A startup is trying to do for Nvidia GPUs what bond markets did for mortgages: turn them into standardized, predictable, bankable assets. According to The Information, the company is building pricing infrastructure that lets lenders, lessors, and data center operators treat H100s and H200s less like volatile tech gear and more like the kind of “boring” collateral banks actually want to finance.
This is significant because GPU financing has been a mess. Hyperscalers buy chips outright. Everyone else has been scrambling, with neoclouds raising billions in private credit at eye-watering rates, and lenders struggling to price risk on hardware that could be worth a fortune or obsolete inside three years.
What’s actually changing
The Information reports the startup’s pitch centers on transparent, observable pricing for GPUs across their lifecycle. Think of it as a “Kelley Blue Book” layer for AI compute:
- Standardized resale and lease values for current and prior-gen chips
- Depreciation curves that lenders can underwrite against
- A reference point for secondary-market trades as Blackwell ramps and Hopper ages out
If that data becomes the industry standard, the downstream effects are real. Cheaper debt for GPU buyers. Lower hurdle rates for new data centers. A genuine secondary market for chips that today mostly sit locked inside one operator’s racks.
Why this matters for the AI build-out
The AI infrastructure spend is projected past a trillion dollars over the next few years. Most of that has been funded by hyperscaler cash flow and increasingly aggressive private credit. What’s been missing is the plumbing that turns GPUs into a normal asset class, the kind insurance companies, pension funds, and regional banks can touch without panicking.
What stands out here is the timing. Blackwell shipments are accelerating. Hopper-generation chips are about to face their first real “is this still worth anything?” stress test. A reliable pricing layer arriving right now gives the market a way to answer that question with data instead of vibes.
What to watch next
A few things will tell us if this actually sticks:
- Adoption by lenders. If major private credit shops start citing this pricing in term sheets, it’s real.
- Hopper resale floors. As H100 leases roll off in 2026 and 2027, transparent pricing will either stabilize the secondary market or expose just how fast these chips depreciate.
- Neocloud refinancing. Companies like CoreWeave and Crusoe have stacked debt against GPU fleets. Standardized valuations could let them refinance at lower rates, or force markdowns if reality bites.
- Nvidia’s response. Nvidia benefits from scarcity pricing. A healthy secondary market complicates that, but also expands the buyer pool.
The broader trend is clear: AI infrastructure is moving from a frontier asset class to a financial product. Boring is the goal. Boring is what unlocks the next leg of capital.
Full details at The Information.