Chamber founder Charles Ding had everything going for him at Amazon. Year after year, he exceeded expectations. The scope kept growing, the impact kept expanding, the visibility put him on a clear path upward. Most people would have stayed.
He didn’t.
“It was one of the hardest decisions of my career,” Charles said. “But Amazon taught me something I couldn’t ignore. It’s always Day 1. And Day 1 only matters if you’re willing to start again.”
So he walked away. As a second-time founder, Charles launched Chamber, an AI Infrastructure optimization platform that just got accepted into Y Combinator’s Winter 2026 batch. He’s not doing it alone. Three fellow Amazon veterans – Shaocheng Wang, Jason Ong, and Andreas Bloomquist – left their own comfortable positions to join him. Between them, they’ve spent decades building and shipping products from zero to one at a massive scale.
Y Combinator officially announced Chamber’s launch while highlighting how the platform puts AI infrastructure orchestration, governance, and resource optimization on autopilot. The accelerator noted that Chamber enables companies to run on average 50 percent more workloads using their existing AI infrastructure—without adding hardware or headcount.
“I’m humbled that they took this leap of faith,” Charles said. “We all had safe jobs. We all had reasons to stay. But sometimes the pull toward something bigger is just too strong to ignore.”
That “something bigger” turned out to be a problem hiding in plain sight. Despite billions that pour into AI infrastructure every year, somewhere between 30 and 60 percent of enterprise GPU capacity sits completely idle. Teams reserve hardware they don’t fully use. Resources get siloed. Meanwhile, other teams wait in endless queues for access to GPUs that are technically available but locked away in someone else’s allocation.
The numbers are staggering. Industry estimates put the annual waste at over $240 billion – and that figure keeps climbing as AI research accelerates. Companies are spending fortunes on hardware that spends most of its time doing nothing.
Chamber fixes that. The platform brings GPU Optimization to enterprises through agentic AI that acts like an autonomous infrastructure team. It monitors GPU clusters continuously, predicts resource needs, detects bad nodes, and reallocates GPUs in real time so AI teams can move faster without manual work.
“Most companies have no idea how much GPU capacity they’re actually wasting,” Charles explained. “They can’t see it. The data is scattered across teams, clusters, and tools that don’t talk to each other. We give them visibility they’ve never had before.”
The platform handles GPU Efficiency through several integrated capabilities. It automatically discovers underutilized GPUs and surfaces optimization opportunities that would otherwise stay hidden. An intelligent scheduling system right-sizes workloads and manages allocations based on real-time demand and priority. When hardware starts failing – something that happens more often than most people realize – Chamber’s self-healing infrastructure detects the problem and routes around it before corrupted training runs waste days of compute time.
For engineering leaders and executives, there’s a layer of analytics and reporting that turns raw utilization data into actionable insights. Native integrations with Slack, email, and PagerDuty mean the right people get alerts when something needs attention.
The founding team incubated had built and scaled large-scale infrastructure optimization at Amazon, delivering hundreds of millions in cost savings, and are now bringing those learnings to help other enterprises uncover idle AI infra, accelerate workloads and reduce wasted money on reserved idle GPUs.
Charles draws inspiration from Jeff Bezos’ regret minimization framework – the idea of people asking themselves what they’ll regret not trying when hey look back years from now. When he frames decisions that way, the answer usually becomes obvious.
As part of their launch, Chamber is offering a free GPU Intelligence Dashboard that gives teams immediate visibility into their AI infrastructure. The monitoring agent discovers unused capacity in real time, breaks down utilization patterns across teams and workloads, and delivers executive-ready reports on GPU usage trends.
For AI and ML teams tired of fighting for resources, and for executives who need accountability over GPU spend, Chamber represents a new way forward. One built by people who witnessed and pain first hand and understood the challenges and problems at scale – and were crazy enough to leave everything behind to solve it.
More information and free account registration are available on the official website.