We can finance compute. Why do we struggle to finance nature’s capacity?


· 6 min read
This is article 7 in the Environmental Infrastructure series. The previous piece looked at what happens when financing exists but the capability to deploy it doesn’t. This one turns to a new lens: what the compute boom’s financing ambitions might teach environmental finance about sustaining capacity over time.
NVIDIA and six major financial institutions have announced plans to mobilise more than $500 billion for AI infrastructure over time. The ambition is to make computing capacity easier to finance against the services customers are expected to buy. It is an extraordinary effort to organise capital around something the economy expects to need.
Our economy already depends on another kind of capacity: watersheds that support reliable water supplies, wetlands that accommodate floodwater and mangroves that buffer waves. We recognise these functions as economically important yet often struggle to arrange dependable funding to maintain them. What accounts for the difference, and what might the effort to finance compute teach us about closing it?
Computing capacity has customers who can commit to paying for access. Ecological capacity often supports many beneficiaries without any one of them being responsible for its upkeep. A port, a utility, and a coastal community may all depend on a functioning landscape while treating its maintenance as someone else’s expense. Recognising that dependence has economic value does not, by itself, produce a funding commitment.
The debate over “compute as an asset class” offers a timely opportunity to examine this problem. Compute is the processing capacity supplied by chips, networks and data centres. Whether it merits a separate asset-class label remains contested, and the $500 billion figure represents an ambition rather than deployed capital. The useful question for natural-capital finance is how comparable effort could secure long-term commitments to maintain ecological functions, alongside payments for restoration and verified improvements. Payments for ecosystem services have a long history; the comparison invites us to reconsider what those payments should sustain over time.
Computing capacity earns revenue when customers use it. A wetland’s protective capacity can remain valuable even during a year without a major flood, because maintaining it helps a community prepare for future storms. Yet that contribution is harder to demonstrate than hours of processing delivered to a paying customer. How should funding recognise the value of keeping protection available when the event it is intended to buffer may not occur for years?
Although estimates will carry uncertainty, we can assess that contribution and monitor the conditions that support it. A persuasive valuation helps establish the case for protection, but it still leaves someone to make a budget decision. If maintaining ecological function is treated as optional while the assets depending on it are treated as essential, better estimates of its value may change little.
A customer reserving processing capacity makes a commitment against an expected future need. Could a city and its major infrastructure owners make a similarly durable commitment to maintaining ecological protection? Imagine them jointly funding local wetland stewardship over a decade, with independent monitoring of conditions linked to flood-buffering function. The important change would be the continuity of the commitment: protection would have a funding basis through quiet years, rather than having to make its case again at the end of each grant.
Such an agreement could never guarantee protection against flooding. It would fund continuing stewardship through changing conditions, extending responsibility beyond completing a restoration project to maintaining its function.
Public institutions would remain central. Essential need does not automatically create paying demand, and a household’s inability to contribute does not diminish its need for protection. Collective arrangements can establish who contributes, on what basis and in whose interests, while public funding covers benefits that cannot reasonably be charged to individual users.
Dependable funding can, in turn, support upfront finance. Blue Forest’s Forest Resilience Bond demonstrates how such commitments can be assembled: investors provide money for restoration, while public agencies and other beneficiaries commit funding that supports repayment over time. Its first project, Yuba I, launched in 2018, completed restoration work in 2023 and returned investor capital in line with expectations. It shows how separate interests in a functioning landscape can become a funded commitment, providing a foundation for continuing payments to maintain ecological capacity.
Such agreements need not make ecosystems interchangeable. A wetland protecting one community cannot be substituted by restoring equivalent acreage elsewhere. Greater consistency in project preparation, contractual provisions and reporting could nevertheless reduce transaction costs while retaining locally appropriate ecological measures. A project can become financeable through a credible local agreement without creating a globally tradable unit of resilience.
Technology companies are already helping fund ecological restoration. In August 2025, Peatland Finance Ireland announced that Google, Microsoft and Meta had pledged more than €3 million to restore up to 450 hectares of degraded Irish peatlands. The initiative envisages verification under the Irish Peatland Standard. Google subsequently described support for Wicklow Mountains restoration expected to improve water-holding capacity, biodiversity and local flood resilience.
These are commitments and anticipated benefits, rather than evidence of completed restoration. The next test is whether maintenance funding, verification and responsibility endure after initial restoration ends. That will determine whether the commitment sustains ecological function over time.
For all its associations with the cloud, computing remains grounded in particular places and their water systems. Data centres depend on reliable electricity and cooling, and the International Energy Agency identifies water stress and physical risks among the sector’s challenges. Where evidence establishes a local benefit, operators have reason to join other beneficiaries in funding watershed management. These contributions belong alongside efficient operations, responsible siting and respect for existing water users; a replenishment claim alone cannot establish local water security.
Durable ecological agreements must sustain the people and institutions doing the work. Where Indigenous rights and jurisdiction are involved, arrangements must be developed through the relevant rights holders and governing institutions. A financial contribution confers no authority to determine a territory’s future. Nor does demonstrating an avoided loss automatically produce cash to repay a loan; financing still depends on someone being willing and able to fund the work.
The uncertainties surrounding compute finance reinforce the need for care. Demand may disappoint and equipment may lose economic value faster than anticipated. Financing terms must reflect those risks, just as ecological agreements must accommodate changing conditions and the cost of continuing stewardship. Even the compute market is still working through differences in service quality and contract terms. Its experience suggests that building confidence in delivery is part of making capacity financeable, rather than a task that must be perfected before financing can begin.
The case for maintaining ecological capacity should be considered before its deterioration becomes an operating constraint. A functioning watershed takes time to restore, and a community cannot commission a replacement wetland when a storm is approaching. The compute boom shows the effort investors will devote to securing capacity they expect the economy to need. There is a strong case for applying comparable effort to the ecological capacity on which it already depends.
illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.
The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.
Monty Simus

Sustainable Finance · AI
Praveen Gupta

Insurance · Climate Change
Charlene Norman

Sustainable Business · Corporate Sustainability
The Insurer

Insurance · Sustainable Finance
World Economic Forum

Sustainable Investment · Nature
CNN

Sustainable Investment · Sustainable Finance