In a shocking twist that surprised absolutely no one with a fully functioning frontal lobe, former Google CEO Eric Schmidt took the TED 2025 stage last week to declare that artificial intelligence – the technology currently receiving more media coverage than oxygen and Donald Trump – is actually “underhyped.” Yes, you read that correctly. The man who helped steer one of the world’s largest tech companies believes we’re not talking enough about AI, in the same way that fish might not be talking enough about water.
Schmidt’s talk, delivered to an audience of head-nodding tech enthusiasts who would applaud a toaster if it had “AI-powered” in its name, argued that we are “drastically underestimating the scope and speed of the AI revolution”.1 This from the man whose company once thought Google+ would be a Facebook killer.
When Machines Play Board Games Better Than Your Retirement Portfolio
The crux of Schmidt’s argument rests partially on AlphaGo’s legendary 2016 victory over Go champion Lee Sedol, which Schmidt frames as a watershed moment for artificial intelligence.2 For the uninitiated (or those with actual hobbies like Chess), Go is an ancient Chinese board game with more possible configurations than there are atoms in the known universe. AlphaGo’s victory was indeed impressive – the AI made a move so unexpected that human experts initially thought it was a mistake.3
“What happened in this particular set of games was in roughly the second game, there was a new move invented by AI in a game that had been around for 2,500 years that no one had ever seen,” Schmidt gushed during his TED talk, conveniently glossing over the fact that Lee Sedol ultimately won one game against the machine.4
But here’s where our inner Sherlock Holmes starts twirling his metaphorical mustache. AlphaGo’s victory, while impressive, bears a suspicious resemblance to other carefully controlled AI demonstrations. Consider Meta’s recent Llama 4 controversy, where the company submitted a specially crafted, non-public variant called “Llama-4-Maverick-03-26-Experimental” to benchmark tests.5 When the actual public model was released, users reported “lackluster results” compared to the benchmark claims. One might reasonably ask: Was AlphaGo similarly “optimized” specifically for its match against Lee Sedol?
As one unnamed AI researcher who asked to remain anonymous because they “enjoy having a career” told us: “Winning at Go is impressive, but it’s also a closed system with perfect information. Real-world problems are messy. It’s like saying you’re ready for the Daytona 500 because you’re really good at Mario Kart.”
90 Gigawatts? Great Scott!
Perhaps the most glaring omission in Schmidt’s techno-utopian TED sermon was any meaningful discussion of the absolutely eye-watering energy requirements of his ‘underhyped’ AI revolution. According to recent projections, AI data centers could consume a staggering 90 gigawatts of power globally by 2028.6 For context, that’s roughly the equivalent of Denmark’s entire power consumption.7 Not a neighborhood in Denmark. Not a city in Denmark. The ENTIRE country of Denmark!
Schneider Electric’s latest report spells it out in terrifying clarity: the overall power consumption associated with AI workloads will reach approximately 4.3 gigawatts, “equivalent to the total power consumption of a country”. And that’s just for starters. The International Energy Agency projects that data centers will consume 945 terawatt-hours by 2030 – roughly equivalent to Japan’s entire annual electricity consumption.8
Meanwhile, India is desperately trying to meet its AI ambitions by building out 10 gigawatts of capacity,9 falling hilariously short of the 40-50 terawatt-hours of additional electricity the country will require for its projected AI data centers by 2030.10 When asked about this small discrepancy, India’s Ministry of Actually Getting Things Done reportedly replied, “We’re working on it, possibly by harnessing the hot air from tech conference keynotes.”
But wait, it gets better! The energy efficiency of these AI models is about as impressive as my attempts at sobriety during a TechCrunch conference after-party. According to Sasha Luccioni, a top AI researcher, generative AI models use up to 30 times more energy than traditional search engines11. That’s right – one simple high-definition image generation uses the same amount of energy as fully charging your phone!12
As one energy analyst who wished to remain anonymous because “I enjoy having electricity” told us: “By 2030, AI might consume up to 25% of US power requirements. We’re basically building a technology that will either solve climate change or cause rolling blackouts across America. It’s a race to see which happens first.”
The Strategic Under hype
When a former Google CEO gets on stage and claims something is “under hyped,” your BS detector should be screaming louder than a startup founder who just lost their Series A funding. There’s an art to the strategic under hype – it’s the corporate equivalent of saying “I’m actually really humble” at a job interview.
Schmidt’s declaration that AI is “underhyped” is the tech world equivalent of yelling “FIRE!” in a theater that’s already on fire, where everyone is already screaming about the fire, and firefighters are actively spraying water on the flames. It’s not just redundant; it’s suspiciously so.13
Consider the metrics: AI is receiving unprecedented investment, media coverage, and academic attention. Companies are tripping over themselves to slap “AI-powered” on literally anything with an on/off switch. Meta’s Mark Zuckerberg rebounded from his metaverse debacle by pivoting so hard to AI that he probably gave himself whiplash. Microsoft bet its entire future on OpenAI. Google launched Bard…then Gemini…then apologized for Gemini…then relaunched Gemini. Every startup pitch deck now contains the phrase “AI” approximately 84 times per slide.
Yet according to Schmidt, this isn’t enough hype. One wonders if he’s been measuring hype in some alternate dimension where people talk more about sustainable farming practices than they do about ChatGPT.
But here’s the brilliance of Schmidt’s move: claiming something is “underhyped” is perfect headline bait. It’s contrarian. It’s provocative (In Will Ferrell’s voice in Blades of Glory) . It guarantees coverage. If he had said “AI is exactly hyped the correct amount,” would we be writing this article? Would TED have uploaded the video? Would you be reading this right now? No, you would be doing something productive, and nobody wants that.
The Artificial Interviewer
Perhaps the most telling moment of Schmidt’s TED appearance wasn’t what he said, but rather the questions he was asked. The interviewer, Bilawal Sidhu, engaged with Schmidt in what appeared to be a series of pre-planned softballs that would make a White House press secretary blush.
Our investigative team (one intern with too much time and not enough supervision) conducted a linguistic analysis of Sidhu’s questions and found a 78% probability that they were generated by an AI, possibly the very technology they were discussing. The questions featured that distinctive blend of sounding intelligent while actually saying nothing -the verbal equivalent of a LinkedIn post about “synergy” and “disruption.”
One particularly revealing exchange:
Sidhu: “If you fast forward to today, it seems that all anyone can talk about is AI, especially here at TED. But you’ve taken a contrarian stance. You actually think AI is underhyped. Why is that?”
Notice the setup: acknowledge the hype, frame Schmidt’s view as “contrarian” (despite it being the dominant view among tech executives), then lob the softball. It’s the conversational equivalent of placing a basketball hoop three feet off the ground and asking Michael Jordan if he thinks he can dunk.
Schmidt’s response, naturally, was to talk about ChatGPT and ignore the more complex reality that AI adoption in actual businesses remains modest. According to real data, only about 20% of workers use generative AI on their jobs, meaning a whopping 80% still do not use these tools regularly. Moreover, only 5.4% of firms have officially deployed generative AI in a formal way.14 But why let facts get in the way of a good ol’ TED talk?
Powering Delusion: The Energy Elephant in the Room
The most glaring contradiction in Schmidt’s underhyped revolution is the simple fact that we don’t have enough electricity to power it. This isn’t a small problem; it’s the equivalent of Elon Musk announcing plans to move the entire human population to Mars without mentioning the minor detail that we don’t have spaceships that can get us there.
The projections are frankly terrifying. Crypto mining – the previous energy villain – pales in comparison. AI’s projected electricity use by 2026 (~1,000 TWh) would equal Germany’s total annual power consumption.15 It’s roughly 10 times the power demand of Google’s entire global infrastructure in 2021.
Goldman Sachs projects that 85-90 gigawatts of new nuclear capacity would be needed just to meet data center power demand growth.16 To put that in perspective, that’s approximately 85-90 new nuclear reactors. And we all know how quickly and uncontroversially those get built.
When confronted with these energy requirements, most AI evangelists mumble something about “efficiency improvements” before changing the subject faster than a politician caught in a scandal. But the math remains stubbornly consistent: more AI means more energy, and more energy means more problems.
As one power grid engineer told us off the record: “We’re building the world’s most advanced technology on the world’s most outdated energy infrastructure. It’s like putting a Ferrari engine in a horse carriage and wondering why it keeps catching fire.”
The Satire Writes Itself
In the end, perhaps the most ironic aspect of Schmidt’s under hype claim is that it came just weeks after a Mozilla Foundation – funded performance art project called “Artificial Life Coach” launched specifically to critique AI hype.17 The project’s creator warned: “Don’t believe all the marketing hype around AI. There are some serious downsides.”
Schmidt either missed this memo or, more likely, recognized that the greatest form of power in the tech industry is controlling the narrative. By claiming AI is “under hyped,” he’s not making a factual statement – he’s attempting to reset the conversation on his terms.
As we hurtle toward an AI-powered future that will require more electricity than many countries can produce, perhaps it’s time to ask the obvious question: Who benefits from this narrative? Certainly not the average consumer, who will face higher electricity bills. Certainly not developing nations, which will struggle to build the necessary infrastructure. Certainly not the climate, which will bear the burden of increased energy production.
The beneficiaries are clear: tech companies like Google, the chip manufacturers like NVIDIA (which Schmidt specifically mentioned as “the big winner right now”), and the venture capitalists funding the next generation of AI startups.
In a world where satire and reality have become increasingly difficult to distinguish, Schmidt’s claim that AI is “underhyped” may be the most unintentionally hilarious statement of 2025. It would be funnier if it weren’t going to potentially leave us all sitting in the dark.
So what do you think, dear TechOnion readers? Is AI truly underhyped as Schmidt suggests, or are we witnessing the greatest case of technological wishful thinking since the Juicero? Drop your hottest takes in the comments below. Extra points if you can craft a response that uses less electricity than training a small language model.
Support Independent Tech Journalism by Donating to TechOnion
If this article saved you from wasting venture capital on an AI startup that promises to revolutionize the way people tie their shoelaces, consider supporting our electricity bill. We’re currently mining cryptocurrency to power our office, but the neighbors keep complaining about the blackouts. Donate any amount you like-we accept all currencies, including those that actually exist. Your support helps us continue peeling back the layers of tech nonsense until everyone’s eyes water.
References
- https://www.linkedin.com/pulse/why-eric-schmidt-says-ai-still-underhyped-matters-now-derek-madden-alsyc ↩︎
- https://deepmind.google/research/breakthroughs/alphago/ ↩︎
- https://www.wired.com/2016/05/google-alpha-go-ai/ ↩︎
- https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol ↩︎
- https://www.theregister.com/2025/04/08/meta_llama4_cheating/ ↩︎
- https://www2.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2025/genai-power-consumption-creates-need-for-more-sustainable-data-centers.html ↩︎
- https://www.powerelectronicsnews.com/schneider-electric-predicts-substantial-energy-consumption-for-ai-workloads-globally/ ↩︎
- https://www.nature.com/articles/d41586-025-01113-z ↩︎
- https://asian-power.com/ipp/exclusive/avaada-boost-re-load-meet-demand-indias-data-centres ↩︎
- https://www.eqmagpro.com/can-india-meet-the-power-demand-for-ai-data-centres-by-2030-eq/ ↩︎
- https://www.thedailystar.net/tech-startup/news/generative-ai-uses-30-times-more-energy-search-engines-research-3706636 ↩︎
- https://artofprocurement.com/blog/supply-the-surging-problem-of-ai-energy-consumption ↩︎
- https://theaiinsider.tech/2024/05/08/this-stuff-is-underhyped-former-google-ceo-eric-schmidt-on-ais-transformative-potential/ ↩︎
- https://www.akooda.co/blog/state-of-generative-ai-adoption ↩︎
- https://evolutionoftheprogress.com/ai-power-consumption-exploding/ ↩︎
- https://www.goldmansachs.com/insights/articles/is-nuclear-energy-the-answer-to-ai-data-centers-power-consumption ↩︎
- https://foundation.mozilla.org/en/blog/an-antidote-for-ai-hype-through-satirical-performance-art-activist-exposes-the-limitations-of-ai-tools-and-examines-their-ties-to-systemic-inequality-and-injustice/ ↩︎
