Atlanta’s been known for heat, but this kind burns in teraflops. QumulusAI just secured a $500M non-recourse financing facility arranged by Permian Labs and powered by the USD.AI Protocol, a DeFi structure that lets them borrow stablecoins against up to 70% of approved GPU deployments. Translation: this isn’t another equity round. It’s on-chain credit that doesn’t dilute ownership, redefining what capital efficiency looks like in AI infrastructure.
Founded in 2019 by Chris Bissell and Patrick Gahan, QumulusAI started with a simple idea, if AI’s future runs on compute, then control the power, the hardware, and the latency that fuels it. Bissell, who once built e-commerce systems at the Home Depot and mined Bitcoin before it was cocktail talk, teamed up with Gahan, a former Lockheed Martin and Booz Allen engineer turned strategist who built and exited Seismic LLC. Together, they built a vertically integrated machine: GPU cloud, power generation, and purpose-built Tier 3+ data centers operating below a 1.1 PUE. That’s efficiency most hyperscalers only brag about in theory.
Today, under CEO Michael Maniscalco, CTO Ryan DiRocco, and CMO Stephen Hunton, the company’s hitting stride. With 550+ GPUs deployed across a 108MW power portfolio and 60MW of active capacity across GA, TX, and OK, QumulusAI is proving that performance isn’t just about compute density, it’s about design discipline. Backed by Chairman Steve Gertz and guided by advisor Conor Moore of Permian Labs, this crew’s blending old-school infrastructure grit with new-school blockchain finance.
What makes this $500M move more than just another number on the board is its timing. While most AI infrastructure players are chasing equity headlines, QumulusAI is tokenizing reality, literally. Their GPU Warehouse Receipt Token (GWRT) model with Permian Labs turns deployed GPUs into on-chain assets, collateralized and liquid within the USD.AI Protocol. It’s compute as currency, built for an era where demand scales faster than supply chains.
This capital infusion accelerates two new Tier 3+ U.S. Midwest data centers by mid-2026, sets the stage for European expansion, and fuels an automated GPU orchestration platform coming early next year. Enterprise AI teams in finance, biotech, and autonomous systems, and even the SMB devs running pay-as-you-go models, now get access to serious horsepower without the waitlist.
In a market projected to top $6.7T by 2030, QumulusAI isn’t just supplying infrastructure. They’re shaping how it’s financed, deployed, and owned. The message is clear: the future of AI won’t just be trained on Qumulus, it’ll be powered by it.
