Setting a New Bar for AI Infrastructure

Google’s Water Pledge

A typical mid‑sized data center uses something like 300,000 gallons of water a day. That’s roughly equivalent to the amount used by 1,000 U.S. households. And large facilities—the kind we are told will be built in significant numbers—can use millions of gallons per day.

AI is already the most rapidly adopted general-purpose technology in history. The data centers are rapidly following suit. Since realistically, usage will not abate anytime any time soon, we can expect the amount of water usage to soar and quickly.

That’s, of course, cause for concern.

For serious people in the tech and NGO community who prioritize making meaningful progress above ideological purity, the important question now isn’t whether to build new data centers. It’s how we should be building them. In some places with extreme water stress or very weak governance or critically important ecosystems, the right answer may be not to build at all. But for most projects that cross the desks of corporate deal teams and NGOs, the more productive question is: “Under what terms do we say yes?”

Water is a good place to start.¹ The volumes are big enough to be controversial—especially in places like Arizona or West Texas where even disciplined water use can endanger finite aquifers—but likely can be managed if we are careful and transparent.

On this front, Google made a positive announcement this past week. It pledged to replenish more water than its data centers consume across its portfolio by 2030. To do this, the company says it will avoid water‑intensive cooling in water‑stressed regions, prioritize reclaimed wastewater where possible, fund local water infrastructure, and—crucially—disclose its water use annually at the site level.

Google is not alone in making this kind of commitment. Microsoft and Meta have made similar pledges, and just last week four hyperscalers (Google, Amazon, Microsoft, and Meta) jointly formed the Data Center Innovation Initiative, a coalition focused on sustainable data center technologies. In other words, Google’s announcement is one in a string of indicators that sustainability performance is becoming a competitive baseline.

That’s good news, but it raises the stakes for scrutiny. And the differentiator will be the credibility of such commitments.

On Disclosures

Google’s announcement made me think of two imperatives we talk about a lot at the Instigator and that apply equally to the tech hyperscalers as to the NGOs trying to keep them in line.

1. Provide / Demand Transparency

Some of Google’s pledge is really about codifying their current practices. But turning these into public, quantified commitments matters. It creates a baseline for peer companies and NGOs. It also creates opportunities for bankers, communities, and policymakers.

The part of it I like best is the commitment to annual, site‑level disclosure of water use. Without that, everything else is essentially marketing. You can’t set standards, hold anyone accountable, or build community trust if nobody knows exactly what is being withdrawn, from where, and under what conditions.

2. Ask / Address the Hard Questions

We, the interested community at large, should not be passive recipients of these commitments.

Google points to 165 stewardship projects in 97 watersheds, expected to replenish roughly 19 billion gallons of water a year by 2030—equivalent to the annual cooling needs of many mid‑sized data centers.

That sounds promising. But “water-positive” is a squishy term. Most replenishment comes through projects like agricultural efficiency, wetland and forest restoration, and urban water upgrades. The benefits are often modeled in a different part of the basin than the facility itself. For example, a data center drawing from a stressed aquifer in Phoenix won’t be meaningfully offset by a wetland restoration 100 miles away. Such details really matter. That’s why people on corporate sustainability teams and at NGOs should treat Google’s pledge as a starting point and then scrutinize the details.

There are equity dimensions to consider too. Data centers are disproportionately sited in communities with less political power, often rural and lower-income. When a large facility draws heavily from a local aquifer or municipal water system, it competes with farmers, households, and small businesses that have no seat at the permitting table. Water-positive accounting that looks good at the portfolio level can mask real harm at the local level. That’s why site-level disclosure isn’t just a technical nicety; it’s a matter of basic fairness.

We should think critically and ask hard questions like: How exactly is “water-positive” defined here? In which watershed? Over what time frame? And who verifies the numbers? The verification gap is big, and regulators and NGOs should figure out how to fill it.

Transparent disclosure requires:

• Site‑level data on withdrawals and discharges

• Clear identification of sources (potable, reclaimed, brackish, etc.)

• A standardized efficiency metric that allows comparisons across facilities

Following these imperatives is not an extra burden; it’s a business process. Data center developers regularly encounter permitting friction and community opposition centered on water and energy. Investors are beginning to press companies like Amazon, Microsoft, and Google for more clarity on these impacts. Firms that can produce credible, consistent water data should be able to get projects approved faster, with fewer surprises and lower political risk. Transparency here is not just virtue signaling. It’s a smarter way to do business.

If you are an executive at a hyperscaler, this is the level of transparency you should insist on from your own teams and suppliers. If you are at a major financial institution, this is also what you should ask for in due diligence and ongoing monitoring. And if you work for a leading NGO, this is the baseline you should expect before treating any “water positive” claim as serious.

Moving from Pilot Pledge to Industry Standard

Google’s move is a useful start. But a single corporate pledge doesn’t solve a systemic problem. If you work in or around this space, you have a role to play in turning this into the new normal.

For hyperscalers:

If you’re at Google, Microsoft, Amazon, Meta, or any other company pushing big AI workloads, you should treat “water positive by 2030 with site‑level disclosure and a bias toward non‑potable sources in stressed regions” as a minimum operating standard, not a stretch aspiration.

More specifically:

• Push your organization to publicly commit to at least water‑neutral, ideally water‑positive performance in every water‑stressed basin where you operate.

• Require a clear water strategy—source mix, efficiency, replenishment projects, verification—for every major data center investment, reviewed at the same level as financial and cyber risk.

• Use your buyer power: require contractors and colocation partners to meet the same standards, not just your wholly owned facilities.

If you are a senior executive, you should assume that in a few years, regulators and investors will expect this as a baseline. Getting ahead of that curve is both environmental prudence and competitive advantage.

For NGOs:

This is a huge opportunity to shape what “good” actually means, rather than just reacting to whatever companies announce.

Three priorities stand out:

1. Help define “water positive” in real, science‑based terms, watershed by watershed. That means getting specific about baselines, additionality, and verification.

2. Build practical guidance for communities and policymakers: short, clear checklists that can be used in real permit hearings, not just long reports.

3. Play the role you’re uniquely good at: watchdog or partner.

When I was at The Nature Conservancy, I saw up close how these roles can reinforce each other. Often, activist groups like Greenpeace would campaign against companies or publish tough reports on specific projects, including our’s. In our case, the criticism stung, but it highlighted real shortcomings and pushed us to improve. We all benefit from watchdog NGOs willing to call out weak performance and fuzzy accounting. And we also need pragmatic NGOs willing to sit at the table with companies that are prepared to do better, and help hammer out stronger standards and verification.

If you work at TNC, EDF, or similar organizations, Google’s pledge is a chance to say: “Good. Now here’s how we’re going to measure and improve it, and apply the same bar to everyone else.”

Others have key roles to play too, of course. Communities have significant leverage (and are now doing a good job of using it), financial institutions can push companies hard to do better, and ultimately, we need legislation that makes commitments like this mandatory. There are opportunities for all of us to encourage these different actors to do more.

I hope that Google’s pledge has a ripple effect. But we should all do our part to encourage this by insisting that any company hoping to ride the AI wave earns its place in the watershed—in terms that make sense not only for their shareholders, but for communities and ecosystems too.

Onward,

1. There are other/bigger environmental issues with data centers beyond water that also need to be addressed. These include: energy use and carbon emissions, land use and habitat damage, heat discharge, grid reliability and strain on local infrastructure, and noise and light pollution.