale of energy demands now being created.