The AI Bubble Explained Like You're 5 — Transcript
Full transcript
- 0:00Picture this. The CEO of the most hyped
- 0:03tech company on Earth just admitted his
- 0:05$200 per month subscription loses money
- 0:08on every single customer. Not a little
- 0:10money, not break even with growth
- 0:12potential. Loses money. And Wall
- 0:16Street's response, they valued his
- 0:18company at $500 [music] billion. Anyway,
- 0:21welcome to the AI bubble, where we went
- 0:23from cool chatbot to $7 trillion gamble
- 0:26in 24 months, and why the people writing
- 0:28the checks might be in for the rudest
- 0:30awakening since 2008.
- 0:35November 30th, 2022.
- 0:38Just a regular Wednesday. Unremarkable
- 0:40cold brew flowing. Engineers debugging
- 0:43code. Then Open AAI dropped chat GPT.
- 0:47Within 48 hours, the internet had a new
- 0:49obsession. People were using it to write
- 0:52breakup texts, debug code, explain black
- 0:54holes, and generate bedtime stories.
- 0:57Dad, I'm talking to an AI that knows
- 1:00everything. It just explained quantum
- 1:02entanglement using Minecraft terms. And
- 1:04did I hear Minecraft? It has 1 million
- 1:07users in the first 5 days of launch. 2
- 1:10months later, 100 million users. Fastest
- 1:13growing app in history. Instagram took
- 1:172.5 years to hit 100 million. Facebook
- 1:20took 4.5 years. Chat GPT did it in 60
- 1:24days and Silicon Valley absolute panic.
- 1:27Inside Google's headquarters, executives
- 1:29were having a meltdown. We invented this
- 1:32technology. The transformer model that
- 1:35was us. The research paper has our names
- 1:37on it. Then why is my son using chat GPT
- 1:40and not Google AI? because we were too
- 1:43scared to release it. Well, release
- 1:46[music] something now. Call it barred,
- 1:48Gemini, whatever. Ship it. Google
- 1:51scrambled. They had the tech. They'd had
- 1:54it for years, but they were terrified of
- 1:56what it might say. Of lawsuits, of bad
- 1:59press. Open AI had no such fears. They
- 2:02shipped first. And in tech, first mover
- 2:05advantage is everything. Within weeks,
- 2:08Microsoft threw $10 billion at OpenAI
- 2:11with another $3 billion on standby.
- 2:13Google panic launched Bard, which
- 2:16immediately hallucinated in its first
- 2:17[music] demo. Amazon started dumping
- 2:20billions into Anthropic. We're a dog
- 2:23walking app, but the root optimization
- 2:24uses machine learning. Say no more.
- 2:27Here's $50 million. Don't you want to
- 2:30see our demo? Shut up and take my money.
- 2:33By mid 2023, if your slide deck
- 2:36mentioned neural networks or large
- 2:38language models, you were basically
- 2:39printing money. Companies with zero
- 2:42revenue, no product, just a team and a
- 2:45Google doc or raising $100 million
- 2:48rounds. The bubble gave us technological
- 2:50marvels like AI powered toothbrushes, AI
- 2:54recipe generators, AI dating coaches.
- 2:57One company raised $30 million to build
- 2:59an AI that summarizes potentially other
- 3:02AI outputs. We'd reached peak absurdity.
- 3:06But the real insanity, [music] the GPU
- 3:09shortage, these little computer chips,
- 3:11which look like harmless circuit boards,
- 3:13became more valuable than gold or oil.
- 3:16Nvidia became one of the most valuable
- 3:18companies on Earth. Their CEO Jensen
- 3:20Huang started wearing leather jackets to
- 3:22conferences like he was the Mcjagger of
- 3:25semiconductors and he kind of was
- 3:27[music]
- 3:28renting a single top tier age 100 GPU
- 3:31costs about $3 an hour. Sounds cheap,
- 3:34right? But startups don't need one, they
- 3:36need hundreds. A small startup trying to
- 3:39train [music] just one model might rent
- 3:41a 500 chip cluster. That's $1,500 per
- 3:44hour. $36,000 per day. over $1.1 million
- 3:49per month just for renting GPU. Yet,
- 3:52companies were still hoarding GPUs like
- 3:54doomsday [music] preppers stockpile
- 3:56canned beans. Nvidia couldn't
- 3:58manufacture them fast enough. The wait
- 4:00list was at one point up to 12 months
- 4:02long. Black markets [music] emerged.
- 4:05Seriously, people were smuggling $100
- 4:07bills across borders. By 2024,
- 4:10everything [music] was AI powered. Your
- 4:13toaster burns bread with 10% more
- 4:15efficiency. [music]
- 4:16Your calendar app suggests meeting times
- 4:19using algorithms. Your grocery store
- 4:21selfch [music] checkout. Just a camera,
- 4:23but now it's machine learning. The party
- 4:26was in full swing. Champagne flowing.
- 4:29Checks clearing. [music]
- 4:30Then somebody asked the uncomfortable
- 4:32question. How much does this actually
- 4:34cost?
- 4:36January 2025.
- 4:39Sam Alman, CEO of OpenAI, face of the AI
- 4:42revolution, tweeted something that made
- 4:44every investor's blood pressure spike.
- 4:47Insane thing. We are currently losing
- 4:49money on OpenAI Pro subscriptions.
- 4:52People use it much more than we
- 4:53expected. Yep, you heard it right. Open
- 4:56AAI charges $200 per month for their pro
- 4:59tier, and they lose money on every
- 5:01single subscriber, which means the more
- 5:04successful they are, the more money they
- 5:06lose at this rate. When they launched
- 5:08ChatGpt Plus at $20 per month, their
- 5:11entire pricing strategy was test two
- 5:14prices, $20 and $42. [music] See which
- 5:17people complain about less. Pick that
- 5:19one. No market research, no cost
- 5:22analysis, just Vibes. That's how a $500
- 5:25billion of company priced its flagship
- 5:27product, Vibes. But wait, it gets worse.
- 5:31Open AAI's 2024 finances revenue $3.7
- 5:36billion losses $5 billion. They didn't
- 5:40just fail to make a profit. They lost
- 5:42more money than they made by a lot. Now,
- 5:45here's where it gets truly insane.
- 5:47Industry forecasts show AI
- 5:49infrastructure will require $6.7
- 5:51trillion by 2030. Let me break that
- 5:54down. $5.2 $2 trillion directly for AI
- 5:58infrastructure, GPUs, accelerators,
- 6:00power upgrades, cooling, and data center
- 6:03construction. While the remaining $ 1.5
- 6:05trillion goes to traditional nonAI cloud
- 6:08infrastructure, the four largest tech
- 6:10companies alone are expected to spend
- 6:13$350 billion in 2025,
- 6:16one year. And here's the kicker. To
- 6:19justify these costs, the AI industry
- 6:21needs to generate $2 trillion in annual
- 6:23revenue by 2030. Current total AI
- 6:26revenues, somewhere around $200 billion.
- 6:30That's a 10 times gap.
- 6:33100 billion. Yep, [music] that's the
- 6:36number. Sam, our projections show us at
- 6:3920 billion in 2025. You're proposing we
- 6:42hit 100 billion by 2029. [music] It's
- 6:45the necessary growth trajectory.
- 6:48That's not a trajectory, Sam. That's a
- 6:50hockey stick drawn by someone who's
- 6:52never played hockey. Do you know what it
- 6:54costs to even attempt this?
- 6:57115 billion in cash burn by 2029.
- 7:00When do we become profitable?
- 7:032029, [music] maybe 2030 or 2031?
- 7:07And if revenue doesn't hit $100 billion,
- 7:11if this hockey stick of yours snaps,
- 7:13then we raise more money. From whom? Who
- 7:16is left? Microsoft? Saudi Arabia? A
- 7:20really optimistic pension fund? Plenty
- 7:22of fish in the sea, guys. Since 2021,
- 7:26private investors have pumped hundreds
- 7:27of billions into AI startups. Most of
- 7:30that went to [music] just 10 companies
- 7:32whose combined valuations swelled by
- 7:34nearly $1 trillion. Everyone's betting
- 7:37they'll become the next Google. Then the
- 7:39failures started rolling in. Chat GPT
- 7:42hallucinated. Tech speak for completely
- 7:44making things up and cited research
- 7:46papers [music] that don't exist, legal
- 7:48cases that never happened, and
- 7:50statistics from an alternate dimension.
- 7:52Google's AI told users to put glue on
- 7:55pizza to make cheese stick better. It
- 7:57learned this from a Reddit joke. Yes,
- 8:00really. The lawsuits arrived next.
- 8:02artists, writers, musicians,
- 8:05journalists, all asking the same
- 8:06question. [music] Did you train your AI
- 8:09on my copyrighted work? Answer: Um, yes.
- 8:14The New York Times sued Open AI. Getty
- 8:17Images sued Stability AI. Authors sued
- 8:21everyone. Musicians sued everyone else.
- 8:25[music] We didn't steal. We learned from
- 8:27publicly available data. You scraped
- 8:30[music] the entire internet without
- 8:31permission. and used it commercially.
- 8:34Well, when you say it like that, it
- 8:35sounds
- 8:36>> [music]
- 8:36>> illegal. By mid 2024, the cracks were
- 8:39showing. The vibe shifted from euphoria
- 8:41to something more dangerous. Doubt.
- 8:48The hype was loud. The math was ugly.
- 8:51And now everyone had the same
- 8:52uncomfortable question buzzing in their
- 8:54heads like a mosquito at 3:00 a.m. So,
- 8:58when does this whole [music] thing
- 8:59explode? This looks like a bubble. The
- 9:02losses are insane. The valuations are
- 9:04dulu. But this bubble isn't popping the
- 9:07same way the dot bubble did. Why?
- 9:09[music] Because the people inflating it
- 9:11are also the people with the biggest
- 9:12wallets on Earth. Unlike 1999 when
- 9:16startups had zero revenue and 17
- 9:18foosball tables, today's AI bubble is
- 9:20being funded by big tech companies that
- 9:22actually make money. Real money. Buy an
- 9:26entire country if we feel bored. Money.
- 9:28Microsoft prints billions from Azure.
- 9:31Google prints billions from ads. Meta
- 9:34prints billions from showing you memes
- 9:35your cousin already sent on WhatsApp.
- 9:38These giants aren't [music] investing in
- 9:39AI because it's profitable. They're
- 9:42investing because they're terrified of
- 9:44being the ones who didn't. If AI becomes
- 9:47the next [music] iPhone moment, the next
- 9:49Google moment, the next entire world
- 9:51changes overnight moment, then the
- 9:53company that slowed down becomes
- 9:54Blackberry. And nobody wants to be
- 9:57Blackberry. Think of big tech like poker
- 9:59players who all brought unlimited chips.
- 10:02If Microsoft folds, Google wins. If
- 10:06Google folds, Amazon wins. If Amazon
- 10:09folds, Meta [music] wins. No one blinks,
- 10:13no one folds, everyone keeps raising the
- 10:16stakes. And that brings [music] us to
- 10:18the real uncomfortable question. If
- 10:20everyone keeps spending, how do bubbles
- 10:22ever pop? Short answer is they burst
- 10:26when a real world event forces everyone
- 10:28to stop pretending. [music] And
- 10:30historically, that event usually comes
- 10:33in three forms. One, the money suddenly
- 10:36disappears. The dot bubble didn't
- 10:38collapse because websites were dumb. It
- 10:40collapsed because the cash spigot shut
- 10:42off. [music]
- 10:43Interest rates rose. Borrowing got
- 10:46expensive. Funding vanished. And
- 10:48thousands of [music] companies
- 10:48evaporated overnight. If interest rates
- 10:51rise anytime soon and big techs tighten
- 10:53their budgets, the AI bubble ends
- 10:56instantly.
- 10:58Two, a big [music] embarrassing meltdown
- 11:00happens. Every bubble has a moment. One
- 11:03spectacular failure that freaks everyone
- 11:05out. Do had earnings disasters. 2008 had
- 11:08layman. Crypto had FTX. AI's version
- 11:12could be a giant lab revealing massive
- 11:15losses, a botched model launch or a
- 11:18lawsuit that freezes everything. So,
- 11:21where are we [music] right now? Big tech
- 11:24is still spending. No meltdown earnings
- 11:26call yet. No FTX [music] level scandal.
- 11:29Reality is showing cracks, but not
- 11:31enough to break confidence. So, yes,
- 11:34this bubble is inflated. Yes, the
- 11:37numbers are insane, but it won't burst
- 11:39because of hype. It bursts when one of
- 11:42the giants finally says, "All right,
- 11:44that's enough." Until then, the bubble
- 11:48floats. We're not at the explosion.
- 11:51We're at the tightening. The inhale
- 11:53before the scream. The tension before
- 11:55the snap. The episode right before the
- 11:58season finale.
About this transcript
This page contains the full transcript of The AI Bubble Explained Like You're 5 by Crayon Capital, generated from the public captions YouTube serves with the video. The transcript has 1,689 words across 278 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.