The Hidden Cost of AI - How ChatGPT is Exacerbating America's Looming Water Crisis

The Hidden Cost of AI: How ChatGPT's Iowa Origins Foreshadow Looming Water Shortages

In a revelation that has sent shockwaves through the tech world, Microsoft recently disclosed that one of the world’s most powerful AI supercomputers, built exclusively to train ChatGPT, is secretly located near Des Moines, Iowa.

This highly advanced system, shrouded in secrecy and possessing over 285,000 computing cores, was instrumental in developing OpenAI’s groundbreaking GPT-4 model that powers ChatGPT. While Silicon Valley has become synonymous with leading edge AI research, it turns out an unassuming Midwest locale is where this revolutionary technology actually came to life.

So how did Iowa of all places end up as the unexpected birthplace of ChatGPT? What drew Microsoft here was the region's abundant land and infrastructure to rapidly construct massive data centers. With its cool climate for much of the year, Microsoft could utilize outside air to naturally chill its AI systems, only needing water for cooling during the hottest summer months.

And cool them it does, using water on a mind-boggling scale. The AI supercomputer consumed 11.5 million gallons in just July 2022 alone. That's approximately 6% of all water used in the entire West Des Moines district, putting substantial strain on local residential supply.

In fact, this abundance of water may be the secret behind GPT-4's intelligence. Creating advanced AI requires analyzing absolutely massive sets of text data, a computing process demanding intensive power that generates huge amounts of heat. With Iowa's plentiful and inexpensive water supply, Microsoft could cost-effectively cool the AI system even during marathon training sessions.

So while Silicon Valley pioneered bleeding edge AI, good old fashioned Midwest values like industriousness and resource abundance enabled Iowa to nurture this technology to maturity. Microsoft and OpenAI have closely guarded details, but make no mistake, without Iowa's infrastructure and natural resources, ChatGPT would never have been possible.

Yet there is a catch. The exponential growth of data centers and AI systems like the one powering ChatGPT may be on a collision course with one of America's most severe yet overlooked emerging crises - widespread water scarcity.

The West and Southwest United States are already gripped by historic drought and water supply shortfalls. As climate change accelerates, groundwater depletion worsens, and usage grows, researchers warn that within decades over half the country could face regular water shortages.

America's outdated water infrastructure leaks trillions of gallons annually. Increasing periods of drought will further strain supplies, especially in fast growing states like Arizona, Utah and Texas. Yet thus far, the tech industry has given little consideration to its vast water appetite when expanding operations.

ChatGPT's Iowa origins reveal the vast quantities of water needed to develop powerful AI, and this demand will balloon as the technology advances. Each new AI system sucks up exponentially more water than the last during its power-hungry training.

With plans to build several new data centers in West Des Moines alone, Microsoft's presence is already commanding substantial local water reserves. As competitors like Google follow suit, small Iowa towns may become unexpected hotspots of intensifying water conflict.

Beyond Iowa, some project AI systems in America could be consuming over 5 trillion gallons annually by 2028 - enough to drain the entire lower Colorado River basin! Without strategies to curb usage, AI expansion could strain critical water resources to the breaking point.

Making matters worse is AI's appetite for electricity, which in the US often comes from water-cooled thermoelectric power plants. Experts estimate up to half of water withdrawn nationwide goes towards power plant cooling. Ramping up energy-hungry data centers will drive electricity demand, further stressing water supplies.

With the tech industry racing to roll out generative AI products and hungry for computing power, they have given little public thought to these urgent environmental constraints. AI shouldn't mean picking winners and losers in terms of whose communities get water.

Solutions exist, but require decisiveness and investment. Water efficient cooling systems, on-site greywater reuse, and shifting computing to temperate locales can greatly reduce data centers' water footprint. AI developers must also standardize efficiency benchmarks and optimize algorithms to use less power.

States projected to face future water problems should be cautious of giving data centers blank checks to tap local aquifers. Communities must assess if they have sufficient current and future water surplus before allowing unrestrained AI expansion.

This may mean saying no to arid regions powering further growth. But it protects citizens and ensures AI remains an engine of progress, not a drain on critical resources.

The tech world buzzes about GPT-4 and ChatGPT today, but may soon buzz about the next big AI technology labeled "Made in Iowa." As unexpected places shape the AI future, ChatGPT's Iowa roots are only the beginning. This heartland state that nourishes our bodies may now also nourish our minds.

Yet enjoying these fruits requires foresight and recognition from the AI industry that its water impacts are no longer negligible - they are an urgent priority. If we are wise, ChatGPT's parched origin story will become a cautionary tale of AI's hidden costs, not the norm. With American innovation and values, we can build an AI future that uplifts society, while also conserving the precious water we depend upon for life.