Every ChatGPT Question Comes at a Cost: Enough Power for a Lightbulb and a Teaspoon of Water
In an increasingly digital world, the convenience of artificial intelligence like ChatGPT has become commonplace.
Yet, behind each seemingly effortless query lies a hidden environmental toll.
From the energy powering vast data centres to the surprising amount of water used for cooling, the true cost of our AI interactions extends far beyond a simple internet connection.
OpenAI CEO Sam Altman has now shed more light on the environmental footprint of AI tools like ChatGPT.
In a recent blog post, he revealed that an average ChatGPT query uses just 0.000085 gallons of water, which he described as 'roughly one-fifteenth of a teaspoon.' The Environmental Cost of ChatGPT
This insight comes amidst growing concerns over how much energy and water artificial intelligence systems consume. Altman's recent blog post, while discussing the future of AI, also presented specific figures regarding ChatGPT's resource consumption.
'People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second, or a high-efficiency lightbulb would use in a couple of minutes,' Altman wrote.
This was less than almost every estimate I have seem: according to the latest Sam Altman post, the average ChatGPT query uses about the same amount of power as the average Google search in 2009 (the last time they released a per-search number)… 0.0003 kWh pic.twitter.com/AgVQB7zkOu — Ethan Mollick (@emollick) June 10, 2025
'It also uses about 0.000085 gallons of water, roughly one-fifteenth of a teaspoon,' he added. The water usage of one-fifteenth of a teaspoon per query might appear tiny for one person. Yet, given the billions of queries AI systems handle daily, the total effect becomes substantial. OpenAI did not explain how these figures were arrived at.
'As data centre production gets automated, the cost of intelligence should eventually converge to near the cost of electricity.' Altman continued. However, this perspective on electricity cost also underscores the larger environmental concerns now under close examination. Increased Focus on AI's Environmental Burden
As AI becomes more widely adopted, experts and researchers are concerned about its environmental burden. Studies earlier this year forecast that AI could potentially use more electricity than Bitcoin mining by late 2025.
Water consumption also presents a major worry, particularly for water-based cooling data centres. According to a 2024 report from The Washington Post, creating a 100-word email with GPT-4 might use roughly one bottle of water, influenced by the data centre's site and cooling approach. This demonstrates how environmental effects can differ depending on the infrastructure.
Altman suggests that, in time, the expense of producing intelligence via AI will nearly match that of electricity. Until then, discussions concerning AI's environmental impact will become more vocal. Every Drop Counts: AI's Thirst for Water
This isn't the only occasion Altman has foreseen AI becoming more affordable to run. In February, Altman stated in a blog post that AI's operating expenses would reduce tenfold yearly.
'You can see this in the token cost from GPT-4 in early 2023 to GPT-4o in mid-2024, where the price per token dropped about 150x in that time period,' Altman wrote. 'Moore's law changed the world at 2x every 18 months; this is unbelievably stronger,' he added. Big Tech's Ambitious Energy Drive
Leading tech firms in the AI race are exploring nuclear energy as a power source for their data centres. In September, Microsoft finalised a 20-year agreement with Constellation Energy to restart a dormant nuclear plant at Three Mile Island.
In October, Google revealed it had partnered with Kairos Power, a nuclear energy company, to produce three small modular nuclear reactors. These reactors, which can supply up to 500 megawatts of electricity, are expected to be operational by 2035.
Google, Amazon, and Microsoft have signed deals for nuclear energy projects to power their AI and data capabilities. Google's agreement with Kairos to buy power from multiple small modular reactors is a world first with 500MW planned across 6-7 reactors. pic.twitter.com/xNxQpKKkdr — Works in Progress (@WorksInProgMag) November 13, 2024
In an October interview, Google's CEO, Sundar Pichai, told Nikkei Asia that the search giant aims for net-zero emissions across its entire operations by 2030. He further stated that Google also assessed solar energy beyond just nuclear power.
'It was a very ambitious target, and we are still going to be working very ambitiously towards it. Obviously, the trajectory of AI investments has added to the scale of the task needed,' Pichai said.
Originally published on IBTimes UK
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Int'l Business Times
10 hours ago
- Int'l Business Times
AI Isn't Fully Automated — It Runs on Hidden Human Labor
Welcome to Tech Times' AI EXPLAINED, where we look at the tech of today and tomorrow. Brought to you by Imagine this scenario, one that's increasingly common: You have a voice AI listen to your meeting at work, you get a summary and analysis of that meeting, and you assume AI did all the work. In reality, though, none of these tools work alone. PLAUD AI, Rabbit, ChatGPT, and more all rely on a layer of human labor that most of us don't hear about. Behind that clean chat interface on your phone or computer, there are data labelers that tag speech samples, contractors that rate AI answers , and testers feeding the system more examples to learn from. Some are highly trained while others focus on more of the tedious aspects of the work. No matter what, though, your AI isn't just automated - it's a complex blend of code and human effort. Without it, your AI wouldn't work at all. The Invisible Workforce Behind Everyday AI AI tools don't just appear out of thin air, of course. They learn similarly to the way we do: by example. That learning process often relies on what's called human-in-the-loop (HITL) training. As data-annotation company Encord says in a blog post: "In machine learning and computer vision training, Human-in-the-Loop (HITL) is a concept whereby humans play an interactive and iterative role in a model's development. To create and deploy most machine learning models, humans are needed to curate and annotate the data before it is fed back to the AI. The interaction is key for the model to learn and function successfully," the company wrote. Annotators, data scientists, and data operations teams play a significant role in collecting, supplying, and annotating the necessary data, the post continued. The amount of human input varies with how involved the data is and how much human interaction it will be expected to offer. Of course, as with many business activities, there are ethical concerns. Many content moderators complain of low pay and traumatic content to review. There can also be a language bias in AI training , something researchers and companies are likely working on to solve as AI becomes more complex and global. Case Study: PLAUD AI Various ways users wear the PLAUD Note device—on a wristband, clipped to a lapel, or hanging as a pendant—highlighting its flexibility for hands-free voice capture throughout the day. PLAUD AI PLAUD AI's voice assistant offers an easy, one-button experience. Just press a button, speak, and then let it handle the rest. As the company said on its website , the voice assistant lets you "turn voices and conversations into actionable insights." Behind the scenes, this "magic" started with pre-trained automatic speech recognition (ASR) models like Whisper or other custom variants , that have been refined with actual user recordings. The models not only have to transcribe words, but also try to understand the structure, detect speakers , and interpret tone of voice. The training involves hours and hours of labeled audio and feedback from real conversations. It's likely that every time you see an improvement in the output, it's thanks to thousands of micro-adjustments based on user corrections or behind-the-scenes testing. According to reviewers, PLAUD AI leverages OpenAI's Whisper speech-to-text model running on its own servers. There are likely many people managing the PLAUD AI version of the model for its products, too. Every neat paragraph that comes out of the voice assistant likely reflects countless iterations of fine-tuning and A/B testing by prompt engineers and quality reviewers. That's how you get your results without having to deal with all that back-end work yourself. Case Study 2: ChatGPT and The ChatGPT logo represents one of the most widely used AI assistants—powered not just by models, but by human trainers, raters, and user feedback. ilgmyzin/Unsplash When you use ChatGPT, it can feel like an all-knowing helpful assistant with a polished tone and helpful answers. Those are based, of course, on a foundation of human work. OpenAI used reinforcement learning from human feedback , or RLHF, to train its models. That means actual humans rating responses so the system could learn what responses were the most helpful or accurate, not to mention the most polite. "On prompts submitted by our customers to the API, our labelers provide demonstrations of the desired model behavior and rank several outputs from our models," wrote the company in a blog post . "We then use(d) this data to fine-tune GPT‑3." a popular online voice transcription service, also relies on human work to improve its output. It doesn't use RLHF like OpenAI does, but it does include feedback tools for users to note inaccurate transcriptions, which the company then uses to fine-tune its own models. The company also uses synthetic data (generated pairs of audio and text) to help train its models, but without user corrections, these synthetic transcripts can struggle with accents, cross talk, or industry jargon; things only humans can fix. Case Study 3: Rabbit R1's Big Promise Still Needs Human Help The Rabbit R1 made a splash with its debut: a palm-sized orange gadget promising to run your apps for you, no screen-tapping required. Just talk to it, and it's supposed to handle things like ordering takeout or cueing up a playlist. At least, that's the idea. Rabbit says it built the device around something called a Large Action Model (LAM), which is supposed to "learn" how apps work by watching people use them. What that means in practice is that humans record themselves doing things like opening apps, clicking through menus, or completing tasks and those recordings become training data. The R1 didn't figure all this out on its own; it was shown how to do it, over and over. Since launch, people testing the R1 have noticed that it doesn't always feel as fluid or "intelligent" as expected. Some features seem more like pre-programmed flows than adaptive tools. In short, it's not magic—it's a system that still leans on human-made examples, feedback, and fixes to keep improving. That's the pattern with almost every AI assistant right now: what feels effortless in the moment is usually the result of hours of grunt work—labeling, testing, and tuning—done by people you'll never see. AI Still Relies On Human Labor For all the talk of artificial intelligence replacing human jobs, the truth is that AI still leans hard on human labor to work at all. From data labelers and prompt raters to everyday users correcting transcripts, real people are constantly training, guiding, and cleaning up after the machines. The smartest AI you use today is only as good as the humans behind it. For now, that's the part no algorithm can automate away. Originally published on Tech Times
DW
13 hours ago
- DW
German drone industry takes off – DW – 05/30/2025
Carmakers shift to drone production Germany, long known for its automotive prowess, is witnessing a quiet but significant transformation. As traditional car production slows, a new industry is taking flight—literally. Across southern Germany, start-ups and engineers are pivoting from cars to combat drones. One such innovation is the 'Falke,' a drone designed for both civilian and military use, capable of high speeds and long-range missions. Its creators emphasize affordability and mass production, leveraging materials and techniques from the automotive sector. This shift is not isolated. Near Munich, drone developers are collaborating with car part suppliers, repurposing their expertise and infrastructure to meet the growing demand for military-grade drones. Ukraine war has changed German mindset The war in Ukraine has accelerated this trend. Companies like High Cat are already supplying drones to the front lines, designed to resist jamming and deliver real-time reconnaissance. These aren't just tech experiments—they're the result of serious engineering, often funded by private investors who once shied away from defense ventures. Even packaging companies are adapting, creating climate-controlled transport cases for drones. The transformation is industrial and cultural: Engineers and entrepreneurs are learning to think like soldiers, designing for battlefield conditions. Automotive suppliers, facing declining orders, are finding new life in defense manufacturing, applying their precision and scale to drone production. Tapping into the €500 billion fund Germany's defense sector is poised for further growth, thanks to a €500 billion special fund aimed at strengthening national security. This fund opens doors for manufacturers across the country, especially those with existing capabilities in high-precision production. Companies that once hesitated to enter the arms industry are now reconsidering, driven by both financial incentives and a growing sense of responsibility to defend democratic values. With the right support and strategic partnerships, Germany's factories could become key players in Europe's defense landscape—transforming economic uncertainty into industrial opportunity. This video summary was created by AI from the original DW script. It was edited by a journalist before publication.
Int'l Business Times
20 hours ago
- Int'l Business Times
Anna Forsythe Built Oncoscope to Give Doctors What They Actually Need: Usable Intelligence
In today's healthcare landscape, oncologists are drowning in data. Thousands of studies are published each month, new FDA approvals roll out regularly, and guidelines change constantly. Yet, despite this flood of information, many doctors still struggle to make the best treatment decisions for their patients. The problem is not a lack of data. The problem is a lack of usable insights. Nowhere is this more critical than in oncology, where every day and every decision can change the course of a patient's life. Anna Forsythe, founder of Oncoscope-AI, an AI-powered oncology intelligence platform, has spent her career at the intersection of science, economics, and clinical practice. She believes the answer is not to add more data to the pile but to transform it into clarity. "We do not need to give physicians more to read," Forsythe says. "We need to give them the right information, in real-time, that is relevant to the patient in front of them." It is a reasonable expectation, but one that the current system sometimes fails to meet. Oncology guidelines can span hundreds of pages, often outdated by the time they reach clinical use. "While there have been recent attempts to address this problem with a chatbot approach, it is the human/AI combination that is key in achieving usability and physicians' trust." Forsythe's platform, Oncoscope, solves this challenge by merging human expertise with trained artificial intelligence. It automatically reviews and organizes clinical trials, cross-referencing them with regulatory approvals and treatment guidelines. The result is a curated, reliable, and immediately usable knowledge base that oncologists can access in seconds. It works like this: a doctor inputs three basic clinical parameters—the stage of the disease, the genetic profile, and any prior treatments. In return, they receive a tailored, human-reviewed list of relevant studies, including survival outcomes and progression data, with immediate links to guidelines, approvals, and original publications. There is no need to scroll through irrelevant abstracts or search multiple databases. Everything is in one place, organized and actionable. This kind of tool is not just a convenience. It is a necessity. According to Forsythe, she was inspired to create Oncoscope after seeing people in her own life receive outdated or suboptimal cancer treatments. In one case, a friend with late-stage breast cancer was placed on chemotherapy despite the existence of a newer, more targeted therapy. The doctor had not yet seen the recent study supporting it. Forsythe found it in three clicks. "That story is not an exception," she says. "It is happening every day. And it is not because doctors are careless. It is because the information is not delivered in a way they can use quickly." This is a systemic flaw, and it has consequences. When oncologists default to older treatments because they cannot keep up with new evidence, patients can miss out on therapies that could extend or improve their lives. In a field where even a few months of added survival can mean everything, delays in information are delays in care. Forsythe and her team designed Oncoscope to cut out these delays. The system prioritizes the most rigorous research, flagged for relevance and clinical significance. Each entry is reviewed by experts to ensure it meets regulatory-grade standards. Doctors are not asked to interpret raw data—they are given the insights they need to make decisions now. Critically, Oncoscope is free for verified healthcare professionals. Forsythe calls it an altruistic venture, at least for now. Her goal is simple—to give doctors a tool they can trust and patients the care they deserve. "If we want better outcomes in cancer care," she says, "we do not need more information. We need smarter information. That is what changes lives."
