I don’t wish to avoid or withdraw from technological progress and have tried to embrace technology within my career and personal life. I often consider this quote from a book I read fifteen years ago; pre iPhone, long before generative AI, self-driving vehicles, delivery drones, gene technology, and endless data centers. If I were alive one hundred years ago, I would not be able to imagine the technological and societal changes driven through the introduction of new technologies. Perhaps we can take comfort in the great improvement in our quality of life. Though at the expense of our global-wide climate and environment for sure.
Artificial Intelligence (AI) And so, it is true, progress we now have. We have no choice but to accept the emergence of AI-enabled technology; it was always going to happen. Every time I google a topic in AI mode the response bar churns and provides an answer – with references, deep dives and related links. I like reading in-depth answers on my inquiries, or design drawings such as for home improvements, or bird identifications, and detailed information on a wide variety of research topics. But at what cost?
Let’s ask how much energy AI searches use. The electricity, gas and water it takes for data centers to process generative AI requests is huge in aggregate. Historically, AI queries have been estimated to require roughly 10 times more energy than a traditional, non-AI Google search engine query. Traditional searches primarily retrieve indexes and existing links from database servers. AI Searches enabled through complex Large Language Models (LLM’s) generate entirely new content, forcing the energy-intensive graphics processing units (GPUs) within the data centers to run high capacity as demand ever increases. We did the math on AI’s energy footprint. Here’s the story you haven’t heard. | MIT Technology Review Google’s AI search summaries use 10x more energy
A single text-based AI search or prompt uses roughly 0.24 to 0.3 watt-hoursof electricity. While this specific amount of power is relatively small – the equivalent of powering a low-energy LED light bulb for about two minutes or running a microwave for a fraction of a second, it scales up to very high KWH usage. If an AI search involves creating or analyzing multimedia, energy usage spikes exponentially. Image Generation uses significantly more compute cycles, consuming energy equivalent to charging a smartphone. Video Generation is the most demanding task of all. Generating a single short AI video can require enough electricity to run a household microwave for an entire hour. On an individual scale, a handful of AI searches won’t heavily impact electric demand. However, because hundreds of millions of people rely on these tools daily, the collective pressure on global power grids is severe. Data centers globally consume an immense portion of national grids-for example, accounting for roughly 17% of all electricity demand in Ireland, and more than 10% in several US states. What’s the impact of artificial intelligence on energy demand?
Creating User:حمزة مستقيم
Can we return to traditional searches? Not a chance. The best opportunity to at least partially address the rising energy footprint (and cost) of AI lies with research and application of efficiencies to manage these applications. Current research focuses on two main areas: technical optimization (making AI models process queries faster) and content optimization (how to structure information so AI can retrieve and cite it easily). Hardware Innovations related to memory & chip design can also drive efficiencies. Advancements including 3D chips, dedicated AI accelerators, and extreme chip cooling techniques enable data centers to run exponentially more computations per kilowatt. Early improvements have resulted in already impressive hardware and software gains, with inference costs falling over 200-fold and specialized hardware delivering up to 30X performance boosts. Still, I think that we are running uphill as demand load projections continue to increase.
Perhaps AI driven manufacturing efficiencies, transportation optimization, logistics and other yet unknown improvements can in theory drive the net energy demand lower. When could that occur? Read a fascinating article on the trends in data center investment, construction and energy demand with forecasts on future efficiency gains. The Rise of AI: A Reality Check on Energy and Economic Impacts – National Center for Energy Analytics
Queries (AI queries of course!) suggest that “AI-driven efficiencies in physical industries can eventually drive net global energy demand lower relative to computing costs, but this shift is unlikely to occur before the mid-to-late 2030s.” While AI computing costs and data center power consumption are spiking heavily right now, these physical sectors—manufacturing, transportation, and logistics—are the largest energy consumers on earth. Because their baseline energy footprints are massive, even single-digit percentage efficiency gains in these sectors can mathematically offset the massive electrical demands of the data centers powering them. The core reason AI can (eventually) reduce net energy usage is the scale asymmetry (for energy usage) between the digital brain (the data center) and the physical muscle (the global supply chain).
- Logistics & Transportation: AI minimizes empty miles, optimizes vehicle load distribution, and adjusts real-time routing. A 1% reduction in global maritime and aviation fuel via AI routing saves more energy than small countries consume.
- Smart Manufacturing: AI-driven predictive maintenance prevents massive industrial restarts, while computer vision drastically reduces material waste and scraping. Optimizing steel manufacturing systems by 5–10%, it saves gigawatts of heavy fuel and industrial electricity.
When Will the Tipping Point Occur? Energy analysts and institutional models suggest a net-energy savings turnaround will likely happen in phases:
| Timeframe | Phase | Net Energy Dynamic |
| 2026 – 2030 | The Deficit Phase | Net-Positive Demand. The scaling of Large Language Models (LLMs) and generative media requires massive, upfront electricity for training and inference. Data center electricity consumption is projected to double or triple. Infrastructure bottlenecks and slow corporate adoption mean physical supply chain energy savings cannot keep pace with tech expansion. AI integration is still localized during this phase, meaning the computing “cost” outpaces immediate physical energy savings. [1, 2, 3] |
| 2030 – 2035 | The Break-Even Era | Equilibrium. Massive hardware efficiency improvements (more tokens per watt) will take hold. Concurrently, autonomous trucking fleets, smart grids, and AI-optimized factories will scale globally, beginning to equalize the energy ledger. |
| Post-2035 | The Dividend Phase | Net-Negative Demand. The energy savings realized across macro-industries will scale globally. At this point, the operational fuel and electrical reductions in manufacturing and logistics will reliably exceed the energy required to compute those solutions. |
According to a comprehensive macro-economic impact study by PwC, the additional power consumed by data centers will be completely offset by the energy savings realized in the rest of the economy by 2035. At this point, the global economy is projected to experience a net energy demand reduction of 0.5% to 1.1% exclusively due to systemic AI efficiencies. [1] However, there are potential bottlenecks. The primary bottleneck to achieving this net reduction is the rebound effect (Jevons Paradox). As AI makes freight transportation, aviation, and factory production significantly cheaper and less energy-intensive, the market demand for those services may skyrocket. If the sheer volume of global manufacturing and shipping increases drastically because it is cheaper, it could erode the net energy savings achieved by the AI algorithms. [1, 2, 3, 4, 5] For net energy to drop, AI must be used to actively substitute or reduce physical processes rather than just inventing new digital related growth opportunities within the economy. Secondly there is a limit to the pace of change in the physical World: Software can update in seconds, but retrofitting factories with sensors, replacing standard commercial trucks with autonomous EVs, and transforming shipping networks takes decades. The lag in physical infrastructure deployment is the primary reason the net-energy dividend is delayed until the mid-2030s.
Does this fall into the category of “sit tight, nothing that I do in the short term will make a measurable difference” in future energy demand? Though as an individual I can’t alter these global trajectories; I will watch and learn and observe in real time the impacts on our society and economy, and consider these opportunities across my personal use of AI tools: Experts: You have some power to reduce your AI environmental footprint | AP News Let’s all do our part in support of mitigating these global impacts.
______________________________________________________________________________________
Electric powered vehicles (EV’s) – why don’t I own one? I have been following these technologies for years, yet I have still not pulled the trigger to purchase an EV. Five or six years ago I could convince myself that the charging infrastructure was not yet in place; I had concerns about charging times as well as vehicle reliability and tire / battery replacement costs. Perhaps all good reasons to hold off and allow the technology to mature.
With EVs, the challenge is called range anxiety, which is something that I can relate to as a person who always re-fills my gas tank as soon as it drops below the halfway full point. One full battery charge cannot take an EV as far as a full tank of an internal combustion engine (ICE) vehicle can. In recent years charging infrastructure and charging technology advancements have served to help lessen this anxiety. EV Charging Infrastructure: Trends, Requirements & Costs This is where location technology steps in. By providing accurate real-time data on the vehicle’s range, based on factors such as route topography and traffic conditions, location technology can help alleviate range anxiety. Moreover, it can guide drivers to the nearest charging station, considering variables like the remaining battery charge and speed limits along the route.
User: Nick-D
And what about cost? Studies show that while the purchase costs for EV’s may be higher, the full lifecycle maintenance and fuel costs offset these initial costs, and over the average seven-year life of a typical purchase one is expected to come out ahead on an expected value payout. OK, range improvements, cost of ownership, check, check.
Then I have this thought in mind that vehicle technology will change at too fast a pace and that any purchase will quickly fall behind capabilities of new innovations. In truth, my current seven-year-old Audi suffers from the same technology fate, as items like camera functions, lane and cruise control assistance have advanced significantly in the past several years. My control screen is ancient now, so I use my phone for navigation, and the audio system works just fine – so you adapt. Electric vs. Gas Cars: Is It Cheaper to Drive an EV? In the meantime, I have talked myself out of hybrids as have friends and recent press coverage. Which Is Better to Buy: A Fully Electric Car or Hybrid? – Kelley Blue Book
So now I am left with the choice of buying one more internal combustion powered vehicle or making the leap as I buy new in the next two years. I think that I am ready.
_______________________________________________________________________________
The protein dilemma: I have been trying to consume more protein in my diet, to increase muscle mass and maintain my strength. I serve myself cottage cheese daily, lean lunch meats when available, eggs and other occasional protein sources like fish and nuts. I do crave a thick rare steak or hamburger whenever I get the chance, though not daily.
Here is the issue: Ruminant meats like beef and lamb have the highest net CO2 emissions per gram produced (~50 kg CO₂e per 100 g protein), and in addition drive deforestation and biodiversity loss (e.g. across South America) FairPlanet+1. Plant proteins and alternatives can cut GHG emissions, land use, and water consumption dramatically —chickpeas emit about 1/20 of the GHGs of beef per kg of protein World Resources Institute. Sadly, I am not a big fan of beans or lentils. While still more resource-intensive than plant proteins, some animal sources have lower environmental costs:
- Poultry – ~85% fewer GHGs than beef per kg of protein World Resources Institute.
- Pork – Slightly lower emissions than beef, but still higher than plant proteins FairPlanet.
- Farmed Fish – Lower emissions than beef, but still higher than many plant sources FairPlanet.
Feed conversion efficiency (FCR): Chickens have a much lower FCR than cattle — they convert feed into body mass more efficiently, so less feed (and thus less energy) is needed to produce the same amount of protein Feed-to-Meat – Conversion Inefficiency Ratios – A Well-Fed World
Photo credit: Malcolm Paterson
I can also convince myself that the conditions of animal welfare in the beef, pork and chicken industries must also be improved, though a losing battle within industries that compete on efficiencies of production. What a world we live in. Through research I learned that some beef is produced through pasture fed sustainable farming, both here in the US and in countries such as Argentina and Uruguay. Why Uruguay Is a Global Leader in Sustainable Beef Production? – Farms in Uruguay driving efforts towards ‘carbon neutral’ beef | Dialogue Earth
So, I will remind myself to lean towards chicken or fish meals and I do when available, though an occasional Omaha Steak, especially prepared through a holiday cookout, may remain on my menu. I would like to believe that I am informed and self-limited, at least to a degree in my beef consumption.
________________________________________________________________________________
Amazon, what the heck: I don’t remember the exact date of my first purchase from Amazon, but it was shortly after they started their online bookstore in 1995. I was living in Rangely Colorado, a very remote town, and we were accustomed to ordering personal and household items through catalogs and by phone directly to suppliers. Having just recently discovered access to the internet, I asked a co-worker if it was safe to enter credit card information for an online purchase. He “assured” me that this type of transaction was generally safe. One quiet evening in the basement of my home I pulled up the title of a book on the Amazon site, entered a payment and received confirmation, followed several days later by a book delivery. My family was skeptical of sharing my card data online, as we had never tried this before. Thirty years later we have Amazon Prime, and any item in the universe can be ordered through the Amazon site, directly from them or through a reseller. Need a book, need a sparkplug, need shoes, how about a replacement gasket for your bathroom faucet? They are all just a few clicks away. Our purchases have become frictionless.
Amazon went public on May 15, 1997, at an initial public offering (IPO) price of $18.00 per share. If You Bought 1 Share of Amazon at Its IPO, Here’s How Many Shares You Would Own Now. Amazon’s market capitalization has grown from about $438 million at its May 1997 IPO to over $2.8 trillion today, representing a staggering increase of over 6,400×. This monumental amount of wealth created reflects its transformation from a niche online bookseller into a dominant global e-commerce and cloud computing giant. [1, 2, 3] A major contributor to this exponential growth has been the company’s ability to diversify its revenue streams well beyond retail—most notably into cloud computing (AWS) and digital advertising. The growth of Amazon has landed the founder, Jeff Bezos, as the third richest person in the world. History of Amazon – Wikipedia
Amazon procurement center Photo Credit: Auledas
Today, Amazon relies on algorithmic dynamic pricing, loss leading, and price matching to undercut competitors. This approach pressures small businesses by compressing profit margins, forcing reliance on ad spending to maintain visibility, and in some cases, restricting sellers from offering lower prices elsewhere.
How Amazon’s Pricing Works Dynamic & Algorithmic Pricing: Amazon uses AI to constantly adjust prices, sometimes multiple times a day, to match or beat competitors and stimulate sales velocity. [1, 2] Amazon often drops prices on essential or high-volume goods to attract shoppers, relying on auxiliary services (like Prime or AWS) to make up the difference. Visibility heavily favors products that offer highly competitive pricing, optimized shipping, and high inventory levels. Impact on Small Businesses To compete, small sellers often must match Amazon’s incredibly low prices, and with already thin margins, this leaves little room for profit. Algorithmic repricing tools adopted by many sellers trigger automated price wars, rapidly dropping the value of a product. To sell successfully on Amazon, venders are structurally prevented from offering lower prices on their own direct-to-consumer sites. If small businesses don’t participate in the aggressive pricing game, they lose product ranking and visibility, often forcing them to rely heavily on expensive Amazon ads to drive traffic.
Because competing solely on price with Amazon is difficult, many small businesses adopt specialized survival and growth strategies, such as selling unique, specialized, or custom products that aren’t easily commoditized by large manufacturers or attempting to drive organic traffic to standalone websites using platforms like Shopify or Squarespace to bypass Amazon’s heavy platform fees and regain control over pricing and customer data. Issue Brief: How Amazon Exploits and Undermines Small Businesses, and Why Breaking It Up Would Revive American Entrepreneurship | Independent Business
Over the years I have benefited from cost efficiencies in the price of the products I have purchased. What is wrong with that? There is an independent bookstore located a mile from my house. I can go in there and typically will have to order any book unless it is a current best seller, wait a week, and return for pick up. Or I can order through Amazon Prime and receive it on my doorstep. And this quirky little bookstore is destined to soon go out of business.
AI assisted searches, gas powered vehicles, beef purchases on demand, Amazon products; I live a very convenient life. Am I just swimming upstream against the tides of technology and progress if I make a conscious attempt to change the intentions of my purchases. Will many other buyers do the same? Or can I make small, informed choices to steer my footprint in the right direction, along with everyone else? I can think of several places to start.