OpenAI CEO Sam Altman claimed during a September 2026 podcast appearance that producing a single California almond uses as much water as roughly 38,000 ChatGPT queries.
Here is the math.
The Bottom Line
- The Arithmetic Check: Altman’s cited ratio aligns with UC Davis estimates placing a single almond’s water footprint near 12 liters, divided by OpenAI’s reported 0.32-milliliter per-query usage.
- The Scaling Flaw: Comparing a marginal agricultural unit to a single digital prompt ignores billions of daily queries and power-intensive model training runs.
- Local Impact vs. Macro Average: Spreading digital water consumption across global averages masks severe localized resource draws on municipal supplies near data center hubs.
Decoding the Numbers Behind the OpenAI CEO’s Viral Claim
When an offhand remark on a podcast sparks global debate, financial analysts look past the rhetoric to examine the unit economics. According to coverage by explainx.ai, a viral screenshot circulating on Reddit on September 3, 2026, thrust Sam Altman’s water comparison back into the spotlight. Altman argued that public anxiety over data center resource consumption relies on outdated metrics, offering the almond comparison to illustrate scale.
The comparison traces back to operational estimates published by Altman in June 2025 via his essay, The Gentle Singularity. That text pegged an average ChatGPT query at roughly 0.000085 gallons of water, or about 0.32 milliliters. Meanwhile, academic benchmarks from UC Davis and the Water Footprint Network estimate that growing a single almond in California requires roughly 12 liters of water due to intensive irrigation requirements in arid conditions.
Divide 12 liters by 0.32 milliliters, and the resulting ratio lands right around 38,000 queries per almond. Alternative U.S. Geological Survey-affiliated analyses from 2019 point to a smaller footprint of 3.56 liters per almond, which yields roughly 11,000 queries. Either way, the arithmetic is defensible. But accuracy in division does not equate to equivalence in macroeconomic impact.
Data Center Infrastructure Versus Central Valley Agriculture
But the balance sheet tells a different story about how resources actually move through regional economies. The primary flaw in Altman’s comparison lies in denominator mismatch. Agricultural water use in California’s Central Valley forms part of an established, heavily scrutinized irrigation network supporting a mature commodity market. By contrast, generative AI infrastructure represents an escalating, concentrated capital expenditure cycle.
Data centers do not distribute their cooling needs evenly across a global grid. Instead, hyperscalers cluster massive server farms in specific municipalities where power and land are cheap. This local concentration can place severe strain on municipal utilities. As noted in industry research cited by explainx.ai, individual data center campuses can occasionally draw upwards of 25% of a small town’s total water supply.
| Metric Type | Resource Unit | Estimated Water Volume | Primary Geographic Impact |
|---|---|---|---|
| AI Query (OpenAI Estimate) | 1 ChatGPT Prompt | ~0.32 milliliters | Concentrated data center hubs |
| Agricultural Commodity | 1 California Almond | ~12 liters (UC Davis) | Central Valley irrigation network |
| Agricultural Commodity (Alt. Study) | 1 California Almond | ~3.56 liters (USGS-linked) | Central Valley irrigation network |
Averaging these intensive local draws across billions of daily internet prompts obscures the localized stress points.
The Regulatory Horizon and Market Implications
As artificial intelligence firms race to secure power and cooling capacity, the tech industry faces mounting pressure regarding resource transparency.
Ultimately, comparing a digital prompt to a nut serves as an effective rhetorical device, but it offers little comfort to municipal utility operators managing localized supply shocks. For institutional investors, the core metric to watch is not the water footprint of a single query, but the aggregate infrastructure buildout velocity and its friction with local resource constraints.
Disclaimer: The information provided in this article is for educational and informational purposes only and does not constitute financial advice.
