IonQ and Classiq say counting T gates no longer captures quantum costs

[DodgyDepot]

The way we estimate the cost of future quantum computers may need rethinking. Quantum Computing Report covers a joint brief from IonQ and Classiq titled From Magic States to Full Stack Accounting

For years, non Clifford gates such as T gates were assumed to make up over 90 percent of a fault tolerant machine's physical cost. Magic state cultivation has cut that cost sharply

The brief argues that routing, layout and idle memory now account for roughly 50 to over 90 percent of the cost. In a Classiq benchmark using Grover search, cutting T states by 3 percent raised the overall physical cost by 41 percent, and magic state production made up only 6 to 26 percent of the total

An IonQ estimate for a 256 bit elliptic curve discrete log problem found that 39 million Toffoli gates would take 13.3 days, but total runtime would be 25.7 days. These are company figures from one algorithm and one fault tolerance model

It is a useful reminder that optimising one number can make the overall system worse. As error correction improves, the real bottlenecks move to other parts of the machine. Full stack thinking should lead to better designs and more realistic resource estimates

Should the industry change how it measures quantum algorithm costs? And which part of the stack do you think will be the hardest to scale?

IonQ and Classiq Outline Architectural Trade-Offs in Fault-Tolerant Resource Estimation - Quantum Computing Reportquantumcomputingreport.com