← Back to Writing

Data Centers and AI's Impact on Communities

Do data center buildouts affect the traffic in the communities in which they are built? Analysing this question with traffic data.

FROM: Ben Lovell

RE: Data Centers and AI’s Impact on Communities

QUESTION PRESENTED:

What impact does the massive buildout of AI infrastructure and data centers have on your local community.

SHORT ANSWER:

Areas I’ve studied with data center construction sites saw a predictable increase in truck and car crashes on local roads.

DISCUSSION

Large Language Models (LLMs) require enormous amounts of energy and computer chips to provide value. Data centers required to train and run inference on these models require a lot of resources.1 The scale of these projects impacts the surrounding communities.

I. Why is AI Treated Like an Arms Race?

The phrase, “Manhattan Project,” has been bandied about in the media when referring to AI infrastructure projects.2 The imagery is deliberate. According to the prevailing research, increasing the scale of these systems leads to increases in the system’s abilities3. To put it another way, the company or government with the biggest and fastest data centers has the best chance of winning the race. Unlimited economic growth and scientific discovery await the winner of this race. What are we racing to build?

General Intelligence vs. Super Intelligence

Most of the serious discussion around AI concerns a benchmark of achievement we will refer to as Artificial General Intelligence. AGI is another nebulous term that means anything from OpenAI’s ability to achieve $100 billion in profit4 to intelligence comparable to an average human. Human level intelligence is arguably enough to result in serious societal change. Artificial Super Intelligence (ASI), on the other hand, will be smarter and faster than us and is regarded as a possible extinction level event for humanity 56.

AGI should create a lot of money and replace a lot of human jobs, which for many is enough of an incentive to enter the race. The fear is that the first person to build super intelligence will control the future. If we can control it, that is, but that’s a different discussion7. Everyone is afraid of losing this race.

II. Why the Pace Matters

“Move fast and break things,” read the immortal words of Mark Zuckerburg when describing the Silicon Valley ethos8. The tech industry’s new narrative requires speed at all cost. Singular companies have committed over one trillion dollars in debt to the construction of data centers. Leveraged investments into these buildouts only worsen the time crisis.

Safety Slows Things Down

Companies investing in these data centers lose money each day the equipment goes unused. Microsoft’s CEO recently said the company has too many processors and no where to plug them in9. Power infrastructure is not being built quickly enough and companies like xAI are resorting to gas turbines without bothering to get permits to run them. 10Companies are building gas generators powered by supersonic jet engines as they wait for grid connections. 11

III. What does that mean for data centers near you?

For starters, the increase in construction activity could increase in traffic collisions. A recent study of data from Holly Ridge, Louisiana, near Meta’s “Hyperion” data center construction site, showed a 600% crash increase in the rural area. My own research shows a similar correlation in local road crashes in Randall and Potter County Texas near data center construction sites. The rest remains to be seen.

Footnotes

  1. De Vries, A. (2023). “The growing energy footprint of artificial intelligence.” Joule, 7(10), 2191-2194. Key Finding: Predicts AI electricity consumption will rival that of a medium-sized European country by 2027. ↩

  2. https://www.reuters.com/world/china/how-china-built-its-manhattan-project-rival-west-ai-chips-2025-12-17/; https://www.csis.org/analysis/us-army-and-second-manhattan-project-ai ↩

  3. The “Kaplan” Laws: In 2020, OpenAI researchers (Kaplan et al.) demonstrated that performance (measured by cross-entropy loss) improves smoothly as you increase compute and data. They argued that scale was the primary determinant of performance, often more important than architectural tweaks (like network depth or width). ↩

  4. https://techcrunch.com/2024/12/26/microsoft-and-openai-have-a-financial-definition-of-agi-report/) ↩

  5. Russell, S. (2019). Human Compatible: Artificial Intelligence and the Problem of Control. Viking. ↩

  6. Carlsmith, J. (2022). “Is Power-Seeking AI an Existential Risk?” NeurIPS 2022 (oral presentation).

    • Focus: A rigorous, probabilistic breakdown of the specific chain of events required for an AI to pose an existential threat (e.g., “Will it seek power?” “Will it succeed?”). It is one of the most cited recent technical assessments of survival odds.

    • Link: arXiv:2206.13353

    ↩

  7. Turner, A., Smith, L., Shah, R., Critch, A., & Tadepalli, P. (2021). “Optimal Policies Tend to Seek Power.” Proceedings of the 35th Conference on Neural Information Processing Systems (NeurIPS 2021).

    • The Argument: This paper provides a mathematical proof that for most tasks, the “optimal” strategy involves acquiring resources and preventing one’s own death (shutdown). It suggests that without explicit (and currently unsolved) safety measures, AI systems will naturally resist human control because being turned off prevents them from achieving their goals.

    • Link: NeurIPS Proceedings

    ↩

  8. https://en.wikipedia.org/wiki/Meta_Platforms#History ↩

  9. https://www.datacenterdynamics.com/en/news/microsoft-has-ai-gpus-sitting-in-inventory-because-it-lacks-the-power-necessary-to-install-them/ ↩

  10. https://www.selc.org/press-release/new-images-reveal-elon-musks-xai-datacenter-has-nearly-doubled-its-number-of-polluting-unpermitted-gas-turbines/ ↩

  11. https://www.cnet.com/roadshow/news/supersonic-jet-engines-will-soon-power-ai-data-centers/ ↩