The AI Blame Game — And Why We're Getting It Wrong

Anxiety is the ambient noise of the AI moment. The technology feels sudden, total, and irreversible — an internet-scale disruption, compressed into a news cycle. And in that atmosphere of accelerating change, a familiar cognitive shortcut kicks in: when something big goes wrong, blame the biggest thing in the room.

Right now, that thing is AI.

But pattern recognition isn't analysis. And in two of the most consequential debates of this economic moment — rising electricity bills and white-collar job loss — the AI narrative is obscuring more than it's revealing.

Let's look at the actual numbers.

The Power Surge Problem: What's Really Driving Your Electricity Bill

The rate increases are real. The cause is more complicated.

In September 2025, the U.S. Energy Information Administration (EIA) reported that residential retail electricity prices reached approximately 18 cents per kWh — a 7.4% increase year-over-year. In a May 2025 report, the EIA projected that electricity price growth would outpace general inflation through 2026. The trajectory is clear. The explanation is where things get murky.

A growing chorus of energy analysts has pointed to AI data centers as the culprit. The logic is intuitive: data centers are energy-intensive by design, consuming massive amounts of power to train and run AI models, and even more to cool the hardware doing the work. A 2025 Department of Energy report — commissioned from Lawrence Berkeley National Laboratory — projected that data centers accounted for 4.4% of total U.S. electricity consumption in 2023, with that share potentially climbing to 12% by 2028.

From there, it's a short rhetorical leap to: data centers are inflating your electricity bill.

But the timeline doesn't hold up.

U.S. electricity costs began their steep ascent well before ChatGPT launched in November 2022 — the moment most people mark as the beginning of the generative AI era. The actual drivers? Post-pandemic inflation, a natural gas price shock triggered by the Russia-Ukraine war, the catastrophic Texas grid failure in winter 2021, and decades of deferred infrastructure investment finally coming due.

The distribution transformer shortage tells the story clearly: demand exceeded supply by 10%, power transformers by 30%, and lead times stretched from 50 weeks in 2021 to over 120 weeks today. Crucially, the same Lawrence Berkeley National Laboratory study that projected data center growth was explicit: data center load was not a primary driver of electricity price increases through 2024.

AI does consume significant energy. The International Energy Agency puts a standard Google search at 0.3 Wh; a ChatGPT query runs about 2.9 Wh — roughly 10x. But in context, data centers represent approximately 4% of total U.S. electricity consumption, nestled within the commercial sector's 36% share. Even at the projected 12% by 2028, that's still a third of residential consumption alone. Calling it the primary variable distorts the picture.

The deeper irony: the companies most exposed to energy risk are the ones most invested in solving it.

Grid instability isn't an abstraction for hyperscalers — it's an existential operational risk. Which is why Big Tech isn't just consuming power; it's vertically integrating into energy infrastructure. Microsoft signed a long-term agreement to restart the Three Mile Island nuclear reactor. Google entered a ~$5 billion acquisition agreement for solar and battery storage developer Intersect Power. Utility PG&E has estimated that adding large-scale commercial loads could actually reduce residential rates by up to 2%.

Google DeepMind demonstrated this logic as far back as 2016, using AI to cut data center cooling energy consumption by 40% — a 15% improvement in overall Power Usage Effectiveness (PUE).

PUE (Power Usage Effectiveness) measures data center energy efficiency: total facility energy divided by IT equipment energy. A score closer to 1.0 indicates minimal waste from cooling, lighting, and other overhead systems.

AI as energy consumer. AI as grid optimizer. The same technology plays both roles — and that complexity deserves more than a single-cause narrative.

The Automation Anxiety: Are AI Systems Actually Taking Jobs?

The headlines are alarming. The data is more nuanced.

The warnings have been coming from credible places. Andrew Yang has projected that hundreds of thousands of white-collar workers could be displaced within 12 to 18 months. AWS carried out significant workforce reductions in 2025. Amazon CEO Andy Jassy stated plainly that AI-driven efficiencies would reduce the company's overall headcount over the coming years.

Against that backdrop, a reasonable person might conclude: productivity is rising, and the people generating it are becoming optional.

So what do the actual numbers show?

Total U.S. layoff announcements in 2025 reached 1.17 million — the highest since the COVID-19 pandemic. But according to outplacement firm Challenger, Gray & Christmas, only 4.5% of those separations — roughly 55,000 positions — cited AI as a contributing factor. In New York State, where employers are now required to check a box attributing layoffs to "technological innovation or automation," exactly zero of 160 reporting companies selected it.

The narrative around mass tech layoffs from late 2022 onward tracks more cleanly to a different explanation: pandemic-era over-hiring, a rapid interest rate environment, and a structural reset toward profitability over growth. AI provided convenient cover in some cases. But the data doesn't support making it the primary cause.

The adoption gap is also wider than the discourse suggests.

McKinsey's 2025 global survey found that while 88% of companies had adopted AI in at least one functional area, only one-third had reached enterprise-wide scale, and just 7% had fully deployed AI across the organization. Only 6% of companies reported generating meaningful economic value from AI initiatives.

A Harvard Business Review survey of 1,006 executives globally found that just 2% of organizations had already executed large-scale workforce reductions attributable to AI adoption.

The capability-to-deployment gap is real and structural. MIT research indicates AI can theoretically handle tasks equivalent to 11.7% of the current U.S. workforce — but capability isn't displacement. In high-stakes verticals like healthcare and financial services, legal liability and regulatory exposure make human-in-the-loop oversight not just preferable, but effectively mandatory.

Human-in-the-Loop refers to operational models where humans review, validate, or override AI decisions before they take effect. In regulated industries, this isn't optional — it's the baseline for managing liability and maintaining auditability.

At the same time, AI is creating roles that didn't exist a few years ago.

The Information Technology & Innovation Foundation (ITIF) estimates that in 2024, AI-driven growth directly created approximately 119,000 jobs — compared to roughly 12,700 displaced. Autodesk's AI Jobs Report found consistent year-over-year growth in AI-related job postings across design, manufacturing, and engineering, with AI engineers and prompt engineers among the fastest-growing new roles.

Goldman Sachs Research has made the structural case: technological shifts historically stimulate demand for new labor categories, and employment disruptions from AI are likely to prove transitory in the aggregate.

The long-run data is worth sitting with: approximately 60% of U.S. workers today hold jobs that didn't exist in 1940, and more than 85% of employment growth since then has been driven by technology-led job creation.

The Takeaway: Calibrated Thinking in a Noisy Moment

We are unambiguously in a period of technological transition. That transition is real, it's accelerating, and it will reshape industries, labor markets, and infrastructure in ways that aren't yet fully legible.

But that's exactly why precision matters.

Long-accumulated structural problems — aging energy infrastructure, pandemic-era hiring distortions, macroeconomic dislocation — are being collapsed into a single variable: AI. That framing is intellectually convenient, and it's misleading. The forces reshaping the world aren't singular. Demographic shifts, energy geopolitics, and monetary cycles are all operating simultaneously.

The risk of over-attributing change to AI runs in both directions. It feeds unfounded fear. And it lets the actual, more complex causes off the hook.

In the middle of a transition this consequential, the most valuable posture isn't alarm and it isn't boosterism. It's the discipline to read what's actually happening — and to act on that, rather than on the noise.

About Kakao Ventures

Founded in 2012 and backed by Kakao — Korea's leading tech platform — Kakao Ventures is one of Korea's most active Seed-stage venture capital firms, with approximately $280M USD in AUM. We partner with founders before the path is fully defined, when conviction in people matters more than proof in numbers.

Our portfolio includes Lunit (AI cancer diagnostics), Rebellions (AI semiconductors), and Dunamu (operator of Upbit, one of Asia's largest crypto exchanges).

If you're building at the edge of what's possible — we'd like to hear from you