Quantum computing startup says it will leapfrog everybody - Ars Technica
Skip to content
AI
Biz & IT
Cars
Culture
Gaming
Health
Policy
Science
Security
Space
Tech
Forum
Subscribe
Story text
Size
Small<br>Standard<br>Large
Width
Standard<br>Wide
Links
Standard<br>Orange
* Subscribers only
Learn more
Pin to story
Theme
Search
Sign In
Sign in dialog...
Text<br>settings
Story text
Size
Small<br>Standard<br>Large
Width
Standard<br>Wide
Links
Standard<br>Orange
* Subscribers only
Learn more
Minimize to nav
A short time back, we covered an announcement by Amazon that it would be hosting a useful quantum computer from its partner QuEra as soon as sometime in 2028. The system promised some eye-popping numbers compared to anything on the market today: over 10,000 individual qubits, each with an error rate low enough that the system could support hundreds of error-corrected logical qubits. But QuEra has to get there from its current hardware, which sits at 260 qubits that are relatively error-prone.
Those details about how it was going to get there were left for last Wednesday, when QuEra announced its roadmap. But the announcement only accentuated the gap: There will be no new hardware releases between now and the useful machine, and QuEra is promising to deliver an even more powerful machine just one year later.
“The company made a strategic decision not to sell NISQ [noisy intermediate scale quantum] systems anymore,” QuEra’s Yuval Borger told Ars. The two systems it had previously made available have similar capabilities, with about 250 hardware qubits and an appreciable error rate—enough to test some error correction codes, but not sufficient for using logical qubits in applications.
The machine it proposes to build for Amazon will be a very different beast. It will have over 10,000 physical qubits, which QuEra intends to use to create 256 error corrected logical qubits. With error correction active, these will experience 99.9999 percent error-free operations, which the company expects will enable millions of operations to be performed successfully. And there’s a follow on. By 2029, QuEra promises a successor with twice as many hardware qubits, capable of supporting over 1,000 logical qubits. Error resistance will rise to 99.9999999 percent.
QuEra’s machines operate using neutral atoms that are held in a grid by lasers, so raising the qubit count is largely a matter of boosting the laser capacity. And the two academic labs that launched QuEra and licensed IP to it have already demonstrated a 3,000-qubit system. Scaling that will be a challenge, but there’s an obvious route to it. The labs have also demonstrated the ability to replace atoms lost during operations, an ability that’s critical to keeping these machines running.
QuEra’s existing hardware (left) and the new machines (right) promised for the next few years. That new hardware will require very rapid progress. “Quop” stands for quantum operations.
Credit:<br>QuEra
QuEra’s existing hardware (left) and the new machines (right) promised for the next few years. That new hardware will require very rapid progress. “Quop” stands for quantum operations.
Credit:
QuEra
The bigger challenge QuEra faces will be that these future machines require getting the error rates of the hardware qubits down dramatically. As a rough approximation, the error rates of logical qubits depend on a combination of how many hardware qubits are used to create the logical ones, and the error rate of the hardware qubits. If you get the hardware error rate down, you can get similar error correction performance out of a logical qubit built with fewer hardware ones. (Some of the more efficient error correction schemes use fewer hardware qubits, but there are limits to how far this can be pushed.)
In its 2028 machine, QuEra will be using an error correction code that requires 40 hardware qubits for each logical qubit. One year later, it intends to cut the hardware qubits per logical qubit to 20. Cutting it to that number will require the hardware error rate to go down just to keep the same level of performance, except QuEra is also promising dramatically better performance, which means the hardware error rate will drop even further.
It’s unclear where that sort of drop is going to come from, especially given it’s only expected to take one year. When we asked QuEra about the biggest challenges remaining, the error rate wasn’t mentioned specifically. “A lot of the challenges between now and a production system on the cloud are classical challenges, not quantum challenges: control electronics, real-time quantum error correction, building compilers that allow users to easily tap into the power of the system,” Borger told Ars.
Although the company has listed the current error rates for its hardware (which are very good, but not presently good enough for this), it is not specifying the level of performance that will be needed to achieve its planned goals. The announcement is...