Martin Splitt mentioned on a LinkedIn post[1] as a follow up to this that larger sites may have crawl budget applied.
> That was a pretty defensive stance in 2018 and, to be fair, using server-side rendering still likely gives you a more robust and faster-for-users setup than CSR, but in general our queue times are significantly lower than people assumed and crawl budget only applies to very large (think 1 million pages or more) sites and matter mostly to those, who have large quantities of content they need updated and crawled very frequently (think hourly tops).
We have also tested smaller websites and found that Google consistently renders them all. What was very surprising about this research is how fast the render occured after crawling the webpage.
I couldn't share any pricing data since the discussions with providers are private. Instead, I added a graph of prices from gpulist.ai. For an Infiniband cluster, median is $2.3 per H100 hour, average is $2.47.
I'm surprised that ~$10 million dollars of GPUs, @ $40k per H100 and excluding operational costs like the energy bill, only rents for $455k per month. Sounds like a really tough business since the amount of time required to recoup the costs of ownership (~21 months) seems like a really long time. A new generation or two of chips will have hit the market in that time, depreciating the recurring rental income. Leads me to wonder how much if any profit can be made renting GPUs.
I made a fork of the Chrome DevTools that adds "Copy as Python" to the right click menu of each request in the Network tab. You can tell Chrome to use a different version of the DevTools if you start it from the command line
Eliot here, I work with Vincent on PhotoRoom. Our current focus is shipping an incredible app on the web and on mobile, but we'll tackle video very soon!