Interview with Dr. Danica Marsden, Principal Quantum Computing Scientist at the Bank of Canada

Quantum computing seems to be the next big thing, Qubits instead of bits will shape the most complex payment processes. You recently co-published a research study demonstrating the effectiveness of such a novel algorithm in terms of liquidity savings for settlement in the Canadian RTGS system. How is the reduction in opportunity cost calculated?

Improving the efficiency of payment systems that need to settle on a gross basis, such as at central banks, requires optimization routines that can’t currently be run in the real system time. Since future fully-fledged quantum computers are believed to excel at optimization problems, we developed an algorithm to tackle this problem. Running our simulation on currently available prototypes, we found a significant reduction in liquidity needed by the system to settle a fixed value of payments – about $240 million dollars per day. The savings can be calculated from the opportunity cost of the collateral that is posted to the central bank in exchange for that liquidity. Since the eligible collateral is typically government securities, which in repo markets trades at a small premium to other high-quality collateral, we estimate that the savings are between five and ten basis points per dollar of liquidity saved, or 240,000 dollars per day – divided among system participants in proportion to the value of their transactions.

The time to run your algorithm is five seconds before the transactions are executed. Is this algorithm also suitable for instant payments? 
Currently, many RTGS systems use liquidity savings mechanisms which are based on rules of thumb, such as bypassing larger payments if there is insufficient liquidity. Since they don’t check all possible payment orders, this results in suboptimal solutions. Delay costs are often increased, too, by an order of minutes to hours. Our method evaluates every possible reordering for a batch of payments and finds one that settles every payment, while also minimizing the liquidity cost. This takes about 90 seconds – the same amount of time to accumulate the batch of payments in the system – of which about five are spent running on the quantum annealer. This type of pre-processor, quantum or otherwise, is not suitable for payments that need to be carried out instantly, such as retail payments, but is appropriate when the benefit of the efficiency gain significantly outweighs the delay cost.

What other use cases could see big changes in the quantum age? 
Wherever similar transaction settlement optimization problems arise, such as in finding netting sets, securities settlement, trade matching, foreign exchange settlement, arbitrage, etc. In addition, quantum computing combined with machine learning could be used for real-time monitoring to identify patterns and detect anomalies in payments.