# Powering The Internet is Becoming a Big Problem ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem.webp]] **Creator:** The B1M · **Published:** 2026-04-20 · **Length:** 16:42 · [Watch on YouTube](https://www.youtube.com/watch?v=sLEnn-sgpIk) > Caption source: YouTube English (United Kingdom) manual captions via yt-dlp (`en-GB`; `subs/subs.en-GB.vtt`), timed lines parsed from the same track (TubeAlfred not needed this run). Figures and spellings (**3–4%** of global electricity; **>11,000** data centres in **174** countries; **>1/3** in the US; Virginia most, then Texas/California clusters; UK and Germany **2nd/3rd**; growth stratospheric after ~**2017**; installed capacity **>122 GW**; Hinkley Point C **3.2 GW** → **>38** HPC-sized plants (companion: "around 38 … if everything was operating at 100% capacity"); pre-AI **<100 MW** vs AI **500 MW–1 GW+**; **2030** > Japan's consumption, **2×** **2022**; AI ~**20%** of data-centre electricity end-**2024**, ~**half** by end-**2025**; IES case: **10→50 kW**/rack, **−95%** water, PUE **1.16**; Ireland **10** data centres ~**−40%** space-cooling loads; Virginia **≥25** projects cancelled in **2025**; **>2 billion** gallons of water in **2023**) cross-checked with The B1M's companion article, [The World's Biggest Building Boom Isn't What You Think](https://theb1m.com/video/data-centre-construction-boom) (byline Adam Savage, 20 April 2026), which names the two interviewees — **Alex De Vries-Gao** (founder of Digiconomist; researcher at VU Amsterdam) and **Shaolei Ren** (associate professor of electrical and computer engineering, University of California, Riverside). YouTube chapters used for section headings; they line up with the captions. Screenshots from the 720p stream (format 136). Host **Fred Mills**. Contains paid promotion for **IES** (free data-centre whitepaper — link in description); description also promotes Fred's book ***Mega Builds*** (pre-order). Additional footage/images: Bahnhof AB, Good Morning America, INSITE, KTLA5, TODAY and WFYI. ## Outline ### 0:00 Intro - **0:02** One of the fastest building booms the world has ever seen — described as "a global construction frenzy that shows no sign of slowing." - **0:14** Humanity now depends on large-scale **data centres**, and demand keeps rising; the strain on **power and water** is enormous, and people are noticing. - **0:30** Neighbouring homeowners say "**not in my back yard**"; opposition is forcing some tech giants to rethink where to build. - **0:39** **Alex De Vries-Gao**: the data centre industry is heading for **3–4%** of global electricity consumption — "a huge sector." - **0:46** AI is taking reliance to a new level; the sector is under pressure to clean up "— and fast." With buildings consuming as much energy as entire countries, what can be done? ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0005.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0022.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0040.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0060.webp]] ### 1:11 Rise of Data Centres - **1:16** What a typical data centre looks like in **2026**; if you live in or near a major city, there's probably one nearby. - **1:27** More than **11,000** worldwide across **174** countries; over **a third** in the **USA**. - **1:39** Big US clusters in **Texas** and **California** — but the state with the most is **Virginia**. The **UK** and **Germany** are "way behind" in second and third; the rest spread across the globe. - **1:56** Why they're everywhere: online shopping, banking, construction videos on YouTube — and doomscrolling, video calls that could've been an email, and memes that refuse to die. ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0076.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0090.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0112.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0130.webp]] ### 2:23 Energy Use - **2:23** Data centres weren't always omnipresent — "we can't do a video on data centres without digging into some… well, data." - **2:39** Graph of global **installed capacity** (total power available for equipment) over **20 years** — growth goes "stratospheric" after about **2017**. - **2:59** Total now **more than 122 GW**. For scale, **Hinkley Point C** (UK nuclear plant under construction) is **3.2 GW** → you'd need **more than 38** Hinkley Point C-sized plants just to power the world's data centres (companion qualifies: ~**38** at **100%** capacity) — and numbers and sizes keep soaring. ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0160.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0180.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0195.webp]] ### 3:34 AI's Impact - **3:34** **Shaolei Ren**: pre-AI data centres tended to be under **100 MW**; today AI data centres can easily be **500 MW** or **more than 1 GW**. Concentrating that power demand — and water demand — over a small area is a big infrastructure challenge. - **3:59** By **2030** data centres are expected to consume **more electricity than all of Japan** — **double** their **2022** grid draw. - **4:11** Main cause: the meteoric rise of AI. Fred's fake-out — "every time you chuck something into **ChatGPT**…" — "Just kidding. Some of the channels that imitate might do that, but we certainly don't." - **4:34** ChatGPT alone processes **billions** of prompts a day, so supporting infrastructure must scale fast. - **4:45** De Vries-Gao: AI was about **20%** of total data-centre electricity by end-**2024**, and could be about **half** by end-**2025** — "really big, really fast." ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0225.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0275.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0300.webp]] ### 5:11 The Construction - **5:11** Not just a big hall full of servers — the largest facilities today are considered **megaprojects**. - **5:27** Siting: good **fibre** connections, **water** sources and — for those handling several hundred MW — **high-capacity power grids**. - **5:41** De Vries-Gao: they consume as much power as entire cities; it's "really hard" to find grids with capacity even to start such projects — **power capacity becomes the number one concern**. - **6:02** Plenty of **space** for **expansion**, often soon after completion — the companies behind them are called **hyperscalers** "for a reason." - **6:20** **Climate and geography**: lower temperatures ease cooling; avoid extreme-weather and natural-disaster risk. - **6:35** Downtime is disastrous — "a lot worse than… missing out on the latest TikTok craze": financial losses, service failures, reputational damage. ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0320.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0345.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0365.webp]] ### 6:54 Cooling - **6:55** Racks run **24/7**; cooling is a key priority, usually one of two ways. - **7:03** **Air cooling**: cold air circulated into racks, normally through the floor, waste heat exiting the top — effective, but consumes massive electricity even with some air recycled. - **7:23** **Liquid cooling**: water or coolant carries heat away — less electricity and often less space; increasingly common in the AI era. - **7:40** Flip side: the water has to come from somewhere. **Ren**: mostly from **municipal** supplies, stressing many public water systems, especially in the US — most are small or medium-sized without much surplus for large industrial users. ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0420.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0445.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0480.webp]] ### 8:06 What Can Be Done? - **8:11** Many companies are prioritising **liquid cooling**, investing in their own **renewables**, and **funding grid upgrades** so they don't drive up residents' bills. - **8:31** Several of the biggest tech firms have partnered with **nuclear** providers for reliable low-carbon power; some are **restarting previously closed plants** (on-screen: a news report on the **Three Mile Island** reactor restart to power AI) or betting on not-yet-arrived tech — **nuclear fusion** and **small modular reactors**. - **8:52** In **Europe**, some build **underground** — inside an active **mine** or former **bunker** (footage credit: **Bahnhof AB**) — good for security and cooling. - **9:06** These options aren't viable for most; delivering future-proof, super-efficient AI-era facilities (new builds and upgrades) at speed is hard to even begin. - **9:36** Sponsor segment — **IES**: physics-based building performance simulation and **Digital Twin** technology; whole-facility, full-lifecycle modelling (dynamic IT loads, rack layouts, advanced cooling, local climate); AI-ready design strategies; **30+ years** of consulting. Case studies: one hyperscaler retrofit with liquid cooling went from **10 to 50 kW per rack**, cut water use **95%**, achieved **PUE 1.16**; in **Ireland**, **ten** data centres cut space-cooling energy loads by ~**40%**. New free whitepaper on dynamic simulation enhancing **CFD** modelling (QR / description link). Paid promotion through ~**11:45**. - **11:45** Beyond racks and cooling: security, networking, power distribution — "might sound dull, but are essential," each with its own energy needs. ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0525.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0545.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0670.webp]] ### 12:05 The Opposition - **12:05** Giant facilities popping up everywhere have consequences for surrounding infrastructure — and communities; opposition grows especially where resources are already limited. - **12:23** De Vries-Gao: it depends on local conditions — water scarcity, droughts. A big new data centre consumes lots of electricity, and generating that electricity consumes lots of water, adding to local water stress. - **12:42** Highest-concentration areas are resistance hotspots. **Virginia** — nicknamed "**Data Center Alley**" — saw **at least 25** projects reportedly cancelled in **2025** after local opposition. - **12:57** Hundreds of data centres there used **over two billion gallons** of water in **2023**; complaints about noise and pollution, since many still rely on fossil-fuel grids and **diesel backup generators**. - **13:15** Elsewhere in the US, residents fear rising energy bills and insufficient mitigation (news clips: public health, local environment, "a huge energy drain," "We are not going to pay double the amount of our utility bills"). ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0735.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0770.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0800.webp]] ### 13:40 The Future - **13:40** Data centres will only get bigger and more widespread, faster than anything seen before. - **13:52** De Vries-Gao: digital growth isn't new (streaming and cloud took off last decade), but with AI "**bigger is better**" — efficiency gains in hardware and software have been spent making models even bigger and more widely adopted; "that's where things really break apart from what we saw previously." - **14:30** The pushback shows things must change — those affected by facilities "almost in people's back yards" can't be ignored. - **14:42** Ren: community opposition is a real concern to address proactively; accommodate industry demand from the infrastructure side. "For power, if you have money, you just build power plants, but for water… [it] is naturally constrained." - **15:03** Yet Ren thinks media focus on water use is "a bit overblown" — not a major issue everywhere; where water is available, it's still the best way to tackle cooling. Many data centres use water only during the hottest weeks/days, so annual use can be low while summer use is high. "We need to use water responsibly… the benefits of using water should not be denied." ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0830.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0860.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0900.webp]] ![[images/Powering-The-Internet-is-Becoming-a-Big-Problem-t0935.webp]] ### 15:50 Conclusion - **15:51** Like the technology itself, data centre construction has become a complicated subject: we can't live without them and they aren't going anywhere — quite the opposite. - **16:03** To keep building these megastructures at ever-greater scale, "we really are going to have to work out a few bugs in the system first." - **16:15** IES whitepaper reminder and thanks to sponsors; subscribe to The B1M. ## Things mentioned ### Projects, facilities and technologies - **Data centres** / AI data centres (pre-AI **<100 MW**; AI **500 MW–1 GW+**); **hyperscalers** - **Hinkley Point C**, UK (nuclear plant under construction, **3.2 GW** — scale comparison) - **Three Mile Island** nuclear plant, Pennsylvania (on-screen news: reactor restart to power AI) - Nuclear fusion; **small modular reactors** (SMRs) - Underground data centres in Europe — active mines and former bunkers (**Bahnhof AB** footage) - Air cooling (raised-floor cold air) vs **liquid cooling** (water/coolant) - Diesel backup generators; fossil-fuel-reliant grids - **ChatGPT** (billions of prompts a day) - IES **Digital Twin**, building performance simulation, **CFD** modelling; **PUE** (power usage effectiveness) ### Places - USA (over a third of the world's data centres); **Virginia** ("Data Center Alley" — most data centres; **≥25** projects cancelled in **2025**); Texas; California - UK and Germany (second and third by count) - Japan (2030 consumption comparison) - Europe (underground data centres); Ireland (IES case study — **10** data centres) ### Organisations and companies - The B1M / Fred Mills (host) - **IES** (Integrated Environmental Solutions — paid promotion; whitepaper) - **Digiconomist**; **VU Amsterdam** (Alex De Vries-Gao) - **University of California, Riverside** (Shaolei Ren) - OpenAI's ChatGPT; TikTok (aside) - Footage: Bahnhof AB, Good Morning America, INSITE, KTLA5, TODAY, WFYI ### Concepts - Installed capacity; grid connection capacity as the number-one siting constraint - Water stress from both direct cooling and power generation; municipal water systems with little surplus - NIMBY / community opposition; energy-bill fears; noise and pollution - "Bigger is better" AI scaling — efficiency gains consumed by larger models (rebound effect) - Responsible water use; seasonal (peak-summer) water demand ### Figures and facts - **3–4%** of global electricity (industry trajectory) - **>11,000** data centres; **174** countries; **>1/3** in the US - Installed capacity **>122 GW**; growth stratospheric after ~**2017**; **>38** Hinkley Point Cs - **2030**: more electricity than Japan, **2×** 2022 - AI share of data-centre electricity: ~**20%** (end-2024) → ~**50%** (end-2025, possible) - Virginia: **≥25** cancellations (2025); **>2BN** gallons of water (2023) - IES case: **10→50 kW**/rack; **−95%** water; PUE **1.16**; Ireland ~**−40%** space-cooling loads; **30+** years of consulting ## Biographies ### Fred Mills Co-founder and on-camera host of **The B1M**. Presents the data centre boom with humour (meme and doomscrolling asides, a ChatGPT fake-out jab at imitator channels), walks through capacity growth, AI-driven demand, siting, cooling, nuclear and underground solutions, community opposition and the future, and presents the mid-video **IES** sponsor segment. The description also promotes his book ***Mega Builds***. ### Alex De Vries-Gao Founder of **Digiconomist** and a researcher at **VU Amsterdam** (per the companion), known for tracking the energy footprint of digital technologies such as cryptocurrency and AI. In the video he projects the industry at **3–4%** of global electricity, estimates AI at ~**20%** of data-centre electricity by end-2024 and possibly ~**half** by end-2025, stresses that grid capacity is the number-one siting concern, links data-centre electricity to local water stress, and argues AI's "bigger is better" dynamic breaks from earlier digital growth because efficiency gains are reinvested in bigger models. ### Shaolei Ren Associate professor of electrical and computer engineering at the **University of California, Riverside** (per the companion), whose research focuses on the water and energy footprint of AI and data centres. In the video he contrasts sub-**100 MW** pre-AI data centres with **500 MW–1 GW+** AI campuses, explains reliance on small and medium municipal water systems, urges proactively addressing community opposition (power can be bought; water is naturally constrained), and argues media coverage of data-centre water use is often overblown because many facilities use water mainly on the hottest days — so water should be used responsibly, not denied. ### Adam Savage (The B1M) Byline on The B1M companion article *[The World's Biggest Building Boom Isn't What You Think](https://theb1m.com/video/data-centre-construction-boom)* (20 April 2026) — a B1M writer, not to be confused with the MythBusters presenter. The article confirms the interviewees' names and affiliations, qualifies the **38** Hinkley Point C figure (at **100%** capacity), and corroborates the video's data-centre counts, capacity, AI-share, Virginia and water figures. ### IES Paid video sponsor (Integrated Environmental Solutions) — physics-based building performance simulation and Digital Twin software plus consulting; promoted for data-centre design, liquid-cooling retrofits and a free whitepaper on dynamic simulation alongside CFD.