The other side of the meter: Who’s financing the AI buildout and why it matters

Sustainability executives have learned to read the AI data center buildout at the meter: How many megawatts will it consume? How many gallons? How many acres? How much carbon emitted? These are the right questions, but they arrive too late. By the time a data center is operating, the decisions that determine its environmental impacts have already been made — not in a procurement meeting, but in a term sheet.

The sustainability community has largely ceded the funding side of the data center boom to the finance world. It’s worth reclaiming, because the numbers have grown large enough that how the build-out is financed now shapes what gets built, where and on what kind of power.

The four largest hyperscalers — Amazon, Microsoft, Alphabet and Meta — plan to spend roughly $725 billion on capital investments in 2026, up about 77 percent from last year. The overwhelming majority of that is on AI infrastructure. Goldman Sachs now forecasts more than $5 trillion of combined hyperscaler capital spending between 2025 and 2030. No corporate balance sheet, however rich, absorbs figures like these without strain. Amazon’s free cash flow is expected to turn negative this year. When Meta raised its capex guidance, its stock fell nearly 10 percent in a day.

So, the money is going off the balance sheet. This is the development that sustainability leaders should understand, because it changes everything downstream.

Need for speed

Meta’s two-gigawatt Hyperion campus in Louisiana will draw more power than many American cities. Meta did not build it with its own cash or with debt. It formed a joint venture with the private-credit manager Blue Owl, contributing a minority equity stake, and let institutional lenders including Pimco, Apollo and BlackRock fund the rest through roughly $27 billion in bonds. Meta keeps operational control and leases the campus back. The debt does not appear on Meta’s books. That structure is now the template: Close to $125 billion moved into similar project financing within a matter of months, and total data-center debt issuance nearly doubled in less than a year, to $182 billion.

Why should a sustainability leader care about the financial plumbing? Because the plumbing determines the outcomes that determine their success. CSOs are urged to interrogate AI suppliers on energy and water, but few analyze the financing terms that determine those inputs. “Speed-to-power” financing tends to favor on-site gas over slower clean interconnection; short-term debt to build a 20-year asset narrows operating choices.

Start with power. Financing rewards speed. A lender underwriting a multi-decade, fully amortizing bond wants the asset generating revenue on schedule, which means it needs power on schedule. Clean-power interconnection queues run five to seven years, while an on-site gas turbine takes 18 months to build. For lenders drafting a term sheet, that’s a no-brainer. The need for firm power now, not years from now, is quietly locking natural gas into 20-year assets across the country.

Next, consider duration. These are long-lived buildings financed over decades, using equipment — the chips inside them — that may be obsolete in three to five years. That mismatch creates refinancing pressure, which shapes operating behavior. The incentive to run hot, defer efficiency retrofits and maximize utilization is a financial incentive before it is an environmental one. A campus that must service its debt has little appetite for taking capacity offline to improve efficiency and reduce emissions.

Follow the financing

Finally, consider who holds the paper, because that determines who bears the risk, and where a sustainability leader actually has leverage. The bonds behind these campuses are increasingly held by pension funds and life insurers who seek long-duration yield. New York and Pennsylvania public pensions are invested in the same infrastructure fund behind Meta’s Louisiana project as well as several of Oracle’s data centers and other AI-related infrastructure. Wall Street has begun bundling this debt into securities. The exposure, in other words, is being distributed into long-horizon, fiduciary pools of capital — exactly the type of vehicles that the sustainable-finance community spends its days trying to steward.

And all the above is correlated. Much of this construction rests on a handful of tenants; OpenAI, for example, has committed to well over a trillion dollars of future capacity. If demand for AI services disappoints, or a marquee tenant stumbles, the losses will not stay contained; they will ripple across the insurers and pensions that funded them. Oracle bondholders have already sued over losses tied to its build-out. The last time capital poured into a physical technology bet on this scale — the fiber boom of the late 1990s — the majority of the physical infrastructure sat dark and unused for years. “Stranded asset” is a phrase the climate world coined. It applies here with equal force.

This is why the financing architecture belongs on the sustainability agenda. The environmental footprint and the financial risk of the AI build-out are being poured, together, into the same concrete — and both are being determined in documents most sustainability teams never see.

What to do about it 

So what can a sustainability executive do? Look upstream.

• Ask how the deal is powered, not just the building. The financing timeline will reveal whether clean interconnection is realistic or whether gas is the default.
• Ask to see those assumptions.
• Ask what the lease and offtake terms require of the operator, because those terms, not the ESG report, govern behavior once the plant is running.
• Join in the discussions while the capital structure is still being negotiated, when a power commitment or a residual-value guarantee can still be shaped, rather than after the bonds are placed and the incentives set in amber.

Sustainability leaders must recognize that the term sheet is now an environmental document, and to treat it as one. The profession spent a decade learning to audit supply chains it did not own. The capital stack is the next supply chain, and it sets the terms for all the others.