AI outpaces today’s infrastructure. Hyperspeed compute closes the gap: distributed, modular GPU access without long buildouts or procurement drag. We're QumulusAI, and we're breaking AI’s biggest barriers. #AI #Hyperspeed #GPU #Infrastructure

“Per-token pricing makes sense when companies are experimenting, but the economics change quickly when AI applications move into sustained production,” said Mike Maniscalco, CEO of QumulusAI. “This research shows why enterprises need a more deliberate infrastructure strategy, using flexible services for experimentation and reserved capacity for predictable, high-volume workloads. The goal is not to replace the public cloud, but to put each workload where it makes the most economic and operational sense.” Read more and download the report for free >> qumulusai.com/articles/agent… #tokenomics #tokens #AIinfrastructure #futurum #hyperspeed

A year ago: 90%+ of our business ran through one customer. Today: 96% of it is direct. Our CEO Mike Maniscalco on how the team rebuilt QumulusAI's customer base — from the Drill Down podcast Full conversation >> piped.video/rHgCrhQsq-c?si=ro8m…

QumulusAI isn't building the AI models. We're building out the infrastructure they run on. CEO Mike Maniscalco tells @StevenDickens3 (@HyperFRAME_Res ) why compute, networking, and storage are the real picks and shovels of the AI wave. #AIInfrastructure #NeoCloud

Pricing per token seemed reasonable until the invoice hit. @TheFuturumGroup breaks down the real cost of per token pricing and the way out of "tokenomics." Read more in the research we supported. 👇 #tokenomics #AIinfrastructure #hyperspeed #qmls

6 customer deployments. All Blackwell. 296 nodes. 2,360 GPUs. This isn't hyperscale. It's hyperspeed. 🚀 #QumulusAI #NVIDIA #Blackwell

Blackwell might be a Honda, not a Ferrari. And that's... fine. Our CEO Mike Maniscalco on why revenue per watt (not raw chip performance) is the metric that matters most in AI infra right now. Full convo on YouTube: piped.video/watch?v=rHgCrhQs…

Deploy an LLM as an agent instead of a prompt, and watch your token count jump as much as 10-100x. Nothing about the model changes. Nothing about your user changes. But with per-token pricing, every one of those tokens shows up on the bill. We sponsored new independent research from @TheFuturumGroup on how enterprises are getting ahead of this before it wrecks their AI budget. A few numbers that stood out to us: → 59% of AI decision-makers already run primarily outside the hyperscalers → 66% of compute is now consumed via reserved or owned capacity vs. just 19% on-demand → Spend outside the big three hyperscalers is projected to grow 8x by 2030 The report also digs into the ~60% utilization threshold where renting compute stops making financial sense (and almost nobody actually knows their own number). Read the full report and grab the toolkit: a decision grid, a cost exposure model, and a CFO packet for the budget conversation. lnkd.in/eu_mqHsx #Tokenomics #AIInfrastructure

“Moving from signed agreement to delivered capacity in months, not years, is what this business is built to do. The B200 capacity is live and serving our customer’s inference workloads today, and the B300 portion is close behind,” said Mike Maniscalco, CEO of QumulusAI. Read more >> qumulusai.com/articles/qumul… #hyperspeed #nvidia #blackwell #AIinfrastructure

If you're securing your Blackwell and rack-scale capacity for 2027, be sure to find Mark Jackson on the floor at the AI Infra Summit in Santa Clara, CA this week. #AIinfrastructure #AIInfraSummit #hyperspeed

"It's all about the ecosystem" — Steve Gertz, CGO, QumulusAI. Steve recently joined the #TechTalk podcast (with host Joey Kline) to discuss QumulusAI's role in Atlanta's AI revolution. In this clip, Steve takes the lid off our secret sauce. #QMLS #ecosystem #hyperspeed

“Building a reliable GPU business starts with reliable supply partners, and that’s the relationship this deployment represents. The agreement we announced in July is now live in Philadelphia and fully in our customer’s hands,” said Mike Maniscalco, CEO of QumulusAI. Read more: qumulusai.com/articles/qumul…

"Hyperspeed" is how we're moving today, and it's the foundation for the distributed inference model we're building for tomorrow. CEO Mike Maniscalco tells @StevenDickens3 (@HyperFRAME_Res) how smaller, faster-to-deploy pockets of compute are building toward a hyper-distributed footprint designed for where inference is headed. #AIInfrastructure #hyperspeed #qmls

AI is the biggest infrastructure build out in history, on the scale of the railroads, the fiber that built the internet, the highway system. Everyone's clocking the models. QumulusAI builds what they run on. In this panel from our @Nasdaq Bell Ringing celebration, CEO Mike Maniscalco, CTO Ryan DiRocco, and Chief Growth Officer Steve Gertz sit down with @StevenDickens3 of @HyperFRAME_Res to talk about what's really driving the AI infrastructure boom (and where QumulusAI fits inside it).

"Teams building on Runpod need inference capacity that lands fast and prices well. QumulusAI got all 616 of these GPUs live ahead of schedule, which means developers get to use them sooner," said Bill Sehmel, Manager of Datacenter Infrastructure at @runpod. Read the full release >> qumulusai.com/articles/qumul… #hyperspeed

Scaling sounds simple until you actually have to deliver it. CEO Mike Maniscalco tells @StevenDickens3 (@HyperFRAME_Res) why bare-metal-as-a-service and managed Kubernetes became QumulusAI's answer: packaging the experience, depth, and capital most organizations don't have in-house.

Supply chain constraints are the biggest bottleneck in AI infrastructure today, and traditional hyperscalers aren't built to solve it. On @Nasdaq Amplify Spotlight, QumulusAI CEO Mike Maniscalco explains how our “hyperspeed” deployment strategy overcomes supply chain gridlock, letting QumulusAI move faster than hyperscalers and even other neoclouds. #Hyperspeed #AIInfrastructure #Nasdaq #EnterpriseAI

QumulusAI is in Austin this week for Datacloud USA x Metro Connect Fall. On the ground: Mike Maniscalco, Stephen Hunton, Steve Gertz, and Michael Halles. Catch our CEO, Mike Maniscalco, on the Metro Connect Fall stage (Level 5) discussing "Building an Ecosystem: The Critical Role of Network Interconnection in Digital Infrastructure": 🗓️ Sep 2, 2026 | 🕒 3:30–4:15 PM If you're there, come find us. #DatacloudUSA #MetroConnect #AIInfrastructure #neocloud

.@GeorgiaTech is one of the top producers of AI-trained graduates in the country, and it's one of @nvidia's largest sources of talent too. Tim Lieuwen on the $20M NSF-funded supercomputer, the AI-native commitment for every graduating student, and why the physical plant behind AI matters as much as the chips themselves. Watch the full conversation on our YouTube channel.

What does it really take to bring enterprise-grade AI infrastructure online? Solving every layer of the physical and operational stack simultaneously. QumulusAI CEO Mike Maniscalco breaks it all down on Nasdaq Amplify Spotlight: ⚡ Sourcing power 🏢 Data center capacity 🧠 Compute 💾 Storage 🌐 Networking 🔗 Services that tie it all together > View full interview: nasdaq.com/news-and-insights… #AIInfrastructure #Nasdaq #DataCenters #CloudComputing #QumulusAI #GPUCompute #HighPerformanceComputing

We just extended our optionality at our Oklahoma site from one renewal option to three, extending potential tenure through January 2044. The site's already operating, carries 19 MW of contracted power, and the amendment opens the door to leasehold financing for future development there. > Read more: qumulusai.com/articles/qumul…