Today’s Contemplation: Collapse Cometh CCLXIII–We’re Saved! Artificial Intelligence.
One of the latest and greatest techno-fixes to our ecological overshoot predicament and its myriad of symptom predicaments–artificial intelligence (AI)–is another in a growing list of oxymoronic “solutions”, doing the exact opposite of “saving” our societies and planet. Our various “solutionists” are shouting from the rooftops incessantly about the benefits and wonders of AI. From “sustainable” resource management to “fixing” climate change to contributing trillions of dollars to the global economy, AI will bring prosperity and well-being to all with no negative impacts upon our environment; for example, see: here, here, here, here, here, here, here, here, here, and here.
humansintheloop.org
If only the universe worked in the manner that the masses are told it does by the elite and those that profit from such technologies. The marketing goes something like this: not only is perpetual growth a God-given certainty but with the help of AI human ingenuity and innovativeness will overcome finite resource concerns and provide humanity with everything it wants and needs. To infinity, and beyond!
I will grant the assertion that AI is disruptive as many of its cheerleaders suggest but not, I think, in the sense “futurists” claim. The following analysis via my techno-fix/salvation questionnaire argues that it is perhaps one of the most resource hungry endeavours to come along in a long time–especially in terms of energy–exacerbating significantly our predicaments, and not the messiah many have been holding their breath for.
For those unfamiliar with my questionnaire, I designed it as a critical evaluation tool that attempts to cut through the marketing hype around “solutions”–particularly technological ones–and assesses whether the “fix” actually addresses the ecological and social crises we have led ourselves into, or whether it serves merely to extend the systems that are causing our issues and exacerbating them (see: Website Medium Substack). It challenges the face value promises by probing four dimensions:
1. Narrative: Is the proposal being honest about the costs, or is it being selective and only highlighting supposed benefits while ignoring such phenomena as rebound effects or its overall addition to material throughput?
2. Biogeophysical reality: What is revealed about energy return, waste, material demands, and effects upon planetary boundaries when a full lifecycle accounting beyond just carbon is performed?
3. Viability: Can the “solution” come about or be maintained without externalised costs, speculative future “breakthroughs”, and/or significant hidden subsidies?
4. Social aspects: Who are the beneficiaries of the proposal, whose power is being reinforced, and is it challenging the perpetual growth paradigm or simply providing it with more “fuel”?
I like to believe this analytical questionnaire helps to separate what might be a genuine and transformative project from one that relies upon material- and energy-intensive “fixes” which end up being leveraged to reinforce status quo power and wealth structures, and generate an exacerbation of our ecological overshoot predicament and its various symptom predicaments.
If you’re new to my writing, check out this overview.
Let’s now take a look through this analytical lens and see what the evidence suggests about AI and its potential for achieving that long promised holy grail of true human sustainability.
Narrative
Drawbacks vs. benefits
The dominant narrative from industry and many governments highlights AI’s potential to solve complex problems — accelerating drug discovery, optimising energy grids, and improving climate modelling. Major environmental and social costs are acknowledged only in passing, then quickly reframed as engineering problems on the cusp of being solved (“more efficient chips”, “carbon‑free data centres by 2030”). The actual doubling time of data centre energy demand, the strain on water resources, the proliferation of e‑waste, and the de-skilling of cognitive labour are routinely downplayed. It is true that some narrow AI applications — such as DeepMind’s AlphaFold for protein structure prediction — deliver genuine scientific insight, but these isolated successes do not salvage the overall techno-fix nor remotely justify the infrastructure’s scale. They are the exceptions that prove the rule.
Replacement or addition
AI is almost never presented as a replacement for a destructive technology with a planned phase‑out; it is overwhelmingly an addition to total human throughput. In most sectors AI is layered on top of existing material and energy flows: AI‑optimised logistics still moves goods in diesel trucks, AI‑enhanced oil exploration accelerates fossil fuel extraction, generative AI bloats consumer demand for more digital content and devices. The Jevons paradox is operating at full force — efficiency gains are swamped by an explosion in usage.
Small‑scale benefits applied to a global scale
Pilot projects that show energy or material savings in a narrow domain (e.g., a 15 % reduction in cooling energy for one Google data centre) are routinely extrapolated to the entire system, ignoring the colossal embedded and induced costs of building the global AI infrastructure in the first place. There is no honest accounting that demonstrates a net planetary benefit when AI’s full lifecycle is scaled to billions of users and trillions of inferences.
Biogeophysical Reality
Lifecycle stages
An honest lifecycle analysis from cradle to grave is missing from public discourse and rarely fully disclosed by developers:
· Raw material extraction: Copper, lithium, cobalt, rare earths, high‑purity quartz, and ultra‑pure water for semiconductor fabrication.
· Manufacturing: GPU/TPU production involves extreme energy and chemical intensity; chip fabs are among the most resource‑hungry facilities on Earth.
· Transportation: Global supply chains for components, servers, and cooling equipment.
· Operation: Electricity for training runs (e.g., a single large model can emit hundreds of tonnes CO₂e) and, far more significantly, inference serving billions of queries daily; water consumption for evaporative cooling and humidification.
· Maintenance: Constant hardware refresh cycles (2‑4 years) driven by performance demands.
· Byproduct disposal: Toxic solvents, heavy metals, and persistent organic pollutants from semiconductor manufacturing.
· Decommissioning: Decommissioned servers and networking gear.
· End‑life disposal: Low recycling rates for complex electronics; most end up in informal e‑waste streams.
· Associated infrastructure: Fibre networks, subsea cables, backup diesel generators, grid expansions, new power plants (often fossil gas or even coal plants re‑opened to meet data centre loads).
No public document offers a full, independent, peer‑reviewed lifecycle footprint for the current AI system. Authoritative partial estimates — such as the IEA’s data centre energy projections, Luccioni et al.’s work on operational emissions, or the UN’s Global E‑waste Monitor — already point to impacts far larger than industry narratives suggest.
Net energy return
AI is a net energy sink, not an energy source. For an energy technology to support societal maintenance the minimum EROI considered necessary is ~10–14:1, and about 3:1 for basic survival functions. Admittedly, EROI is a metric designed for energy‑supply technologies, not information services. Yet the same logic applies: a society can only afford to invest large amounts of energy into a tool if that tool ultimately reduces the total primary energy required to maintain the society. AI, by every credible estimate, does the opposite — it consumes high‑quality electricity to produce digital goods whose net effect is to grow overall energy demand, making it a net drain on the finite energy surplus available for everything else.
As a pure service technology, AI yields useful information work, but when comparing the society‑wide energy investment (building and running the entire hardware/software stack) against the energy savings it genuinely displaces, the ratio is almost certainly well below 3:1, and likely substantially below 1:1 in many applications. In other words, we are burning more energy to accomplish tasks that were previously done with less energy, or generating new tasks that did not exist before.
Finite materials and supply bottlenecks
AI is critically dependent on high‑end GPUs and high‑bandwidth memory (HBM). The supply chain is already strained: extreme ultraviolet (EUV) lithography depends on a single supplier (ASML), advanced packaging faces capacity limits, and rare earth elements for magnets and capacitors are subject to geopolitical chokepoints. Copper for power distribution and data centre construction is encountering declining ore grades. These are not smoothly scalable resources; they show classic patterns of diminishing returns and increasing extraction costs.
Ecological blind spots beyond carbon
Most “green AI” talk focuses exclusively on CO₂ or electricity and ignores:
· Freshwater use: Data centres already compete with agriculture and drinking water in water‑stressed regions.
· Biodiversity loss: Mining expansion for AI hardware directly destroys habitat.
· Land‑use change: Data centre campuses consume large areas, often in arid or semi‑arid ecosystems.
· Novel entities: Chemical pollution from semiconductor fabrication (PFAS, glycols, heavy metals) is unquantified at industry scale.
· Biogeochemical flows: Nitrogen and phosphorus discharges from manufacturing zones are rarely tallied.
The assessments that are commonly discussed in public are firmly stuck in carbon tunnel vision.
Waste and planetary sinks
AI produces a growing mountain of toxic electronic waste with abysmal recycling rates (well under 25 % for rare metals). Planned obsolescence in the hyperscale world is rapid, creating long‑term liabilities: leaching of heavy metals into groundwater, open burning of e‑waste in informal economies, and loss of critical materials. The planetary sinks for electronic waste are already overloaded, and the chemical absorption capacity of ecosystems near manufacturing clusters is similarly exceeded.
Viability
Subsidies, externalised costs, loan guarantees
Current AI expansion would collapse overnight without massive externalised costs. The physical infrastructure piggybacks on publicly funded energy grids, water systems, and transportation networks that are not priced to reflect their true environmental or social cost. Cloud providers secure subsidised electricity rates, tax breaks for data centre construction, and implicit state guarantees for the stable legal and geopolitical environment they require. Venture capital and equity markets fund AI on the expectation of future monopoly rents, not current profitability; the industry as a whole is burning hundreds of billions of dollars a year with no near‑term path to break‑even for many foundational services.
New complex infrastructure
AI demands an entirely new, resource‑intensive global infrastructure layer: hyperscale data centres (each consuming as much electricity as a small city), dedicated high‑voltage transmission lines, upgraded backbones, and an enormous increase in electrical generation capacity. This is not a lightweight overlay — it is a material‑heavy, fast‑growing parallel system.
Dependence on “breakthrough” technologies
The promise that AI’s environmental footprint will be solved rests on things that do not yet exist at scale: commercially viable fusion power, room‑temperature superconductors, ultra‑low‑power neuromorphic chips, or 100% circular semiconductor recycling. Today’s efficiency improvements are incremental and continuously overwhelmed by demand growth. AI’s viability as a “solution” is thus contingent on breakthroughs that remain speculative.
Social Aspects
Infinite Growth Enabler
AI is the ultimate enabler of the infinite economic growth paradigm. It is sold as a tool that will lift productivity growth rates indefinitely, creating new markets, new product categories, and demand for evermore computation. It does not challenge the growth model; it accelerates it by orders of magnitude while masking the physical limits.
Promoters and Profiteers
The primary promoters are a handful of massive technology corporations (Microsoft, Google, Amazon, Meta, Nvidia) and their billionaire investors, along with governments who see AI as a source of geopolitical advantage. Profits concentrate at the apex of the hardware and cloud stack. Even as “open‑source” models appear, the capital requirements for training frontier models ensure that power remains in the hands of those who control the hardware.
Concentration of Wealth and Power
AI is structurally concentrating wealth and power. It automates cognitive and creative labour in ways that tend to commodify skills, suppress wages, and concentrate intellectual property rights. The immense upfront capital cost creates a moat that entrenches incumbents. There is a lively discourse around “AI for everyone”, but the underlying ownership of the means of compute remains hyper‑concentrated.
Eminent Domain and Dispossession
Increasingly, the physical footprint of AI is being carved out through raw state‑corporate force. In multiple jurisdictions, eminent domain is being wielded to seize private homes, small farms, and community land to make way for hyperscale data centre campuses. Residents — disproportionately from lower‑income or rural communities — are removed with minimal compensation and zero meaningful consent, their objections overridden by the imperative of “digital infrastructure.” This dispossession shatters local social fabrics, extinguishes generational homes, and redirects public resources (grid connections, water, tax abatements) toward a handful of global technology firms. It is a stark illustration of how AI, far from distributing power, extends a form of extractive colonialism into the twenty‑first century, using the machinery of the state to subordinate human habitation to server racks.
Fragile Global Supply Chains
AI requires the most globalised, centralised, and fragile supply chain imaginable: Taiwan‑based advanced fabs, rare earths from specific mines, assembly in a handful of Asian sites, data centres owned by US tech giants, and undersea cables connecting a few major hubs. It is the opposite of relocalisation and community resilience. A disruption in any node (geopolitical, seismic, pandemic) cascades globally.
Shutting Down Alternatives
AI is routinely presented as the only realistic path to solving climate change, resource scarcity, and economic stagnation within the current system of continued growth — thereby side‑lining or actively delegitimising degrowth, sufficiency, relocalisation, and simplification. When critics point out the ecological impossibility of exponential AI expansion, the standard reply is to accelerate technological innovation, not to question the underlying commitment to growth. This narrows the Overton window, framing deeper systemic change as either unnecessary or impossible.
Conclusion
When all is said and done, the notion that AI is a “saviour” to help humanity achieve ecological sustainability while ensuring economic prosperity for all is as far removed from reality as one can imagine. The chasm between what cheerleaders of this technology repeatedly promise and what is actually happening is as vast as the growth they insist will follow. Many promoters genuinely believe in AI’s promise; what they fail to see is the structural logic of a system that must turn any efficiency gain into a new frontier for extraction — a blindness that does not make them conspirators, but makes their project no less dangerous.
It boggles my mind that the elite profitting from this latest and greatest undertaking (scam?) are finding as many followers as they do. But perhaps this support is not as widespread as some believe. It may be simply the result of marketing flooding the airwaves with stories of imminent prosperity to help keep the eyes of the masses averted from the pillaging of national treasuries (including natural treasuries) while disasters build. There does seem to be increasing pushback against the entire AI rollout and expansion so maybe the increase in stories of its benefits is the counterpoint being crafted by its boosters (grifters?) to keep the party going for as long as they can and hope another salvation techno-fix can come along to help keep their speculative ventures (Ponzi schemes?) afloat for another quarter or two…
Given the full weight of the analysis above, the potential for AI to achieve sustainable resource management, fix climate change, or deliver true human sustainability is, in a word, negligible — and in practical terms, actively negative.
This is not because AI lacks any useful narrow capability. It can, in controlled contexts, optimise a supply route, improve a weather model, or fold a protein. But the techno-fix questionnaire reveals why these isolated successes are irrelevant to the larger question. The AI system as a whole is a net addition to material and energy throughput, not a replacement for destructive technologies. Its life‑cycle demands — from mining rare earths to cooling hyperscale data centres — are staggering and are accelerating the breaching of multiple planetary boundaries well beyond carbon. Its economic viability rests entirely on externalised costs, subsidies, and speculative capital, not on genuine value creation that reduces ecological pressure.
Most decisively, AI does not challenge the infinite growth paradigm; it is its most powerful engine. Every so‑called efficiency gain is swallowed by the Jevons paradox, generating new demand, new markets, and new forms of extraction. Far from relocalising economies or distributing power, AI concentrates wealth, seizes land via eminent domain, and delegitimises the very sufficiency and degrowth pathways that might actually address overshoot.
In short, AI is being deployed as a substitute for the political and economic changes required for genuine sustainability — a dazzling distraction that keeps the extractive machinery running. The long‑promised holy grail remains a mirage; the chalice is an energy‑hungry server farm, and drinking from it only deepens the thirst.
Recent and Relevant Articles of Interest:
AI Translations Are Adding ‘Hallucinations’ to Wikipedia Articles
AI firms and their US military ties, “a whole civilization will die tonight” edition
AI, Money, Human Nature and the Problem with Problems: circularity leads to delusion
AI Hype Meets Hardware Crunch As US Power Equipment Market Eyes $65 Billion Boom | ZeroHedge
Dissent and AI: The Future Before Us Clear
Are Inequality, AI and Digital Life Undermining Society? Yes.
Inside the plot to cover Europe with gas-powered AI data centres — resilience
Homeowners Face Eminent Domain Bulldozers As Data Centers Demand Ever More Power | ZeroHedge
The AI Bubble: The Hidden Costs of the Data Center Boom — Global Research
AI’s Coming Reality Check: When the Physics Finally Hits the Hype
The AI Takeover Has Arrived — The Honest Sorcerer
How the AI Layoff Shock Is Triggering the Greatest Wealth Transfer in History
AI Data Centers Are Not Like Railroads
AI Data Centers: The Real Reason They’re Going Up Everywhere
What is going to be my standard WARNING/ADVICE going forward and that I have reiterated in various ways before this:
“Only time will tell how this all unfolds but there’s nothing wrong with preparing for the worst by ‘collapsing now to avoid the rush’ and pursuing self-sufficiency. By this I mean removing as many dependencies on the Matrix as is possible and making do, locally. And if one can do this without negative impacts upon our fragile ecosystems or do so while creating more resilient ecosystems, all the better.
Building community (maybe even just household) resilience to as high a level as possible seems prudent given the uncertainties of an unpredictable future. There’s no guarantee it will ensure ‘recovery’ after a significant societal stressor/shock but it should increase the probability of it and that, perhaps, is all we can ‘hope’ for from its pursuit.
If you have arrived here and get something out of my writing, please consider ordering the trilogy of my “fictional” novel series, Olduvai (PDF files; only $9.99 Canadian), via my website or the link below — the “profits” of which help me to keep my internet presence alive and first book available in print (and is available via various online retailers).
Attempting a new payment system as I am contemplating shutting down my site in the future (given the ever-increasing costs to keep it running).
If you are interested in purchasing any of the 3 books individually or the trilogy, please try the link below indicating which book(s) you are purchasing.
Costs (Canadian dollars):
Book 1: $2.99
Book 2: $3.89
Book 3: $3.89
Trilogy: $9.99
Feel free to throw in a “tip” on top of the base cost if you wish; perhaps by paying in U.S. dollars instead of Canadian. Every few cents/dollars helps…
https://paypal.me/olduvaitrilogy?country.x=CA&locale.x=en_US
If you do not hear from me within 48 hours or you are having trouble with the system, please email me: olduvaitrilogy@gmail.com.
You can also find a variety of resources, particularly my summary notes for a handful of texts, especially William Catton’s Overshoot and Joseph Tainter’s Collapse of Complex Societies: see here.