Back during the dot-com bubble, telecom giants were convinced that "traffic demand would grow exponentially," and they went all out laying down submarine cables and backbone networks. But due to the tech limitations at the time, the downstream applications were only web browsing and low-res images, so they couldn't even use that massive bandwidth.
If AI moves past the hype phase without a solid Product Market Fit (PMF), the bubble could burst, and the turning point might be in Q4, coinciding with Nvidia's new generation of inference rolling out, which could drop AI token inference costs by 90%. In the end, we might find that there's no demand, and the whole thing could be proven wrong.
If AI moves past the hype phase without a solid Product Market Fit (PMF), the bubble could burst, and the turning point might be in Q4, coinciding with Nvidia's new generation of inference rolling out, which could drop AI token inference costs by 90%. In the end, we might find that there's no demand, and the whole thing could be proven wrong.