The gamble in this case is pocket change. They could comfortably spend this amount every year and just use precious years for overnight / batch processing jobs when they become less useful for dev work but remain operational.