ChatGPT hit one billion users in three years. The internet took thirty-six.
And now nobody can stop investing.
Meta is spending $600 billion through 2028. OpenAI and Oracle announced $500 billion for Project Stargate. Amazon is investing over $25 billion per quarter. More than a trillion dollars in motion.
The forcing function is complete. With a billion users watching, being left behind means obsolescence. So every company must invest—regardless of whether the math works.
A trillion dollars for 2% growth
Daron Acemoglu at MIT estimates AI will add 0.66% to total factor productivity over ten years. Penn Wharton Budget Model projects 1.5% GDP increase by 2035.
A trillion dollars for 2% growth is the largest capital misallocation in modern history. An order of magnitude worse than the dot-com bubble.
But if AGI arrives—artificial general intelligence, post-scarcity abundance—then a trillion is negligent. Too small, not too large.
The investment scale itself reveals the implicit bet. These aren’t productivity improvement budgets. They’re transformation budgets.
The circular escape hatch
Money circulates between Nvidia, OpenAI, Oracle, and Microsoft. Nvidia invests in OpenAI. OpenAI buys Nvidia chips. Microsoft funds OpenAI. Oracle hosts the infrastructure.
It’s not irrational. It’s Big Tech using each other’s balance sheets to buy runway while unit economics remain broken.
OpenAI loses money on every ChatGPT query. So does Anthropic, and every frontier model lab. The only entities profiting are the infrastructure layer.
Training a frontier model costs $100 million to $1 billion. Post-training adds tens of millions more. Grid constraints require $720 billion in power upgrades.
Every major technology had revenue models before infrastructure. ISPs charged for internet access from day one. Utilities charged when wires reached homes. Telecoms charged monthly plans before building 3G networks.
AI gave away service to a billion users. Then started building trillion-dollar infrastructure.
The prisoner’s dilemma at scale
The trap has three outcomes. If one company invests $600 billion and competitors don’t, the investor gains five years of capability lead—potentially insurmountable. If all companies invest and AGI doesn’t materialize, they all lose billions but maintain relative position. If a company sits out while others invest, it faces certain obsolescence.
The paradox: no one knows if AI capability translates to profit. But no company can risk finding out their competitors were right and they were wrong.
The game theory forces a single conclusion: everyone must invest.
Once hundreds of billions are committed, only abundance justifies the spend. There’s no middle ground where this “works well enough.”
Either AGI arrives and every dollar was mandatory. Or 2% GDP growth makes this the biggest capital allocation failure in economic history.
Public language talks about incremental improvements. Investments reveal bets on abundance.
Someone is very wrong. Either Acemoglu and the consensus economists. Or Zuckerberg, Altman, and Ellison.
We’re witnessing a 3-year adoption curve colliding with a $1 trillion infrastructure buildout to produce an outcome ranging from transformative abundance to economic disruption.
And the forcing function guarantees nobody can walk away.
That’s the bet hiding in plain sight. Not whether AI works. Whether it transforms everything—or becomes the most expensive mistake in corporate history.
References
“OpenAI CEO Sam Altman revealed that ChatGPT’s user base had more than doubled in just a few weeks, pushing it past the 1 billion mark.” Stan Ventures (June 2025)
“It took 36 years for the Internet to get its first billion users.” Nielsen Norman Group (September 2005)
“Meta Platforms CEO Mark Zuckerberg has said the company expects to spend at least $600 billion on U.S. data centers and related infrastructure by 2028.” RCR Wireless News (September 2025)
“The Stargate Project is a new company which intends to invest $500 billion over the next four years building new AI infrastructure for OpenAI in the United States.” OpenAI Official Announcement (January 2025)
“[Amazon] Capital investments were $26.3 billion in the fourth quarter, and we expect a similar quarterly rate in 2025.” The Next Platform (February 2025)
“Using existing estimates on exposure to AI and productivity improvements at the task level, these macroeconomic effects appear nontrivial but modest—no more than a 0.66% increase in total factor productivity (TFP) over 10 years.” Daron Acemoglu, “The Simple Macroeconomics of AI,” NBER Working Paper No. w32487 (May 2024)
“We estimate that AI will increase productivity and GDP by 1.5% by 2035, nearly 3% by 2055, and 3.7% by 2075.” Penn Wharton Budget Model (September 2025)
“OpenAI during the first half of 2025 collected $4.3 billion in revenue while still posting a net loss of $13.5 billion during that six month period.” The Information, reported in The Register (October 2025)
“CEO Sam Altman vaguely pegged the training cost [for GPT-4] at ‘more than’ $100 million… Anthropic CEO Dario Amodei suggested in August that models costing over $1 billion would appear this year and that ‘by 2025 we may have a $10 billion model.’” Fortune (April 2024)
“Llama 3.1 (Q3 2024) >$50M: similar preference data to Llama 2, a ~200-person post-training team, larger models, etc. The number could be much higher.” Nathan Lambert, Interconnects Newsletter (January 2025)
“Goldman Sachs Research estimates that about $720 billion of grid spending through 2030 may be needed… These transmission projects can take several years to permit, and then several more to build, creating another potential bottleneck for data center growth.” Goldman Sachs Research (February 2025)
“In addition to the $500 billion Stargate project, the startup on Monday announced an equity investment deal with Nvidia that will add an estimated $500 billion worth of data centers in the coming years.” CNBC (September 2025)