If their supercomputer was so great, there was very little probability in losing.
Garry ‘s Elo rating was in the 2800s. Deep Blue II’s was comparable or maybe slightly higher. There was no guarantee that Deep Blue would consistently win against Garry and there was even a real chance it could lose if Garry figured out how Deep Blue calculated its moves. That was the other risk. Today even a laptop running freely available Stockfish is rated around 3500 that no human can stand up against. It wasn’t like that back then. IBM was good but not significantly better than the best grandmasters.