Because in Ethereum you don't waste hashpower since stales are also rewarded.
A stale block is called uncle or ommer block in Ethereum and you might want to read about the reward scheme here: What is GHOST and what is its relationship to Frontier and Casper?
And here: What is an uncle/ommer block?
To supplement this with Vitalik's answer on Gitter (quoted):
- Rewarding ommer blocks means miners have lesser incentive to mine on the latest block:
basically means that miners' private incentive to make relaying super-fast is ~3-4x lower than it normally would be this is by design, to limit centralization concerns
- Ethereum is computation-heavy. IO accesses and state Merkle root computation are more of a bottleneck than network latency. In Bitcoin, the transaction Merkle root can be precomputed (before getting the hash of the previous block).
ethereum in practice is a bit more computation-heavy than bitcoin is, though for reasons more subtle than "OMG it's turing complete!!1!", so bandwidth improvements won't help quite as much the actual reasons are (i) you can't pre-compute as much before receiving the block as it's more order-dependent, (ii) there's overhead from Merkle state root computations, (iii) you can't pre-compute IO accesses as those may be dynamic