I’ve spent a lot of time recently reflecting on the sheer scale of the AI infra buildout required for the next 5 years. It is staggering. We are talking about gigawatts of power and acres of cooling infrastructure. Physical reality is becoming more of a bottleneck. The

We are watching Venture Capital try to do the job of Infrastructure Finance, and it's massively inefficient. Right now, VCs are using expensive equity dollars to fund hardware "down payments". Even when a neocloud secures debt from a major lender, they are still forced to pay a

The AI economy is beginning to mirror the Energy sector. You have the raw resource (Electricity/Data), the generation plants (Data Centers/GPUs), and the transmission lines (Networks). Yet, we are still financing buildouts with VC money. It is a fundamental mismatch of capital

We are proud to announce the release of our latest whitepaper, co-authored with The Family Office Association (@TFOA_SFO): "A New Frontier for Family Office Investing: GPU-Based Infrastructure" As AI capital expenditure scales toward a projected $7 trillion by 2030, family

We need to stop financing neoclouds with venture capital logic. In the industrial world, you don't build a power plant by diluting your equity. You finance it against the future electricity it will sell. Compute should be no different. The current mismatch where neoclouds buy

We need to stop financing neoclouds with venture capital logic. In the industrial world, you don't build a power plant by diluting your equity. You finance it against the future electricity it will sell. Compute should be no different. The current mismatch where neoclouds buy

The "Equity Gap" is a silent killer for many neoclouds trying to scale right now. Even the few neoclouds that manage to secure traditional debt face a math problem. Banks are hesitant to lend above 70–80% Loan-to-Value (LTV) to smaller neoclouds. That leaves the operator to fund

The primary constraint on AI infrastructure deployment is capital efficiency. We are seeing a massive disconnect between compute demand and the structures available to finance it. Thinking of GPUs like standard IT equipment traps operators in a CapEx cycle that limits growth.

Solar energy became a bankable asset class only when the risk models matured. Before we had standardized data, solar was treated as venture risk. Once the yield became predictable, it became infrastructure. GPUs are crossing that exact same bridge. We now have the utilization

For years, compute had a coordination problem. Neocloud operators needed GPUs, yield-focused investors were open to new real assets, and there was no straightforward way to link capital to the performance of the hardware. Revenue-share financing closes that gap. Neoclouds

saas founders: "we are building the future of agi" us moving a pallet of gb200s through customs so ur chatbot works:

Hyperscalers became enterprise partners by doing a few simple things very well: staying online, delivering consistent performance, and building trust over time. Neoclouds are now starting that same journey. The difference will be in how they fund growth. With revenue-share

AI infrastructure development accelerated faster than the financing systems around it. Billions in GPUs are on order, but many neoclouds can’t scale because the capital markets haven’t fully adapted to the asset class. Hardware that produces revenue every hour still isn’t

Investors can now get AI exposure through the economics of compute itself, not just through public equities. When GPUs run paid workloads, they generate hourly revenue. With clear utilization data and consistent demand, we can underwrite those deployments and structure the cash

The maturity of any infrastructure market begins the moment performance becomes measurable. Compute is at that point. Neoclouds are collecting utilization, uptime, and workload data with enough consistency for lenders and investors to understand how these deployments actually

The biggest difference between mature infrastructure assets and early-stage compute is how well performance can be understood. Power plants and data centers have decades of standardized metrics. GPUs are just now catching up: utilization, uptime, workload mix and other factors

nvidia earnings call, first sixty seconds "we have line of sight to a half trillion in revenue in 2026" the bubble hasn't started yet

Nowadays people read the word "GPUs" and think “expensive hardware". However, most of the true value comes from how efficiently the hardware is powered and kept online. Power contracts decide the cost of every workload. Cooling and uptime help clients determine whether to stay

Neoclouds track a few core metrics to gauge how well their infrastructure is performing: • GPU Utilization: how often GPUs are running paid workloads • PUE (Power Usage Effectiveness): how efficiently the facility turns power into compute • Uptime: how consistently the site

Most people know GPUs power AI models, but they don’t always know how that turns into revenue. At the simplest level, GPUs earn money when they’re running paid workloads like model training or inference. When the hardware is being utilized, it generates income. When it’s