Even Sequoia Capital thinks we might be in an AI bubble. A recent blog post from David Cahn on the Sequoia Capital blog has everyone in venture capital talking about bubbles, again. How to read this chart? The math in it is relatively simple. → First, this chart doubles Nvidia's run-rate revenue, which is mostly GPUs, to account for the total AI data center costs, which must also include energy and infrastructure. → Then, double that amount again. This is to account for a 50% gross margin for startups or businesses that use AI services from companies like Amazon or Microsoft (who are massive buyers of Nvidia chips). You can see the final number is the much talked about $600B in revenue that this analysis says is required for the industry to earn from AI apps to account for the massive spending on chips and data centers. Can the industry earn that much? How many years will it take? Will there be a huge breakthrough in AI robots that quickly uses up all the capacity? All great questions that are very hard to answer, though I am seeing new meaningful use cases everyday from startups that are solving real world problems using AI (everything from training robots, optimizing trucking, self-driving cars, asking your mailbox questions). You can read the full post here: https://lnkd.in/g7p736Rk
Can it be a bubble if everyone thinks it is a bubble?
Co-Founder at Sweetspot (YC S23) | AI for Government Procurement
1moIt's way too early to say we're in a bubble. 80% of the money raised by AI labs are for compute and those data centers are still under construction. Current models are being trained with last generation hardware. I wouldn't make a direct conclusion until those are out...