The neocloud shakeout sorts survivors by who holds a lab contract

The space is filling up quickly, but each player covers a different piece. Two well-funded firms already sell pricing and residual-value data on GPU collateral: Silicon Data and Ornn, which are also becoming the references for exchange futures. Operators and lenders are building telemetry into their own deals. Aravolta and Hydrahost sell lender dashboards, and USD.AI puts a monitoring node into its own loans. I found no one that is clearly the independent, ABS-style servicer or verification agent. That would be a role sitting between borrower and lender, comparable to a trustee or backup servicer. It would check telemetry against power draw, billing and offtake contracts. It would track renewal and counterparty concentration risk, especially exposure to the same few AI labs. And it would issue standard periodic reports that rating agencies and bank structuring desks could rely on. The best way in is probably to build on the price indices rather than compete with them, and to sell independence and a standard reporting format. Possible buyers are ABS trustees, rating-agency surveillance teams and private-credit funds holding many neocloud loans. A new entrant faces three risks. Aravolta or Silicon Data could add this feature. Large deals are underwritten mostly on the hyperscaler or lab offtake contract rather than the hardware. And getting telemetry access depends on borrowers agreeing to it in the loan documents.

• Aravolta

  The closest direct match. It sells to lenders that finance GPU deployments, offering real-time asset verification, depreciation tracking and risk management. It also describes lease structures where terms adjust to measured GPU utilisation and thermal performance. All of these claims come from Aravolta itself.

• Hydrahost Lender Dashboard

  Private beta. It sends asset-side data on connection, health and monetisation into a lender's own loan systems, but leaves collateral ratios, repayment tracking and credit assessment to the lender. It is the operator's own tool, not an independent servicer.

• Silicon Data

  Raised a $30.5M Series A. It publishes GPU rental price indices, forward curves with tenors from 1 to 36 months, and DCF-based residual value curves that use utilisation and opex assumptions. Its benchmark is reportedly the reference for a CME compute futures product. It covers the pricing and residual-value side of collateral, but does not monitor individual fleets.

• Ornn

  Raised a $33M seed. Its Ornn Compute Price Index is built from cleared transactions and is available on Bloomberg. It is the planned reference for ICE compute futures and AX perpetuals, and it is marketed to lenders for benchmarking and for hedging depreciation. This is market pricing, not loan-level monitoring.

• SiliconLien

  Publishes research arguing for a lender-side verification layer that checks telemetry against power, billing and contracts, at closing and after funding. That is close to this idea, but I could not confirm whether it has a shipping product.

• USD.AI

  A crypto-native GPU lender. It places an on-premises monitoring node in each financed SPV to track utilisation, uptime and asset health for the life of the loan. Monitoring is built into its own lending and not sold to third parties, but it shows the model works in practice.

• SemiAnalysis ClusterMAX

  Rates GPU cloud operators on quality, and lenders are reported to use it as a qualitative input on neocloud borrowers. It also notes that parametric insurers now cover contract-termination risk for neoclouds and their lenders.

• CCIR Compute Credit Tracker

  Turned up in search as a tracker of GPU-collateral financings that carry public agency ratings, with links to filings. That is deal-level market data, not fleet monitoring. I could not verify much beyond the listing.

Checked 2026-10-09