The AI 'Bubble' Critics Sound Exactly Like Idiots Who Called the Internet a Fad
Spoiler alert: The revolution doesn't pause for permission from pessimists.
Let me be blunt.
The articles circulating about AI's inevitable crash reads like a compilation of the greatest hits of every tech bubble prediction since 1995.
Same tired playbook.
Same breathless warnings.
Same fundamental misunderstanding of what's actually happening.
I've built 110+ companies across three major technology cycles.
Lived through the dot-com crash.
Watched crypto's meteoric rise and face-plant.
And here's what bubble doomsayers consistently miss: real utility doesn't disappear when speculation ends.
The Crypto Comparison That Reveals Everything
Crypto's collapse was inevitable because most projects solved problems nobody actually had.
NFT monkey pictures?
Decentralized autonomous organizations for pet grooming?
Come on.
But compare that circus to what AI delivers today: Fortune 500 companies cutting operational costs by 40-60% while improving output quality.
Manufacturing plants are reducing defect rates from 12% to 0.3%.
Financial services firms are processing loan applications in minutes instead of weeks.
This isn't speculative digital gold.
It's productivity infrastructure.
The authors mentions Amazon getting "twenty cents for every dollar" in AI development returns.
That's remarkable ROI for infrastructure investment in year one.
When Amazon built AWS, critics called it a distraction from their "real" retail business.
Today, AWS generates a bigger chunk of operation profit than their entire retail operation.
Revolutionary infrastructure always looks expensive until it becomes indispensable.
Why Founder DNA Actually Matters More Than Ever
Here's where the analysis gets particularly shallow.
Dismissing Mira Murati's $12 billion valuation because she hasn't shipped product yet misses the fundamental dynamics of AI development.
In previous bubbles, execution mattered more than vision.
Any competent team could build a website or crypto exchange.
AI is different.
The gap between brilliant researchers and average technologists isn't 2x or 5x—it's infinite.
Most people can't architect systems that process human language at scale.
They can't design neural networks that generalize across domains.
They can't solve alignment problems that determine whether AI helps humanity or destroys it.
When you're betting on technologies that could reshape civilization, the inventor's brain matters more than their current product lineup.
Would you invest in Einstein's theory of relativity or demand he build a nuclear reactor first?
The Infrastructure That Outlasts The Hype
The dot-com bubble built the infrastructure we've been running on for 25 years.
Fiber optic networks.
Data centers.
Internet protocols.
Payment systems.
When the speculation ended, the infrastructure remained. Google, Amazon, and PayPal emerged stronger because they built real utility on solid foundations.
Today's AI "bubble" is creating similar lasting infrastructure.
Graphics processing units optimized for parallel computation.
Cloud platforms designed for machine learning workloads.
Open-source models that democratize artificial intelligence.
NVIDIA's stock might be overvalued, but their chips aren't going anywhere. The computational capacity being built today will power applications we haven't even imagined yet.
The Wrapper Economy Strawman
Criticising 2024 AI companies for building "ChatGPT wrappers" is like criticising 2001 internet companies for using TCP/IP. Of course, early applications build on existing foundations. That's how technology evolution works.
But here's what the wrapper critics miss: most successful AI companies today aren't wrapping external APIs.
They're fine-tuning open-source models with domain-specific data.
Healthcare AI companies train on medical imaging datasets.
Legal AI systems learn from case law and regulatory documents.
Manufacturing AI analyzes sensor data from specific industrial processes.
This isn't API arbitrage. It's specialized intelligence development using proven foundation models as starting points.
The most valuable AI applications will emerge from deep domain expertise, not general-purpose chatbots.
Training vs. Inference: The Cost Structure Revolution
The analytics fixates on training costs while missing the bigger economic story.
Training large language models is expensive, but it's a one-time cost amortized across millions of users.
Inference—actually running the models—is where the economics get interesting.
Processing costs drop exponentially as hardware improves and algorithms optimize. GPT-4 inference costs fell 90% between launch and today.
This creates a virtuous cycle.
Better models attract more users.
More users justify infrastructure investment.
Better infrastructure reduces processing costs.
Lower costs enable more applications.
Traditional software scaled linearly with users.
AI scales exponentially while costs scale logarithmically.
The IPO Red Herring
Pointing to weak IPO activity as evidence of bubble collapse ignores fundamental market changes. The best companies stay private longer because they can.
SpaceX is worth $350 billion without going public. Stripe processes hundreds of billions in payments while remaining private. ByteDance built TikTok without needing public markets.
Private capital markets provide liquidity without regulatory overhead.
Why would AI companies rush to go public when they can raise billions at attractive valuations from sophisticated investors?
The IPO market isn't dead because companies are failing. It's irrelevant because the best companies don't need it.
What Bubble Critics Miss About Failure Rates
Yes, 90% of AI startups will fail. Startups have always had 90% failure rates. This isn't news.
But the 10% that succeed will generate returns that dwarf the losses from the failures.
Amazon's success more than compensated for Pets.com's implosion.
Google's dominance justified every failed search engine that preceded it.
The authors seems shocked that companies with no products can raise massive rounds. I'm more shocked that anyone finds this shocking.
Venture capital has always been about funding potential, not proven results.
The Real Bubble Nobody Discusses
Here's the actual bubble: human labour costs in knowledge work.
When AI can generate legal briefs, financial analyses, and marketing strategies at 1% the cost of human experts, what happens to traditional consulting?
Law firms?
Investment banks?
The bubble isn't in AI valuations. It's in service businesses built on information asymmetry and labor multiplication. Those business models just became obsolete.
The companies "burning billions on AI" aren't speculating on future potential. They're building defensive moats against technological unemployment.
Why This Time Actually Is Different
Every bubble predictor claims "this time is different" right before the crash.
But sometimes, this time really is different.
The dot-com revolution created digital communication infrastructure.
The mobile revolution put computers in everyone's pocket.
The AI revolution creates systems that can think.
Not simulate thinking. Actually process information, recognize patterns, generate insights, and make decisions at speeds humans can't match.
This isn't about faster horses. It's about inventing the automobile.
The Real Risk Nobody's Talking About
The bubble risk isn't that AI companies will fail. It's that successful ones will succeed too completely.
When artificial intelligence systems can perform most cognitive tasks better and cheaper than humans, what happens to human-centric economic systems? When AI companies control the means of intellectual production, how do wealth and power redistribute?
These questions make Mira Murati's $12 billion valuation look quaint by comparison.
The AI transformation isn't a bubble that will burst. It's a phase transition that makes previous bubbles look like warm-up acts.
Smart money isn't fleeing AI.
It's positioning for a world where artificial intelligence becomes as fundamental as electricity.
The revolution is just getting started.
And the doomsayers will be the last to notice when it succeeds.
Howgh:)
Connect with me at [LinkedIn Profile] for strategic AI implementation insights that separate transformation from hype. The future belongs to those who build it, not those who predict its failure.
Join 25k+ Startup Founders, Fortune 500 Executives, UHNW investors, and VC/PE General Managers understanding the path to General AI in our AI Of The Coast Newsletter: