Investors

The category

GPU-backed private credit is a new asset class, growing directly out of hyperscaler and neocloud capital spending on compute. Lenders are underwriting large, long-dated facilities against hardware that depreciates fast, moves easily, and has no established third-party verification standard yet — the equivalent gap that independent inspection and collateral management firms closed decades ago for physical commodities and warehoused goods.

Goldman Sachs Research now expects the leading hyperscalers to spend a combined $5.3 trillion on AI infrastructure and data centers between 2025 and 2030, up from a $4.5 trillion estimate before this year's first-quarter earnings — spending that increasingly outruns what liquid credit markets alone can absorb. Private credit is filling that gap directly: Morgan Stanley projects it will supply an additional $800 billion in data center financing over the next two years, with funds including Blackstone, Blue Owl, Apollo, Pimco and BlackRock originating most of that debt. CoreWeave's $7.5 billion facility with Blackstone, priced at an average floating rate of roughly 11%, is one of the largest single GPU-backed facilities closed to date.

Sources: Goldman Sachs Research, "Private markets expected to have a growing role in data center financing" (Jun 2026) · Quinn Emanuel client alert, "Emerging Litigation Risks in Financing the AI Data Center Boom" (16 Mar 2026), citing Morgan Stanley research and the CoreWeave–Blackstone facility.

The risk

Today, monitoring on these facilities is largely self-reported — a quarterly PDF from the borrower, taken on trust. There is no independent, continuous check that the collateral is present, operating, and unencumbered, and no reliable way to assess its residual value as it ages. That verification gap sits directly underneath the credit risk lenders are pricing.

Residual value is genuinely contested, which is itself the problem: independent analysis of thousands of H100 resale listings between June 2024 and December 2025 found three-year-old used units trading around 69% of the equivalent new unit's price and refurbished units around 84% — while falling to just 20–30% of their historical peak price. Hyperscalers don't agree with each other either: in the same quarter, Amazon shortened its GPU useful-life assumption while Meta extended its own. A lender pricing a facility today has no independent, continuously updated view of what the collateral underneath it is actually worth.

The failure mode when collateral isn't independently verified is well precedented, just not yet in this asset class. In the 2014 Qingdao port fraud, a single Chinese trading firm raised RMB 12.3 billion (US$1.78 billion) using fake or duplicated warehouse receipts for metal that was pledged, in some cases, to 13 different banks at once — producing over US$1 billion in international losses and up to US$3 billion in exposure for Chinese banks. The market's eventual response was to demand collateral sit with an independent, verified custodian rather than take the borrower's word for what was in the warehouse. GPU-backed credit doesn't have that custodian yet.

Sources: Silicon Data, "The Illusion of Stability: H100 GPU Market Value Trends" (30 Jan 2026) · Global Trade Review, "Qingdao fraud probe ends with jail term" (10 Dec 2018).

The platform thesis

Verittest starts as independent verification and covenant monitoring for a single lender on a single facility. Each engagement produces evidence — presence, condition, utilization, covenant performance over time — that accumulates. Over time, that evidence base is what lets Verittest extend from a per-deal verification service toward becoming the reference valuation and monitoring source for GPU-backed collateral across the asset class.

Where things stand

Verittest is working software today, not a prototype or a deck. We're currently working with a syndicate of private credit funds based across APAC to build the covenant framework for their first GPU-backed facility.

Founder

Verittest is built by a founder with a background in structured credit at a major institutional asset manager, partnership-level experience advising on strategy, and dual-jurisdiction legal qualification in New York and Singapore.