The AI boom has spurred massive investment in new power plants, chips, and data centers. In parallel, a more subtle financial infrastructure is emerging, creating viable channels for asset-backed lending against high-value hardware.
Miami-based fintech startup Barkr is one company building this service layer, specializing in AI chip collateral valuation and having already garnered attention from NVIDIA. The company's CEO and co-founder, Thomas Galbraith, recently discussed his firm's role at a San Francisco forum focused on how lenders assess credit for AI infrastructure projects.
Galbraith operates in a highly specialized niche within the complex world of asset-backed finance. His company provides insured valuation services for physical hard assets, including cars, artwork, commercial equipment, and now, AI chips. Barkr does not originate loans itself. Instead, it provides lenders with authoritative, credit-ready asset valuations.
This business requires immense confidence: if a Barkr valuation is off and a borrower defaults, the company covers the loss differential. Founded in 2023, Barkr initially priced assets like collector cars and private jets to help lenders set appropriate loan amounts. About six months ago, a lender presented a new challenge.
Galbraith recalls being told, "We want to lend against GPUs, but we can't find anyone to price them fairly, and certainly no one to insure that price." While NVIDIA graphics cards have become a new asset class in recent years, asset-backed lending against them is not yet widespread. The core issue is disagreement over residual value, with no neutral third party to resolve disputes, leading to less favorable loan terms compared to traditional assets.
Asset-backed lending carries two primary risks: the borrower's ability to repay and the stability of the collateral's value. Lenders have decades of experience assessing borrower creditworthiness. However, valuing the assets themselves, especially novel ones like GPUs, remains difficult. Galbraith notes that five different firms might give five vastly different valuations for the same cluster of chips.
Barkr develops AI-powered valuation models to appraise collateral and partners with insurers like Munich Re to provide legally binding guarantees on those valuations. Galbraith states this model changes the calculus for lenders. A fair, credible asset price reduces lender risk and helps borrowing companies secure lower interest rates.
In the past six months, Barkr has facilitated roughly $200 million in GPU-backed financing and expects to handle another $300 million by year-end. While modest compared to NVIDIA's overall sales—most chip purchases are still made with large corporations' own capital—it represents growth. For context, three years ago, investment firm Upper90 provided a $200 million loan to cloud provider Crusoe backed by 20,000 GPUs, but such deals haven't scaled until now.
**Standardizing GPUs as Stable Assets**
With just 14 employees and $3.5 million in venture funding, Barkr values GPU clusters similarly to how professionals value commercial equipment: by estimating the market resale price if the borrower defaults.
"We are turning GPUs into a standardized, low-volatility asset," Galbraith said. "When the insurance industry looks at assets and risk, the more uniform, conventional, and stable the price, the more readily it is accepted."
He points out that while rental rates for AI chips are volatile and rising, secondary market resale prices for the physical cards are far more stable. GPUs depreciate, but in a clear, predictable pattern—a trait highly valued by insurers and lenders. For example, NVIDIA's H100 chip launched in 2023 at around $30,000; its current resale price is between $25,000 and $30,000, well below a previous peak over $40,000.
Galbraith's background is in asset valuation, having worked for a global insurer appraising art and luxury goods for high-net-worth clients in the early 2000s. The company name, Barkr, pays homage to this focus, referencing the late host of the classic game show "The Price Is Right," Bob Barker.
**NVIDIA's Quiet Endorsement**
Following the forum, Galbraith visited NVIDIA's headquarters in Santa Clara, meeting with NVIDIA's VP of AI Infrastructure, Nico Kappes, and members of its startup investment team.
NVIDIA recognizes the significant market opportunity in facilitating chip financing. According to Galbraith, NVIDIA personnel have proactively connected him with lenders interested in the GPU financing space.
NVIDIA's support is driven by two key motives: first, to demonstrate through a neutral third party that GPU assets retain value longer than some critics claim; second, to open capital markets financing, allowing companies beyond top-rated tech giants to purchase GPU computing power at favorable rates.
"We are solving a problem that NVIDIA cares deeply about," Galbraith stated.
**Other Industry News**
* New Era Energy has announced an all-stock acquisition of Dominion Energy valued at $66.8 billion. The combined entity, with a market value nearing $250 billion, would create the world's largest utility company. * Grid intelligence company GridCare has closed an oversubscribed $64 million funding round with new investors including Sutter Hill Ventures and John Doerr. The company uses advanced AI simulation to identify underutilized power within electrical grids—which often operate at only about 30% capacity—to help power AI data centers. * Brent Mayo, former head of data center capital markets at consultancy Newmark Group, has left to join investment firm DigitalBridge Group.