We Are in 1905: Electricity, Not the Dot-Com Bubble, Is the Right Analogy for AI

Introduction: A Critical Debate Over Historical Analogies

Every time a new wave of capital frenzy hits the AI sector, a familiar chorus of warnings echoes through the market: "This is the dot-com bubble of 2000 all over again." Skeptics point to NVIDIA's soaring stock price, the endless parade of AI startups, and the still-unfulfilled commercial returns, attempting to draw a direct parallel between the present moment and the catastrophic bubble burst of over two decades ago.

However, a growing number of technology historians and industry observers are putting forward a fundamentally different perspective: AI is not the internet bubble — it is more like the electricity revolution. And where we stand right now is roughly equivalent to 1905.

This shift in analogy is far from a mere rhetorical exercise. It fundamentally changes how we should anticipate the pace of AI development, the logic of investment, and the scale of societal impact.

Why the Dot-Com Bubble Analogy Falls Short

Comparing AI to the dot-com bubble has a certain surface-level plausibility: both involve a frenzy of capital inflows, rapidly inflating valuations, and a yawning gap between public expectations and actual deployment. But upon deeper analysis, this analogy suffers from fundamental flaws.

First, the dot-com bubble was essentially a bubble of business models, not of the technology itself. After the 2000 crash, internet technology didn't disappear — it actually gave rise, in the post-bubble cleansing, to giants like Google, Amazon, and Facebook. What the burst eliminated were companies without sustainable business models, not the underlying technological paradigm.

Second, the internet is fundamentally a "connection technology" — it transformed how information is transmitted but did not fundamentally reshape every physical production process. E-commerce replaced some retail, and social media changed communication, but the manufacturing floor, the agricultural field, and the surgical operating table were not fundamentally altered by the internet.

AI is entirely different. It is not merely a connectivity tool but a general-purpose capability layer — one that can permeate virtually every link in nearly every industry, from drug molecule design to metal defect detection, from legal document review to crop disease identification. This omnipresent pervasive power is precisely the hallmark of electricity, not the internet.

Why Electricity Is the Right Historical Mirror

The diffusion history of electricity as a General Purpose Technology (GPT) provides an extraordinarily precise framework for understanding AI.

A Prolonged Penetration Period

Thomas Edison completed the first commercial power station on Pearl Street in New York in 1882. But it wasn't until the 1920s that electricity truly became the dominant power source for American industry. That gap spans nearly 40 years.

Why? Because merely "inventing electricity" wasn't enough. Factories had to completely redesign their floor layouts — transitioning from centralized power systems arranged around a single steam engine shaft to distributed architectures where each machine had its own electric motor. This meant that building structures, workflows, management practices, and worker skills all had to be thoroughly reinvented. Economist Paul David pointed out in his landmark paper that the productivity gains from electricity didn't show up significantly in statistical data until the 1920s — nearly half a century after the technology was invented.

AI is undergoing a similar process. Large language models have demonstrated astonishing capabilities, but for enterprises to derive systematic benefits from them, they need to rebuild data infrastructure, redesign business processes, train employees in new modes of human-machine collaboration, and establish new governance and compliance frameworks. The construction cycle for these "complementary innovations" is far longer than the time it takes to train a model.

Infrastructure Comes First

In the early stages of the electricity revolution, the largest investment targets were infrastructure — power plants, transmission networks, transformers, and meters. This infrastructure construction itself generated enormous economic value, even before end-user applications had fully matured.

Today's AI landscape exhibits a strikingly similar pattern. NVIDIA's GPUs, massive data centers, cloud computing platforms, and model training frameworks — these are the "power plants and electrical grids" of the AI era. The current concentration of capital flowing into the infrastructure layer is not a signal of a bubble but rather the classic early-stage pattern of general-purpose technology diffusion.

The Unpredictability of Applications

In 1905, no one could have foreseen that electricity would ultimately give rise to the radio, television, refrigerator, air conditioning, the elevator (and the skyscrapers it enabled), and assembly-line production (and the mass consumer society it spawned). The most profound impacts of electricity came from applications that were utterly unimaginable at the time of its invention.

Similarly, our imagination regarding AI is likely severely underestimating its potential. We are still talking about "AI writing articles," "AI generating images," and "AI coding" — these are merely the "light bulbs" of the AI era: the most intuitive, most easily understood, but by no means the most important applications. AI's truly transformative applications are likely entirely new things we haven't yet conceived.

What 1905 Means

If we accept the electricity analogy, the assertion that "we are in 1905" carries several layers of important meaning:

First, the technology has been proven to work, but is far from mature. By 1905, electricity was no longer a laboratory toy — factories and cities had begun using it. But the degree of electrification was still very low, and the vast majority of factories still ran on steam power. Similarly, today's AI has demonstrated real-world value, but its penetration rate remains in the single-digit percentages across most industries.

Second, the biggest winners have yet to emerge. In 1905, General Electric and Westinghouse were already important companies, but the greatest enterprises and industrial forms of the electrical age — from home appliance empires to consumer electronics giants — wouldn't take shape for decades. Today's NVIDIA and OpenAI are certainly important, but the truly "defining companies" of the AI era may not yet have been founded — their founders might still be in college.

Third, the short term may disappoint, but the long term will exceed all expectations. The history of the electricity revolution repeatedly demonstrates that the return curve of general-purpose technologies is highly nonlinear: early progress is frustratingly slow, but once complementary innovations fall into place, growth accelerates explosively. AI's "productivity paradox" — massive investment with no statistically significant output gains yet — is a mirror image of the skepticism electricity faced in the early 20th century.

Fourth, institutional transformation is no less important than the technology itself. The full-scale adoption of electricity required not only technological advances but also regulatory frameworks (electrical safety standards), educational systems (training electrical engineers), financial innovation (financing large-scale infrastructure), and shifts in social attitudes. The same is true for AI — data governance, algorithmic regulation, AI literacy education, and the restructuring of intellectual property regimes. The construction of this "soft infrastructure" will determine whether the dividends of AI can be broadly shared.

Limitations and Caveats of This Analogy

Of course, no historical analogy is perfect. There are important differences between AI and electricity:

• Speed differential: AI may diffuse far faster than electricity did. The marginal cost of replicating digital technology approaches zero, whereas electrical infrastructure required massive physical construction. ChatGPT acquired 100 million users within two months — a diffusion speed unimaginable in the electrical age.
• Centralization risk: Electricity eventually became a highly decentralized resource, accessible to virtually everyone. But the core capabilities of AI are currently concentrated in the hands of a few companies possessing enormous computing power and data. Whether this centralization trend persists will profoundly affect social equity in the AI era.
• The uniqueness of cognitive capability: Electricity replaced physical power; AI replaces or augments cognitive capability. The latter touches the core of human identity and may provoke even more profound ethical and social challenges.

Outlook: An Era Demanding Patience and Vision

If we truly are in the "1905" of the AI era, then perhaps the most important takeaway is this: What we need is not fear of a bubble, but patience with the pace of transformation and vision for the long-term possibilities.

The dot-com bubble narrative implies that "all of this will come crashing down soon." The electricity revolution narrative implies that "all of this has only just begun, and it will prove far more profound than anyone imagines."

This does not mean there are no overheated elements in the current market, nor does it mean that every AI company will survive. The electrical age also witnessed the rise and fall of countless companies. But the critical distinction is this: After the bubble bursts, the underlying trend will not reverse. Just as after 1905, no force could halt the march of electrification transforming the world.

For business leaders, the right strategy is not to sit on the sidelines and wait to see whether the AI bubble will pop, but rather — like the visionary factory owners of the early 20th century — to begin seriously planning their own "electrification overhaul," even if it means seeing no immediate returns in the short term.

For investors, the focus should shift from "Who will be the next Pets.com?" to "Who is building the electrical grid of the AI era, and who is redesigning the factories of the AI age?"