AI and OpenAI existed before 2022, but got the boom from ChatGPT and generative AI. On the other hand, people who were in the boom got the ride of a lifetime, and those on the other side of the AI Boom saw a downfall, but the ones in the middle, the enterprise companies, found a way to balance it, in fact.
The sectors that got severely impacted were VC-driven and Google-driven, i.e., internet companies. Google made sure because ChatGPT worked directly inverse
Post the First wave and reaching parity with Gemini with ChatGPT, the search for differentiation came with AI infrastructure, i.e data centers and chips, and a memory boom
Google has 2 lakh employees, but before you understand the AI bubble, you must understand AI tokens, AI context window, and NVIDIA-driven AI factories
What are Tokens in AI
AI token is the smallest unit of text that the AI models process, and 1 token is equal to 4 characters
Every time you send a text to an AI chatbot, they break it down into tokens to understand the text and then generate a response.
1000 AI tokens are used to model 750 words
AI tokens are billed for both reading the input and writing the output
What is a context window in AI LLMS
The context window is the maximum number of tokens that the model can hold in short-term memory, with 8000 tokens, which is equal to 6000 words. Beyond this, the conversation with a chatbot goes on, and it starts losing the earliest part of the chat.
AI tokens are billed in API, with OpenAI charging GPT-5.6 as $5 for 1 million tokens, Anthropic charges Fable 5 as $10 for a million tokens, which Anthropic calls Mtok
Google took advantage of it to do harm first, and venture capitalists took advantage of this wave, and then others used it to impact productivity in enterprises, which hardly had much impact, and they want more people where FOMO and scare will work
What are AI Factories
AI factories are data centers that have an infrastructure that is specialized in ways that NVIDIA GPUs help in reducing cost per token by 20x
NVIDIA claims that its GPU and some software optimization (software that should ideally be optimized by AI, considering now NVIDIA means AI, but they ignore it) help reduce Cost per Token.
First, this wave went up in 2022, and then they started to keep a circular trade between them
The smart ones know what the dumb vastness knows about and use AI, and CEOs know very well the trick is to balance both sides
What Is Tokenmaxxing
AI tokenmaxxing was used by software engineers in Silicon Valley to maximize AI token burning as a fake display of productivity, especially with the culture of leaderboards used to encourage AI use AI tokenmaxxing came to public knowledge when Meta’s leaderboard strategy came to public where records were
- Meta target was 75% AI-assisted code
- Meta Leaderboard topper used 281 billion tokens in one month
- Average Software engineer in Silicon Valley used 50 million tokens per month
- $6 billion wasted on AI by American engineers, as per Jellyfish; that doesn’t really matter for America since it is a small amount for top American companies
Main Companies used this practice: Google America, Disney ESPN, Meta America, Microsoft America
Tokenmaxxing came to an end when Uber CFO put it in public that he ran out yearly quote due to price increase done by AI companies
Tokenmaxxing was a subsidy to encourage American users to use it; now CFOs don’t want to increase the quantity of subsidy to drive revenue of AI companies and increase the IPO price
As of now, the idea is divided; a part of Silicon Valley wants to continue as it is, the part of finance just wants to be more practical, another 50-50 question, and a good gambit which all the CEOs are following and are sure to follow that
This is not a consultant answer that you get from IBM
Better answer and better visit the sources
What is AI Boom vs AI Bubble
AI boom and AI bubble are essentially about the internet, the absolute King Google, and the VCs who funded and created this, and will make money out of it, and Americans who enabled it
The future is not about AI; it is about how people will continue to do so. It will never be over, but the hype is the bubble that we can stop
Many people can’t take this many cursed moments. How exactly it helped ChatGPT and then new Anthropic grow is the same set of curses which brought down
I mean, ask Siri AI a question: why is 2024 just a blip in memory that can only be allowed for how big Google and Apple are, AI bubble proof
Again, the smartest people in the world have a better hypothesis on this, and so does Reddit; mine is plain and simple
Just like everything else, this is an AI product market fit driven by subsidy and an event causing a bump, and they will be back to doing the same thing again
What is next in AI
Open source AI models can help reduce token costs, so it is a debate right now
Also Try: