Equal1 announced a new quantum processing chip that highlighted the operation at 3.7 degrees kelvin, but we think everyone missed some of the most important attributes of the chip. We talked with Jason Lynch, COO of Equal1 to find out more about the chip.
First, we will discuss the temperature aspects, Equal1 has implemented a strategy similar the Intel’s approach of developing cryogenic qubit control logic that runs at 3.7 degrees kelvin to eliminate the long run of coaxial cables from room temperature electronics down to the low temperature qubits. But, in some respects they have gone two steps beyond Intel’s current status. First, Equal1 has integrated both the qubits and the control logic on the same piece of silicon in a design that has over 10 million transistors. Intel has indicated that this is a future goal, but currently the qubits and the control logic are on separate chips. And second, because of this the Equal1 qubits run at an operating temperature of 3.7 kelvin while Intel’s qubits still require 0.020 kelvin operating temperature. While the difference may not seem too be much, it actually makes a world of difference from a cooling standpoint. Millikelvin operating temperatures requires using dilution refrigerators which cost hundreds of thousands of dollars. A 3.7 kelvin temperature can be cooled with liquid helium at a cost of perhaps a few 10’s of thousand of dollars. Besides the very large cost savings, eliminating the dilution refrigerator would makes it easier to provide an on-premise solution to those customers who prefer to have the machine in-house than access it remotely over the cloud.
One caveat about this comparison is that we do not know how the gate fidelities and other quality metrics compare. Although integrating the qubits and eliminating the dilution refrigerator sounds good, if there is a negative impact on the qubit quality, it might not make these moves worthwhile. Nonetheless, we expect Intel’s future roadmap is to also go in the same direction while being careful to maintain or increase the qubit quality levels.
Another characteristic of the Equal1 chip that surprised us was the gate delay. We were told that it currently is in the 1-3 nanosecond range and that Equal1 has a goal of reducing this to 100’s of picoseconds. For comparison, gate delays in the current generation of superconducting machines are orders of magnitude slower and can range from the ten’s to hundreds of nanoseconds. And gate delays in ion trap machines are slower still and are in the microsecond to millisecond range. Equal1 currently is reporting a coherence time of 150 nanoseconds which is quite short, but the key metric that one should look at is the coherence time divided by the gate delay. So the fast gate times can help make up for the short coherence times.
The current chip, codenamed Alice, is still meant for internal testing of Equal1’s technology. Equal1’s near term focus is to develop quantum chips optimized for use in AI applications and use Tensorflow for programming them. Over the longer term, they will also develop more general purpose versions that will use Qiskit for programming. But AI applications are currently have the larger focus at the company. Besides potential advantages in performance, a quantum approach can provide significant advantages in power consumption over classical AI approaches. Another advantage is that Equal1’s chips are being fabricated at GlobalFoundries using a standard semiconductor 22 nanometer manufacturing process without any special modifications. This could allow them to quickly ramp up production once they have a design ready for production.
Additional information about this quantum processor chip can be found in a news release located on the Equal1 website here.