Don’t Put All Your Plants in One Basin: Why company averages miss the chemical industry’s water risk

Ion Visinovschi and Sara Azeem

Water stress puts an estimated 17% of annual gross profit at risk across 24 major chemical producers, and just ten plants account for 38% of that exposure. Nearly half of the exposure is concentrated in four places: the Gulf, Belgium, China and India. The pressure is likely to intensify as AI data centres and new chemical plants compete for the same water. Investors need to look past company averages and ask which plants carry the risk, and whether new capital is being deployed with water in mind.

When drought left the Rhine unusually low in 2018, barges could not carry enough raw materials upriver to BASF’s Ludwigshafen site. Plants ran below capacity, and BASF put the hit to that year’s earnings at around €250 million. Water scarcity reached the income statement not through damage but through constrained supply and lost output.

Droughts like this tend to be treated as one-offs, but our first facility-level screen points to something more structural: chronic water stress, where demand presses against renewable supply year after year. Our scan covers more than 950 verified facilities in 62 countries, with an estimated rebuild cost of $1 trillion. They belong to 24 of the world’s largest chemical producers, which together account for around 15% of global chemical sales.

The headline finding is concentration. Just ten of the 951 plants we modelled account for 38% of the sample’s gross profit at risk from water stress, more than the roughly 900 least-exposed plants combined. Across all sites, we estimate around $38 billion a year at risk, about 17% of the sample’s annual gross profit: a yearly run-rate under today’s conditions, not a cumulative figure.

The risk is not only concentrated in a few plants; it is also shared by companies drawing on the same water sources, and may grow as new plants are built in water-stressed areas. Little of it shows in company reporting, and a thirsty new neighbour is now arriving in the same basins: the AI data centre.

How we measured it

We start from each company’s reported gross profit and allocate it across its plants in proportion to their estimated replacement value, as a proxy for each plant’s share of the company’s output. We then use WRI Aqueduct’s baseline water-stress score, which measures how much of the local renewable supply is already being withdrawn, to estimate potential production losses at each facility. The model applies loss curves that reflect how water-dependent and flexible different industrial processes are. We estimate gross profit at risk by applying the modelled production loss to gross profit allocated to each plant, then summing the estimates across all plants.

These are screening estimates, not forecasts. Plant-level output and margins are rarely disclosed, so a site’s real profitability may differ from this asset-based allocation, and the loss curves have not yet been calibrated against observed shutdowns. The figures credit no adaptation, such as desalination or recycling, and capture today’s chronic water stress only: acute droughts, floods and future changes in stress come on top. They show where to look harder, not what any single site will lose.

Ten plants, more risk than nine hundred

Exposure to water stress is extremely uneven. The ten most exposed facilities account for about 38% of total modelled gross profit at risk, more than the roughly 900 least-exposed facilities combined. The top 50 carry two-thirds.

Most sites have relatively low modelled exposure: 743 of the 951 facilities modelled carry less than $25 million of annual exposure each. Yet for four companies, a single site accounts for more than half of the group’s entire exposure.

The same skew runs across companies. Five of them, SABIC, Linde, BASF, Mitsubishi Chemical Group and Bayer, account for about half of the total, and SABIC alone for a quarter.

That is why company-level water metrics can mislead. Averaging across dozens of sites dilutes the few plants that matter, and a group-wide water policy says little about whether those plants keep running when water runs short.

Figure 1 · Planet Tracker facility-level water-stress model · 951 modelled facilities, 24 companies · baseline water stress

Similar size, different risk

Similar totals can hide very different risks. BASF and Linde carry almost the same modelled exposure to water stress, but for different reasons. What separates them is how much comes from sites under extremely high water stress, where demand already uses more than 80% of the renewable supply.

Figure 2 · Planet Tracker analysis · exposure from sites in extremely high water stress; asset value = estimated replacement cost of sample facilities in each zone (approximate; China: northern basins)

For BASF, the risk is already present, and investors should price it in today. For Linde, it depends on how water stress evolves: one large German complex accounts for around 60% of its exposure, and if that basin moved from high to extreme stress, its modelled losses would rise sharply.

Linde itself tells CDP that water scarcity does not meet its threshold for a substantive financial risk at enterprise level in the short to medium term, and Mitsubishi Chemical says none of its plants faces water risks that could adversely affect its activities. On the basis of group-wide averages and today’s conditions, that can be a fair reading; site by site, and over plant lives of several decades, it can look quite different.

Shared basins, shared losses

Owning many chemical companies offers less diversification than it seems when their plants depend on the same water sources. Four hotspots account for nearly half of the water-stress exposure we modelled.

The Gulf dominates. Extremely water-stressed sites across Saudi Arabia, the UAE, Qatar, Kuwait and Bahrain carry about $14 billion of annual gross profit at risk, spread across nine companies, including SABIC, Dow, Linde and Yara. Many rely on shared desalination and water-allocation systems, so one policy change, outage or regional shock could hit all of them at once.

Belgium is the surprise: despite its rainfall, demand from people and industry is so high relative to renewable supply that WRI counts it among the 25 countries facing extremely high water stress. Six of the companies we assessed have extremely stressed sites there; for Evonik, a single Belgian site accounts for more than three-quarters of its exposure. China, where such sites appear most often (above all in its northern industrial basins), and India complete the list.

These four regions hold about a fifth of the sample’s estimated replacement value but nearly half of its water-stress exposure. An investor or insurer with broad exposure to the sector is therefore more concentrated in a handful of basins than any single company report would suggest.

A thirsty new neighbour

The AI build-out is landing in the same regions. Seven in ten of our facilities in the four water-stress hotspots have an AI, hyperscale or cloud data centre within 100 km, and three in ten within 50 km. Proximity is not proof of competition, as locations are approximate and cooling designs vary, but it shows where to look.

Belgium shows the scale: Google’s Saint-Ghislain campus, its thirstiest in Europe, used around one million cubic metres of water in 2022, about as much as 10,700 local households use in a year. Our water-stress data are long-term averages that do not yet reflect the AI build-out, even as the IEA expects data-centre electricity use to more than double by 2030. Because modelled losses rise steeply once stress passes a threshold, extra withdrawals for data-centre cooling and power generation could tip nearby plants into much higher losses.

The irony is that AI depends on the chemical industry. Chip fabs¹ run on ultra-pure gases, wet chemicals and silicon wafers, products that several companies in our sample supply. Two industries that need each other may end up competing for the same water.

Tomorrow’s exposure is being built today

At least eight companies in our sample have facilities planned or under construction in places with high water stress or flood risk. In Saudi Arabia, one of the world’s most water-stressed countries, they include new projects by Chevron Phillips Chemical, LyondellBasell, SABIC and Air Products. Indorama Ventures, which already has 38 of its 84 sites in high-stress zones, is adding two more in stressed regions.

Chemical plants run for three to four decades. Site selection is the cheapest moment to deal with water risk, before it is locked in for a generation.

This moves the investor conversation on from “Does the company have a water strategy?” to three sharper questions:

1. Is water risk assessed when sites are chosen for new capital projects, or only after the investment decision?
2. For sites already in extreme water stress, how will the risk evolve over the asset’s life as climate change and new water users, such as data centres, add pressure?
3. How are the long-run costs of water risk, from insurance and mitigation capex to potential stranding, reflected in project returns?

Measure it where it sits

These are high level figures for chronic water stress alone. Acute droughts and floods come on top: flood, for instance, costs less in an average year but can do severe damage in rare extreme events. Even so, the picture is clear enough to act on.

Company disclosure will not close this data gap soon. Investors still cite data as a key obstacle to integrating nature risk, and the EU’s Omnibus I Directive, in force since March 2026, has dropped the sector-specific reporting standards that might have made water disclosure comparable. For chemicals, water risk is concentrated, shared and growing, and it has to be measured where it sits: at the facility.

Sources

• BASF, 2018 full-year results news release, 26 February 2019
• WRI, 25 Countries, Housing One-Quarter of the Population, Face Extremely High Water Stress, 16 August 2023
• Linde, 2025 CDP corporate response, 2025
• Mitsubishi Chemical, Sustainability Report 2025, 2025
• The Brussels Times, Google data centre in Belgium the most ‘water-hungry’ in Europe, 25 July 2023
• IEA, AI is set to drive surging electricity demand from data centres, 10 April 2025
• Responsible Investor, Nature and Investors Survey 2025: Results, 12 June 2025
• DLA Piper, EU Council approves Omnibus I Directive, 31 March 2026
• All other figures: Planet Tracker facility-level analysis of 24 chemical companies, 2026