Tech Disruptors: Myers on Turning Compute Into an Asset Class

Morgan Stanley’s global head of structured products and private capital markets says the market is working to make compute infrastructure a financeable asset class as capital markets fund AI infrastructure expansion. He also discusses how a deeper secondary market for GPUs and compute systems could support that financing; the provided article excerpt gives no figures or specific market impact.

Analysis

The investable mechanism is not simply more AI spending: financing against GPUs could lower the cost of capital for compute operators and broaden the buyer base, supporting near-term hardware orders. But collateral value depends on resale liquidity, utilization and the pace of performance obsolescence. A liquid used-GPU market would help lenders underwrite residual values while potentially shifting some demand away from new systems over time—an important second-order risk for hardware suppliers such as NVIDIA and AMD. Operators could gain financing flexibility, but financing does not solve power, utilization or customer-concentration constraints; it can instead convert those operating risks into leverage and refinancing risk.

Treat this as a market-structure thesis, not evidence that a mature GPU-backed financing market already exists. Over days, the podcast itself is unlikely to establish a durable catalyst. Over 1–3 months, verify whether lenders publish repeatable terms, haircuts and realized resale recoveries, and whether financings close beyond pilot transactions. Over 6–18 months, the key test is whether secondary-market liquidity remains robust through a GPU generation transition. The thesis weakens if resale prices or collateral advances fall, utilization disappoints, or financing costs rise faster than compute revenues. No direct trade is warranted from the supplied information alone.

Key Decisions for Investors

• No immediate position: do not treat discussion of a financeable asset class as proof of scalable GPU-backed lending or incremental orders.
• Add an alert for disclosed GPU-secured loans, leases or asset-backed deals; verify advance rates, pricing, maturities, covenants, collateral marks and actual resale recoveries before underwriting beneficiaries.
• If financing becomes repeatable, assess compute operators as potential beneficiaries only where utilization and contracted revenue support debt service; track refinancing exposure alongside capacity additions.
• Monitor used-GPU prices and availability against new-system demand. A liquid resale market may initially support hardware deployment but could later pressure new-unit volumes for NVIDIA and AMD; reassess on evidence of substitution or falling collateral values.