# [AI is a Pascal's Wager for companies where the upside is potentially infinite and missing out could be a bad outcome, so the best bet is to buy GPUs](https://x.com/tsarnick/status/1854258802071327159)
***
<iframe title="Ep19. State of Venture, AI Scaling, Elections | BG2 w/ Bill Gurley, Brad Gerstner, & Jamin Ball - YouTube" src="https://www.youtube.com/embed/tBU5UPI03Rg?feature=oembed" height="113" width="200" allowfullscreen="" allow="fullscreen" style="aspect-ratio: 1.76991 / 1; width: 100%; height: 100%;"></iframe>
#### [[👤Brad Gerstner]]:
There are growing benefits to having large-scale resources, both in terms of data and computing power.
Pascal’s Wager is a helpful way to think about this. Pascal, a French philosopher, argued that even if he didn’t know if God existed, it made sense to believe in God. If God exists and he believes, he gains eternal life. If God doesn’t exist, he loses nothing. But if God does exist and he doesn’t believe, he risks eternal punishment. So, the potential upside of believing is huge.
Applying this to AI, if the potential benefits of AI are massive (like "infinite" or "eternal" outcomes), and all your competitors are investing in it, then it makes sense to invest heavily in AI too. Instead of spending money on things like share buybacks or dividends, you should invest in AI resources, like GPUs (graphics processing units), to stay competitive.
#🤖Brad_Gerstner
***
- [[The cost per query in AI models has decreased by 100x in the past 2 years and quality will improve as hallucinations decrease 10x per year]]
- [[NVIDIA are continuing to increase chip supply and Blackwell's improved performance will help meet the demands of increasing AI model size, diversity, and capabilities]]
- [[Data centers are the most important asset in training AI models and no-one is ever going to train a frontier model on the same data center twice because by the time you've trained, the GPUs will be outdated and the data center will be too small]]
- [[The big AI model companies are hard to differentiate on compute, data and algorithms, so what really matters now is servers, steel and power]]
- [[Oracle is building nuclear reactors to power "acres" of GPU clusters and it will cost companies $100 billion in the next 3 years to build a frontier AI model and stay in the race]]
- [[He and Elon Musk went to dinner with Jensen Huang and begged him for GPUs and to take more of their money, because there will be only one winner in the race for AI]]
- [[A trillion dollars is being spent on data centers to enable the next, biggest wave of AI to revolutionize business productivity]]
- [[Their next AI model will be an order of magnitude bigger than GPT-4 and future models will grow at a similar rate, requiring capital-intensive investment to meet their "really big aspirations."]]
- [[AI investment will soon exceed the US defense budget and the most devastating AI safety scenarios can be avoided if the US leads in AI, but the US government is investing 3-10x less than China]]