That's why Bitcoin supporters disagree with raising block size as main scaling solution and instead use 2nd layer (such as LN), optimize the protocol/transaction/scripting and optimize the client used for full nodes.
1 block is added to the chain approximately every 10 minutes. That means at most 6MB per hour, 144 MB per day or ~52.5 GB per year. Where do you see an exponential growth here? Even if we consider Segwit, or even if the block size is further increased in the future, that's still a linear growth. I'm not sure how you came up with petabytes, with current growth we'll reach 1 terabyte in about 20 years, that means 20K years you reach 1 petabyte.
Now, even if your calculations were somehow correct, I'm not sure why that would result in centralization. Bandwidth is a much tighter limit than storage. Reaching those numbers in the near future would require a huge bandwidth. If such bandwidth would be maintainable by enough people then surely storage would be cheap enough and most people would still be able to run full nodes.
If we want see Bitcoin become popular/widely usage, obviously we need to increase the block weight even considering almost everyone use 2nd layer protocol and that could risk decentralization.