Silicon Valley utopians imagine AI solutions to ecological crisis, while being oblivious to the real material and ecological harms their fantasies wreak.


Jessica Silva was killed by the Mexican National Guard while protesting against water diversions to the United States. In 2020, long droughts had plagued the state of Chihuahua and endangered its population’s access to water. To show their anger and dissatisfaction, local farmers occupied La Boquilla Dam to close its gates and stop water diversions. They opposed President Lopez Obrador’s decision to continue diverting water resources to the United States in compliance with a 1994 treaty between the two countries that stipulates that Mexico must grant a fixed amount of water every 5 years. Despite the current climate emergency, the United States does not seem willing to revoke that treaty imminently.

Meanwhile, governments around the world and Silicon Valley’s Big Tech companies have announced artificial intelligence (AI)-based projects and large investments in sustainability to mitigate the effects of the climate crisis. For instance, Google has claimed to be “building a carbon-free future for all”; Amazon is allegedly “the world’s largest corporate purchaser of renewable energy”; and Microsoft has promised to “be water positive by 2030”. Yet, the three tech companies own the biggest cloud services and networks of data centres, which all consume large amounts of water to cool their servers. How does Chihuahua’s drought and local communities’ resistance relate to Big Tech companies, data centres and artificial intelligence? Disentangling this provides us with a way to grasp the ecological politics of Silicon Valley more broadly.

AI and its infrastructure

Designing an AI-based algorithm does not require particularly expensive infrastructures. Machine or deep learning algorithms can be programmed and processed on laptops. But if these algorithms are trained with large amounts of data or the algorithmic architecture is sophisticated, then it is necessary to use servers with greater computational capacity to store and process the information. To build these chip-based infrastructures, raw materials are necessary: iron, aluminium, minerals such as lithium, silicon or cobalt and water, among others.

Due to the current geopolitical instabilities, a supply chain crisis and climate emergency, there has been an explosion of resource nationalism in recent years. Governments have announced their resilience plans to guarantee their supply of raw materials and chipmaking. In February 2022, the White House announced in a statement its plans to reduce the USA’s dependency on foreign resources by securing ‘made in America’ supply chains for strategic minerals . In July 2022, the British government published a similar strategy aiming to “maximize what the UK can produce at home”. Still in 2022, Spain announced a European Union’s investment plan of over 12 billion euros for Spain to become a global reference for chipmaking and thus end the EU’s reliance on countries external to the Union. Meanwhile, the states from which most of these resources have been extracted, and which have suffered from the Global North’s voracious extractivism, have started to nationalise raw materials. Chile has proposed nationalising all the mines on its territory, and Mexico has passed a law prohibiting private industries from mining and exploiting lithium.

Given the current situation, the reliance on material resources to build the physical infrastructures of AI could affect its development and maintenance. The future of innovation and efficiency that many governments and private companies dream of runs into ecological and geopolitical limits. But AI does not rely on raw materials only during the construction of its physical infrastructures; it does so throughout its cycle. For instance, data centres and servers need large amounts of water to cool down. According to a study published in Nature in 2021, Google and Microsoft declared using respectively 15.8 billion and 3.6 billion litres of water.

We don’t know if these numbers are trustworthy. As a telling example, Microsoft has been involved in a scandal regarding the water expenditure of one of its data centres in the Netherlands. Whereas the technology company declared to the Dutch authorities that the centre consumed between 12 and 20 million litres, it transpired it was actually consuming 84 million. Meanwhile, in August 2022, Thames Water announced reviewing the water expenditure of data centres in London due to the drought scenario the capital faced that summer. While the average annual cooling system consumption of a small data centre in the US is estimated to be 25 500 000 litres, that of a person in Nigeria is 12 410 litres – 2 000 times less.

AI is also energy intensive. The more data to be analysed, the higher the energy consumption. More sophisticated algorithms, which need long computational time, consume even more. For example, it is estimated that training an algorithm to automatically produce text uses 190,000 kWh; that is, 120 times more than the average annual consumption of a household in Europe in 2020. To generate this energy, raw materials such as organic matter, uranium, coal or water, among others, are again needed.

Although some of the big tech companies claim that their energy is produced sustainably, the data shows another trend. In 2019, Greenpeace published a report about an Amazon Data Centre in Virginia (USA), which is considered to be one of the most important in Amazon’s global infrastructure. Greenpeace warned against the important growth in energy consumption in the region due to this data centre’s activities. Despite Amazon’s pledge to invest in “green” energy for this data centre, the reality is that its investment in fossil fuels has increased shamelessly. In 2021, data centres were estimated to consume 0.9-1.3% of global electricity demand. Given AI’s high energy consumption and the current energy crisis, the techno-optimistic dreams of governments and Silicon Valley’s companies could be dashed by the high price of energy.

AI and its carbon footprint

AI has also contributed to the climate crisis as illustrated by the rising concerns about its carbon footprint. To go back to my previous example, training an algorithm that automatically types text has the same carbon footprint as a round car trip to the moon. There are initiatives to calculate the carbon dioxide emissions of an algorithms based on the hardware, computational time and the cloud service employed. Despite the lack of transparency on how their carbon footprints are calculated, data centres are estimated to have emitted almost 100 megatons of CO2 in 2022. As a gesture to address these concerns and theoretically reduce greenhouse gases, Big Tech companies have started to invest large funds in an eco-washing strategy: carbon offsetting. Some of the most popular carbon offset projects are investing in renewable energy and tree plantations. This is how great techs secure the “carbon neutral” label. In 2019, Google announced that it had been carbon neutral for 12 years as it invested in more than 40 projects to offset its emissions.

But these types of projects and their environmental impact need to be questioned. This summer, 14,000 hectares were burned in a fire in Zaragoza (Spain) caused by a Dutch carbon offsetting company. The fire started when one of the company’s operators Land Life — which uses AI and drones for reforestation — set out to plant trees in the middle of July to whiten a company’s carbon emissions. This case calls into question the strategies that the capitalist system develops with the help of technology to continue destroying the planet. As Peter Kalmus, a NASA climate researcher, pointedly wrote on Twitter: “Remember, the reason carbon offsets are so popular with corporations is because they don’t really work, they’re just a cheap green wash.”

AI and climate justice

But resistance to the ecological programme of big tech is occurring internationally. The Thacker Pass Desert Communities in Nevada (United States) are resisting the technology industriesambition to extract lithium from the region. Following the U.S. government’s plan (which I mentioned above) to invest in national supply chains for minerals, this area is now endangered despite its natural and cultural richness. A similar fight around lithium is taking place in Spain: Extremadura is threatened by European extractive companies, which seek to exploit its mining wealth, disregarding the opposition of much of the local population.

The municipality of Cerrillos (Chile) has also resisted Google’s foray into its territory, in a struggle that offers a striking parallel with the Chihuahua community’s fight for its right to water. Cerrillos’ community organization and fight have affected Google’s plan to build a data centre in the area, which would require exploiting the local aquifer to source 14 million litres of water per day.

Mike Anane, journalist and climate activist, has reported the impact of illegal electronic landfills for years. When AI infrastructures are no longer useful, the parts that cannot be recovered and reused are illegally sent to landfills, like the ones found in Accra (Ghana). In his documentary “E-Waste Hell”, Mike Anane documents the impact these wastes have on bodies and territories. He mentions, among other, the contaminated rivers – whose waters were still clean 20 years ago – as well as the local populations’ health problems caused by the inhalation of toxic gas when burning screens, cables and computers. Although Europe has a legislative framework that prohibits the shipment of these wastes to other countries, the reality is that illegal shipments are common practice. These different examples show how the entire AI cycle has a climate impact beyond carbon emissions.

There is no algorithm that can bring water back into the Chihuahua region, put out the fires that will overwhelm Europe next summer, or achieve net zero emissions. In contrast, counter-power and resistance at the local level have proven to be more efficient than any line of code or policy target to adapt to the current climate emergency. Whereas Silicon Valley companies are striving to promote AI projects to mitigate the effects of climate change or invest in carbon offset projects, reality shows us another path. Many of these projects are empty promises, wasting immense amounts of public money to eco-bleach technology, capitalism, and AI in particular. But the organization and struggle of communities to protect what belongs to them remain.

The people of Chihuahua give us a lesson on how important it is to take care of our community and to organise ourselves to protect the right to life, water and more. It also shows us that power will stifle and will use its monopoly on the legitimate use of force, taking the lives of people like Jessica Silva’s, and of so many other fundamental rights defenders. The only way we can adapt to the devastating consequences of the looming climate crisis and, as political activist Sofía Castillo says, “prevent life resources from becoming ‘vile goods’ and ‘political boots’, is to stay together on the path for the common good”.


Ana Valdivia is a Departmental Research Lecturer in AI, Government & Policy at the Oxford Internet Institute (OII). Building on her experience as a mathematician and computer scientist, her interest lies in investigating the political, technical and environmental impact of AI. The research for this work was supported by funding from the Alan Turing Institute under the Post-Doctoral Enrichment Award (no. 2022PEA\100089).

This is the eighth contribution of PERC’s series on the Silicon Valley ideologyEach week for the next two months, experts from the fields of political economy, political theory, economic history, cultural studies, computer science and law will share their research perspectives on the recent trends that have animated the Silicon Valley bubble. If you wish to get involved or would like to pitch an idea for a contribution, get in touch with our editor Carla Ibled (c.ibled[at]