UpTrajectory Review

Nscale, a European GPU cloud provider, is negotiating a $3.5 billion funding round that would value it among the most capital-intensive infrastructure plays in recent memory. The structure is telling: $1.5 billion in convertible notes led by hedge fund Third Point, plus roughly $2 billion from Nvidia itself. The company is pitching investors on a contract backlog of approximately $103 billion, a figure that ballooned largely because Nscale secured a $45 billion deal with Anthropic after both Microsoft and Google declined to match the terms. For a firm that was relatively obscure even within cloud circles a year ago, this is a staggering acceleration, and it signals how the AI compute layer is being reconstituted around a handful of vertically integrated players willing to absorb enormous capital risk.

For small-business operators, the immediate read is that cloud costs for AI workloads are not following the commoditization trajectory of traditional cloud storage or general compute. Nscale's entire pitch depends on scarcity economics: Nvidia's latest chips are allocation-constrained, and the companies that can secure them at scale can extract premium pricing from AI labs desperate for training and inference capacity. If you are running AI-dependent operations—whether customer service bots, content generation, or analytics pipelines—your negotiating leverage against providers is shrinking, not growing. The concentration of supply in entities like Nscale, which itself depends on Nvidia's favor, means that pricing power sits two or three layers above where most businesses interact with the stack.

What deserves scrutiny is the $103 billion contract figure and the Anthropic deal specifically. The Next Web's reporting that Microsoft and Google 'passed on' the $45 billion arrangement is genuinely significant: it suggests that the most capitalized cloud incumbents, with their existing infrastructure and balance sheets, found the risk-adjusted returns unacceptable or the terms unsustainable. Nscale's willingness to take this on implies either superior access to Nvidia silicon, a different cost of capital, or a bet that Anthropic's growth trajectory will validate economics that larger players deemed too speculative. We are skeptical that a $45 billion commitment to a single AI lab, even one as prominent as Anthropic, represents prudent portfolio construction rather than a moonshot concentrated bet that could unwind spectacularly.

The downstream effects bifurcate sharply. For Nvidia, this is vertical integration by proxy: rather than building cloud capacity directly, it is effectively financing a captive demand channel that absorbs GPUs at scale, insulating pricing and utilization from competitive pressure. For Anthropic, locking in $45 billion of compute potentially overcommits it to a single provider, creating operational risk if Nscale stumbles or if better silicon economics emerge elsewhere. For smaller AI startups and the businesses that depend on their services, the consolidation means fewer independent compute options and greater susceptibility to disruption if any node in this tightly coupled network fails. The cost of redundancy is being socialized downward while the profits of concentration accumulate at the infrastructure layer.

Watch whether Nscale's US listing actually materializes at the implied valuation, and whether the convertible note structure suggests Third Point is hedging downside rather than expressing pure conviction. The conversion terms will reveal how much equity dilution the founders are accepting for capital that may be deployed over years against contracts that themselves may renegotiate or default. For operators, the actionable response is to audit your AI supply chain for single-provider dependency, to pressure vendors for multi-year pricing commitments before any Nscale-driven cost reset, and to evaluate whether open-source models running on commodity hardware can substitute for API-dependent workflows that pass through this increasingly expensive and fragile infrastructure layer.

“Microsoft and Google both passed on” — The Next Web

Takeaway: Lock in multi-year AI compute contracts now before infrastructure consolidation eliminates your pricing leverage.