Doom is turning into a crowded profession. Over the weekend, another wave of financial commentary washed ashore. Again, the tone was apocalyptic. Again, the references were antique: the South Sea Bubble, the Railway Mania, the Dot-Com Bust. Again, the conclusion was inevitable collapse.

In fact, it is now a cacophony of collapse. There is a palpable one-upmanship in pessimism. Merely predicting a standard market correction is passé. Absolute doomerism is the new currency. Analysts compete to name the largest number. $35 trillion in wealth, gone. A 1929-style crash, around the corner. One has to be squeamish or young to only invoke a TMT-level crash. The gold standard of displaying one’s hold on history is through the invocation of the South Sea Bubble. There has to be a mandatory reference to the Tulip Mania, but ideally with new charts.

Every doomer speaks as if they are a lone voice, a prophet in the wilderness. Yet, the profession is more crowded today than all the bears preceding every major bubble combined. It appears as if there is not a financial journal or independent blog that has not joined this dirge. More has been written on manias, panics, and crashes in the last few weeks than perhaps in all the centuries leading up to 2000.

Effectively, history is turned into a haunted house, filled only with ghosts of failures past. This bias has a tailwind. Doomerism sells. We know this. An article we wrote on the Seven Sins of new technology was one of our most-read pieces. A follow-up, Seven Redeeming Virtues, was one of our most ignored. 

If one has to use history to predict the future, the answer is not as assured as we are made to believe. This time, we will weaponize history to show how it is a tool to bolster one’s pre-determined conclusions or pre-formed views. We will show that a selective reading of history is a game two can play. The same patchy, popular history can be used to draw a map to a completely different destination.

Our core view remains unchanged. History is a poor guide. These times are unique. The technological shift is of a different magnitude altogether. This lazy analysis, this patchy history worn as a badge of honor, is a dangerous way to navigate the present. With a selective history as below, we can learn many different lessons, with the most important being to beware of folks waving history books and predicting apocalypses or dooms or anything in between.  Disasters are definitely possible, but not due to events that may have occurred centuries earlier. In summary, do remember that the following is a tongue-in-cheek write-up to show how history can be read in multiple ways and does not have the lessons of any assurity.

The '640K' Error

There is a paucity of imagination at the heart of most pessimism.

A famous, apocryphal story claims Bill Gates once said, "640K ought to be enough for anybody." He denies it. Likely, not just him but no one involved with technology at the time ever said it. But the spirit of the quote, the tendency to misjudge a new technology's scale, is the most consistent error in history.

Ken Olsen, the founder of Digital Equipment Corporation (DEC), stated in 1977, "There is no reason for any individual to have a computer in his home." He was an expert. Western Union dismissed the telephone as an "electrical toy" with no commercial possibilities. They were the experts. Cars were "fool's contraptions" according to some experts in the early 20th century, which would never replace the horse.

History is replete with these ghosts. They are the men who doomed their organizations by misreading a transformative shift. They analyzed a new technology through the lens of the old. IBM walked out of the PC and photocopying businesses. DEC was reduced to irrelevance as mainframes lost to small machines, and we all know about the BlackBerry. 

The same error is likely made by those using history. Critics trivialize Generative AI as a "statistical parrot." They use flimsy examples of personal use cases to prove the technology is useless. They focus on user jadedness or disillusionment as if they are watching a social media app.

But the numbers tell a different story. In early 2022, global token processing was negligible. By late 2024, Google alone processed 1.4 quadrillion tokens in a single month. Projections suggest this could grow 100-fold within a decade. This is not linear adoption. This is not hype. This is infrastructure-scale demand materializing in real time. Tulip bulbs did not go from zero to quadrillions in two years. Railway miles did not multiply a thousandfold in 24 months. The steam engine did not rewrite global logistics in a single business cycle.

The error lies in conflating current applications with ultimate utility. Early telephones couldn’t cross oceans. Early cars couldn’t survive winter. Early PCs couldn’t play video. Yet each unlocked cascading second- and third-order effects that their inventors never foresaw. Generative AI is not just a tool. It is a new layer of intellectual infrastructure. It will not merely automate tasks; it will redefine what tasks exist. Drug discovery, legal reasoning, scientific hypothesis generation, code synthesis: these are not edge cases. They are the foundation of a new productivity frontier.

To dismiss this because today’s models occasionally fabricate a date is like rejecting the internal combustion engine because the Model T stalled in the rain. The cost of this failure of imagination is not a missed trade or a wrong forecast. It is strategic obsolescence. It is becoming the next DEC, the next Western Union, the ghost in someone else’s cautionary tale.

The Myth of Shared Capacity: Your Car is Not a Bus

A new dogma is taking hold. It sounds rational. It is dangerously flawed. The logic is simple. Analysts calculate the total compute the future might need. They divide this by the power of a single GPU. They apply an optimum data center utilization rate of, say, 60%. The math screams one thing: overcapacity. A massive, wasteful build-out is underway.

This is spreadsheet logic. It is not human logic. It is certainly not the logic of power. It is a classic case of survivorship bias. During WWII, the statistician Abraham Wald was asked to analyze returning bombers. The military wanted to add armor to the areas most riddled with bullet holes. Wald advised the opposite. He said to armor the places that had no holes. The holes in the surviving planes showed where a plane could be hit and still fly. The missing data, from the planes that did not return, showed where the fatal hits landed. Analyzing current data center utilization is the same mistake. It only measures the survivors. It ignores the massive, unseen demand from applications and models that cannot exist yet due to a lack of compute.

Looking at it differently, if the demand were based on perpetual functioning of devices, the global car, PC, or even smartphone industries would be a fraction of their size. Calculate the world's total transport needs. Divide by the capacity of a single car. Assume optimum shared utilization. The result would demand a fleet a fraction of the one billion cars on the road today.

But people do not share their cars. The collective utilization of the global automobile fleet is in the single-digit percentages. The utility is personal, immediate, and sovereign, and this will be as true about the new compute as it has been for cars, home space, laptops, or smartphones. The recent AWS outage is a case study in this reality. When a critical cloud service flickers, the world stumbles. It is a stark reminder. For a global enterprise, reliance on another's compute is a systemic risk. Excess capacity is not waste. It is resilience. It is survival.

The AI era is not an incremental step. It is a change in the nature of computation itself. This is not the shift from a horse to a faster horse. This is the invention of the internal combustion engine. The old hardware is becoming obsolete. What is being built is not a shared utility. It is a strategic asset. Nations and corporations are not building for efficiency. They are building for sovereignty. 

The doomers see idle servers and cry "bubble." They are missing the point. They are counting the number of cars in 1910 and predicting the end of the road. They are using the logic of public transit to explain the rise of the private jet. The build-out is not about achieving perfect utilization. It is about owning the means of production for the next century. The goal is not to share. The goal is to win. 

Simply put, the capacity utilization of this new asset class will not be 60%. It may well be 5-10%. Notwithstanding the change in the forms of this new compute, it is likely that even at the utilization, in the way it is counted in spreadsheets, of 5-10%, there could be huge demand in the future. 

The 9.1 Quake: This Time is a Different Scale

Historic analogies are seductive. And, some can be catastrophically wrong. One such is the current popular sport of comparing the AI revolution to the railway boom of the mid-19th century. It is a willful ignorance of the difference in scales. This is not a difference of degree. It is a difference of kind. The railway boom unfolded over 50 years in leading countries and nearly a century if one considers the global impact. The internet took 20. Generative AI rewrote medicine, law, and software in 24 months. The pace is different. The models are moving so randomly and quickly that any 12-month detailed prediction is proven wrong almost immediately.

Consider the evidence. Google now processes 1.4 quadrillion tokens monthly. This figure was near zero just quarters ago. It could grow another hundredfold, or potentially even a thousand-fold, in a decade. This is not the slow, physical laying of track. This is an exponential-on-exponential kind of explosion, for the want of better descriptions.

The change is not just in scale, but in substance. The transformer architecture is a fundamental leap. It is bigger than the shift from analog to digital. It redefines the nature of computation itself. This is not just more compute. It is a different kind of compute. It is a discontinuity event. It is the leap from sequential CPU-based processing to massively parallel GPU-based processing. Legacy hardware is not just slow for this new task. It is increasingly useless in the new world, similar to mainframes and electronic typewriters, with the arrival of PCs.

Doomers trivialize this. They point to a student using AI to cheat on an essay. This is like pointing to the first car that broke down and declaring the automobile a fad. It is as if the telephone were just for gossip, or electricity just for lightbulbs. They are blind to the revolution in every profession. Medicine, law, science, and art are being rewired from the inside.

This kind of lazy parallel that one draws comparing GenAI to railways is bad science. It is allowed only in financial journalism. No scientist would use the average 19th-century tremor to predict the impact of an unprecedented 9.1 quake in a crowded, modern city. 

The Peril and Promise of the Cherry-Picked Past

History is not a verdict. It is an argument. And like any argument, it can be cherry-picked to prove anything. The doomers have their canon: the South Sea Bubble, Tulip Mania, the dot-com bust. These are their sacred texts, recited to prophesy doom. But this is a selective reading of a vast and contradictory library. Let us open a different volume.

The South Sea Bubble was a financial disaster. This is true. It is not the whole truth. The bubble was also a massive act of state-level financial engineering. The British government used the mania to consolidate its crippling national debt. The crash was brutal for speculators, but the state achieved its strategic goal. It emerged financially stronger. A mania, it turns out, can be a tool. And imagine if this is one of the lessons being examined in global government corridors, where the bigger worry is the seemingly unsolvable public sector debt trap.

The Nifty Fifty were a classic bubble. In 1972, fifty glamour stocks traded at insane P/E ratios. The subsequent crash eviscerated them. The narrative was sealed: a lesson in valuation hubris. But the full story is more revealing. A famous 1998 study asked what would have happened if you bought these stocks at their absolute peak in 1972 and held them for 25 years. The result is stunning. That portfolio would have returned 12.2% annually, almost in line with the index returns in the period. The bubble was in the timing, not in the long-term value of the best companies.

Japan’s bubble is another favorite. Yes, it crashed spectacularly. But it is also a warning against premature doom. For nearly three decades, from the 1960s to the late 1980s, Japan’s market was overvalued by Western standards. Doomers who called the top in 1970 were wrong for twenty years. The peak was only identifiable in hindsight. This example is a layup to not repeat one of the most quoted, worn-out phrases of all time: a market can remain irrational far longer than any skeptic can stay solvent.

This is the peril of the single story. For every bubble that popped, there is a boom that lasted for generations. For every crash that caused pain, there is a crisis that cleared the way for a new era of growth. History is a Rorschach test. The doomers see only skulls. But look again. You can just as easily see the foundations of the next empire.

Trained Firefighters in the Building: The Modern Policy Imperative

The doomers play with a historical deck that is missing its most powerful card: the modern state and utility of experience. In simpler terms, there were no doomers armed with the consequences of a collapse when Tulip prices were rising. The policymakers of 1929 had no examples of deflationary doom to learn from. The emerging market policymakers would have behaved differently in 1997 if they were allowed to think that multilateral agencies could be wrong, or Japanese money managers of the 1990s had the lessons of 2008 response available. The TMT-era monetarists did not have to worry about keeping their politicians’ fiscal funding at reasonable levels. 

Reading history differently and superficially, in 1720 or 1929, a financial crash was a natural disaster. Governments watched, helpless, as the contagion spread. The system was left to purge itself. This is no longer the case. The firefighters are now permanently stationed in the building, hoses at the ready. The lesson of 2008 was learned, for better or worse. The systemic collapse of a major financial market is now a political impossibility. At the first sign of a seizure, central banks will open the liquidity valves. They will do whatever it takes. For details, pls see our piece on the world of no ideological policymaker here

This creates a powerful, structural backstop. It is the policy put. It does not prevent corrections, but it fundamentally alters the nature of risk that cannot be captured by reading books of different times. One cannot compare the tendencies of a driver behind the wheel for the first time with those of an F1 driver on the track. A 1929-style uncontrolled deflationary spiral, like any other risk, including a comet hitting the earth into oblivion, is possible, but unlikely, as the states globally will do their darnedest to prevent it, knowing what they know.

This imperative is compounded by a new, overriding geopolitical reality. AI is not just another industry. It is the new domain of national power. It is the engine of economic and military supremacy. No major government can afford to let its domestic AI capabilities falter. A market crash that cripples the technological race would be viewed as a national security failure. The response would be swift and massive. Subsidies, direct investment, and protective measures would be deployed without hesitation.

This does not guarantee endless prosperity. It simply changes the game. The biggest risk is no longer a classic collapse. It is the risk of falling behind. The forces driving investment are no longer just profit and loss. They are sovereignty and survival (a detailed article here). And in that game, the rules of 1720s do not apply.

The Cost of Being Wrong

Let us be clear. We are not unabashed bulls. There are massive financial excesses. There are lofty expectations. There will be gut-wrenching cycles and reversals.

Our point is that this is a messy reality. It is a reality of both amazing technology progress and absolute excesses. And it cannot be navigated with "South Sea bibles." This, however, is not the most important point. The fixation on an imminent crash is a dangerous distraction. It is a sideshow.

The far greater sin is the trivialization of the change itself. It is the arrogance of dismissing a fundamental technological shift because one has read a book about tulips. It is the intellectual laziness of making the same time-invariant argument, quarter after quarter, while the world transforms beneath your feet.

The doomer’s arrogance is twofold. First, a disdain for the innovators and investors who are building the future with their sweat and capital. Second, a refusal to engage with the messy, complex details of the present, preferring the comfort of a pre-digested historical narrative.

They operate with a false purity. They believe that by forecasting doom, they bear no responsibility for being wrong. If the boom continues, they are merely early. If a crash comes, they are prophets. When a cyclical downturn inevitably happens, they will emerge from the crowd to say, "I told you so," behaving as if only those who read history are able to see what's happening.

In real life, the cost of this error is not just a missed investment return. The cost is strategic. It is the cost of inaction. A company that opts out of this race for fear of a bubble may become the next BlackBerry. A nation that hesitates may cede technological leadership for a generation. Risks exist in every direction. The greatest risk of all may be in opting out, armed with nothing but a chart of a 19th-century railway.

Once again, our mandate is not blind optimism. Nor do we advocate joining a blind rat race out of the fear of missing out. It is nuance. Articles like these could turn into dartboards when markets and economies crash with widespread pain sometime in the coming quarters. We all will make mistakes with consequences. But we all need to find the courage and the grit to make decisions based on a fluid present, not a fossilized past. We all will need to make our own, highly divergent decisions on personal journeys and constraints. Our point here is that few of these decisions should be based on glib tales of irrelevant times.

They should not start with the chants of prophets claiming how history always repeats, or at least rhymes.