The Balance Sheet Era
Financing as part of the product
Thank you to Glenn Hutchins who has supported our learning and current thinking around new technology development and capital markets. Glenn’s decades of work give him a historical perspective on how new asset classes and financing structures take shape. Glenn is also very active across the AI ecosystem today, which also gives us an additional lens into how this market is developing. He is a day-one backer of our work at Logos and we are grateful for his support.
“You have four companies that essentially don’t need any net tangible assets. … This has become somewhat, you could call it an asset-light economy.”
- Warren Buffett, 2018[1]
A lot has changed in ten years. In stark contrast the prevailing sentiment in 2018, today’s AI and technology companies require tangible assets to deliver product. Infrastructure requires chips, data centers, and power. Applications commit to compute before they collect customer revenue. Hardware companies carry inventory and deployment costs before they deliver finished systems. Very little of this resembles the technology sector of the 2010s.
The financing markets around these capital needs are in their early innings. The AI infrastructure buildout is changing how the largest participants finance assets and growth. However, application companies (software and services, the users of this compute) as well as smaller compute developers and complementary hardware companies still rely on permanent equity to fund much of their compute, inventory, and deployment. The next evolution is for the capital markets to mature around these participants.
This is the Balance Sheet Era of the technology sector. Financing will become a core company-building skill, alongside product, distribution and deployment. The companies that develop it early will be able to meet more demand with the same equity and may turn their cost of capital into a durable advantage.
FINANCEABILITY AS A COMPANY COMPETENCY
Silicon Valley has spent two decades optimizing companies around the availability of equity. Private credit is more opaque and exists largely outside the venture ecosystem. Early transactions require founders and financing partners to educate each other on how the product, contract, and collateral work.
The capacity to finance the business with non-equity capital is built alongside the product, long before a company approaches credit partners. Contracts and operations should allow capital providers to isolate the assets and cash flows they can underwrite, while equity funds the remaining business risk. The right financing lets a company add capacity and recycle capital faster with a focus on increasing returns to equity. It can also become a product and company advantage, with increasing access to lower-cost capital delivering accelerated growth.
INFRASTRUCTURE FINANCING IS BEING BUILT IN REAL TIME
Tech-sector capital spending is contributing to nearly 50% of U.S. GDP growth, up from less than 5% in the first three quarters of 2023.[2] The first wave was funded largely by hyperscaler free cash flow and is now supported by debt and equity issuances.[3] Neoclouds are combining equipment facilities (GPU financing) with customer prepayments and long-term contracts to unlock third-party financing and drive down their cost of capital.[4] In this phase, capital is moving from the general corporate balance sheet toward more nuanced structures that combine specific assets and projects with corporate support.
This market is still in its infancy, and each financing remains artisanal. Our conversations with lenders reveals that projects are deconstructed to identify the assets, contracts, and counterparties that can support debt. As those units become standardized, financing can become faster and more modular, lowering the cost of capital.
APPLICATIONS AND EMERGING INFRASTRUCTURE
Application companies (software and services) and non-investment-grade hardware companies (compute, energy, robots, etc.) are relying on equity to fund much of the gap between committing capital and collecting customer payments. That gap increasingly resembles working capital, often tied to compute, equipment or deployments that should have collateral value.
Application companies must have compute available to deliver product and, given supply constraints, are required to make significant commitments and prepayments before customer revenue arrives. This dynamic is not going anywhere as the compute constraint will persist for years to come[5] (and probably get worse given increasing complexity of agent work[6]), which will drive compute contracts towards increasingly favorable terms for compute owners. For this cost to move away from pure equity financing, the underlying compute must become financeable. Exchanges, forward contracts, and futures are early attempts to create the fungibility and liquidity that lenders need to treat the reserved capacity as an asset rather than a stranded expense.[7]
Hardware companies (compute, energy, robots, etc.) face the same timing problem around components and deployment. A financing partner may not underwrite the startup’s balance sheet but may finance identifiable equipment or inventory and/or a contracted deployment. We are seeing the first green shoots here with new parties and banks starting to lend to these companies outside of traditional equipment leasing and venture debt[8]. Customer prepayments, off-take agreements, and manufacturer guarantees can make a contract or defined pool of assets increasingly financeable even when the company itself is not standalone.
Despite growing strain in parts of the market, global private credit is roughly $1.8 trillion, with $300+ billion of tracked dry powder.[9] The capital appears available but the market is still working out how to participate. Suppliers (NVIDIA[10], AMD) and large customers (AMZN, GOOG, etc.) are helping bridge that gap, using guarantees, prepayments and other credit support to bring third-party capital into non-investment-grade compute and energy projects. Critics have seen these arrangements as roundtripping (strategic players financing their own customers or suppliers). But that support may be what allows companies to develop the performance history needed for more independent capital to participate.
WHAT WE ARE WATCHING
First, how the multivariable features of chips and compute become normalized well enough for fungibility. We expect this market to become liquid, but the path remains unresolved. A tell here will be if application businesses or inference providers can start to finance future capacity without owning the underlying compute assets.
Second, credit performance (issuance, spreads, CDS) for the neoclouds and other AI infrastructure developers. It is a bellwether for emerging company financing.
Third, examples of companies defining a repeatable financing unit around a customer contract, equipment pool or deployment, and then streamlining the process across a fragmented set of lenders.
[1] Warren Buffett, remarks at the 2018 Berkshire Hathaway annual meeting, CNBC Buffett Archive. Source
[2] Michael Cembalest, “Smothering Heights,” J.P. Morgan Asset Management, January 1, 2026. Source
[3] Alphabet, June 2026 equity offerings; Meta, $25 billion senior notes offering, May 2026. [1] [2]
[4] CoreWeave, DDTL 4.0 and DDTL 5.0 financing announcements, March 31 and May 18, 2026. [1] [2]
[5] Combined 2026 capital-spending guidance of $735–$760 billion and capacity commentary from Amazon, Microsoft, Alphabet and Meta. [1] [2] [3] [4]
[6] “DeepSeek and the Impact of Cheaper Models,” May 9, 2025. Source
[7] NVIDIA Compute Match; Compute Exchange; Ornn Compute; CME Group and Silicon Data planned compute futures. [1] [2] [3] [4]
[8] Erebor Bank, trade-finance facility for Matium and $200 million credit facility for Valar Atomics, led by Erebor, as administrative agent, and J.P. Morgan, April 14 and August 4, 2026. Sources: Matium; Valar Atomics
[9] State Street Global Advisors, global private-debt estimate at Q1 2026; MSCI Q1 2026 dry-powder data; The Wall Street Journal on recent market strain. [1] [2] [3]
[10] NVIDIA, memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish compute-financing platforms designed to mobilize more than $500 billion of third-party capital, August 10, 2026. Source