Simba
On September 5, 1929, financial expert Roger Babson stood before a crowd and declared, “Sooner or later a crash is coming, and it may be terrific.” The market dipped 3%, the establishment dismissed it as a “healthy correction,” and two months later the entire economy imploded. On January 20, 2025, a Chinese AI lab called DeepSeek released models matching GPT-4’s performance for a fraction of the cost, tech stocks briefly wobbled, and Silicon Valley’s finest immediately declared it a “market recalibration moment” before returning to their speculative orgy. If you’re experiencing déjà vu, congratulations—you’re paying attention to history while everyone with a Series B term sheet is actively choosing to ignore it.
The Great Crash didn’t happen because markets randomly decided to collapse. It happened because an entire generation of investors took profits from real innovations—automobiles, radio, electrification—and funneled every dollar into speculative bets on future prosperity that existed only in their collective imagination. Today, Big Tech is taking profits from actual innovations—cloud computing, smartphones, electric vehicles—and burning $200 billion annually on AI models that are fundamentally sophisticated autocomplete engines. The playbook is identical. The only difference is that this time, they’re firing humans to make room in the budget for stochastic parrots.
The Evidence: Following the Money From Prosperity to Panic
Let’s establish the structural parallels, because the similarities aren’t superficial—they’re architectural.
The “New Era” Delusion: The Roaring Twenties were defined by breathless optimism about permanent prosperity. Industrial expansion had generated real wealth, and investors convinced themselves that traditional economic cycles no longer applied. Sound familiar? The 2010s generated real wealth through cloud infrastructure, mobile advertising, and SaaS businesses. Now, in 2025, that same establishment has convinced itself that AI represents a “paradigm shift” that justifies infinite investment with zero return requirements.
The Speculation Cascade: In the 1920s, industrial profits weren’t reinvested in industrial capacity—they were dumped into stock speculation. In 2025, Big Tech isn’t reinvesting cloud profits into better cloud services. Instead:
- Meta made billions from advertising, failed spectacularly with the Metaverse, and is now spending $14 billion to acquire a data labeling company while pouring resources into Llama and ASI initiatives. They’re not improving Instagram; they’re hiring AI researchers at $2 million annual salaries.
- Google declared “Code Red” when ChatGPT launched, pivoted from Bard to Gemini to Nano to VEO3, and invested billions in Anthropic—OpenAI’s rival. Every dollar spent on this AI arms race is a dollar not spent on making Search actually useful.
- Amazon and Microsoft both invested billions in Anthropic and OpenAI, respectively. These aren’t research grants; they’re panic hedges.
- Tesla’s Elon Musk, a co-founder of OpenAI, is now running xAI with Grok while simultaneously pursuing humanoid robots and robotaxis—because when one speculative bet isn’t enough, why not three?
This is the exact pattern from 1929: take proven revenue streams and gamble them on unproven futures. The difference is that in 1929, they at least pretended the investments would generate returns.
The Dismissed Warning: British Chancellor Philip Snowden called American markets a “speculative orgy” in October 1929. The market crashed days later. Today, MIT research explicitly shows that AI investments are not yielding productivity gains. Companies are spending billions on technology that demonstrably doesn’t improve outcomes. The response from Silicon Valley? Double down and fire the humans who might question the ROI.
The Pre-Emptive Unemployment Crisis: The 1929 crash caused mass unemployment. The AI bubble is causing mass unemployment before the crash. Duolingo fired contractors to replace them with GPT-4. Consulting firms, accounting practices, and law firms are shedding staff to fund “AI transformation initiatives.” The cruelty is that these companies aren’t replacing humans with superior intelligence—they’re replacing them with expensive autocomplete that can’t understand context, learn from mistakes, or apply judgment. It’s a stochastic parrot that costs $20 million in compute annually.
The DeepSeek moment in January 2025 was this cycle’s Babson Break—a warning shot that exposed the speculative excess. DeepSeek proved that AI performance isn’t about spending more; it’s about engineering competence. The market dipped briefly, everyone called it a “healthy correction,” and then Big Tech resumed burning cash on GPU futures and model training runs. This is 1929 in high definition.
The Absurdity: Firing Humans to Fund Algorithms That Can’t Think
Here’s where the satire writes itself, because the dialogue is too absurd to fabricate.
The Delusional CEO (composite of several real executives): “We’re investing $50 million in AI infrastructure to stay competitive. Yes, I understand the MIT study shows no productivity gains, but those researchers don’t understand the transformative potential of large language models. We’re betting on the future.”
The Desperate CFO: “Sir, we could use that $50 million to retain the 200 employees you just laid off, who actually understand our business and generate revenue.”
The Delusional CEO: “That’s exactly the kind of old-economy thinking that will get us disrupted. AI is the new electricity. Do you want to be Kodak?”
The Cynical Engineer (who actually builds these systems): “It’s tokens and matrix multiplication. It doesn’t understand anything. It’s pattern matching at scale.”
The Visionary VC: “That’s what they said about the internet in 1995. This is the same revolution.”
The Cynical Engineer: “No, the internet actually did something new. This is just expensive regression to the mean with a chatbot interface.”
The Visionary VC: “Exactly the kind of skepticism that missed the last wave. We’re going all-in.”
This is the groupthink that defines every bubble. The believers aren’t stupid—they’re financially incentivized to sustain the delusion. VCs need portfolio companies to justify their valuations. CEOs need growth narratives to justify their equity packages. No one has an incentive to admit that the emperor is naked and the “AGI” they’re funding is an extremely expensive word predictor.
The “speculative orgy” Snowden described in 1929 was at least investing in companies that built things. The AI orgy is investing in companies that train models to plagiarize the internet at scale and call it “reasoning.” It’s speculation on speculation, funded by firing the humans who generate actual value.
The Judgment: We’re Watching the Pre-Crash, Not the Correction
Here’s the damning verdict: The AI bubble isn’t a new paradigm—it’s a photocopied playbook from 1929, with worse fundamentals and better marketing.
The 1920s boom was built on real innovations that genuinely transformed society. Automobiles changed transportation. Radio changed communication. Electrification changed production. The speculation was irrational, but the underlying technology was revolutionary. The 2020s AI boom is built on models that are fundamentally autocomplete engines that don’t understand, don’t reason, and can’t learn. They’re stochastic parrots that cost billions to train and millions to run, deployed by companies desperate to justify shedding labor costs while maintaining growth narratives.
The DeepSeek moment proved what every honest engineer already knew: you don’t need $100 billion in compute to match GPT-4’s performance. You need competent engineering. But admitting that would require Big Tech to admit they’ve been spending like drunken sailors on a casino floor, funding a speculative bubble with shareholder money and employee livelihoods.
The crash isn’t coming because AI doesn’t work—it’s coming because the economics don’t work. You cannot indefinitely spend billions on technology that generates no measurable returns while firing the humans who generate actual revenue. At some point, the CFOs will demand ROI. The VCs will demand exits. The market will demand proof. And when that moment arrives, the executives will act shocked, the media will write think pieces about “unforeseen risks,” and the workers who were fired to fund the bubble will be told to “learn to code” by algorithms that can’t code themselves.
The warnings are already here. MIT’s research. DeepSeek’s efficiency. The lack of measurable productivity gains. Just like Babson in 1929 and Snowden’s “speculative orgy” comment, they’re being dismissed as noise by people who need the music to keep playing. History doesn’t repeat itself, but it absolutely rhymes—and right now, it’s rhyming in iambic pentameter.
The crash is coming. It may be terrific.
Has your company fired humans to fund an “AI transformation” that’s delivered zero ROI? Are you a VC who knows the bubble is unsustainable but can’t stop because your LPs demand deployment? Have you watched executives compare GPT-4 to electricity while laying off the people who kept the lights on? Share your pre-crash horror stories below.
This article is just a tremor. The earthquake is coming.
The patterns of hype, deception, and greed laid bare here are part of a much larger story. They are the evidence file for the great deception of our time.
The full, unvarnished truth is detailed in the forthcoming book from our founder, Simba Mudonzvo:
The Gilded Cage: How the Quest for Artificial Intelligence Became the Greatest Deception in Human History.
Stay tuned. The reformation is coming.
GIPHY App Key not set. Please check settings