In a comprehensive Investor Day presentation, Rigetti Computing described a detailed product roadmap that covers all aspects of their activities and announced new and enhanced partnerships that will help them achieve their goal of quantum advantage. Here are the details:
Availability on Microsoft Azure
In February of this year, Rigetti announced they were planning to add Microsoft Azure as one of their cloud partners for their 40 qubit Aspen-11 and their 80 qubit Aspen-M-2 processors and this week they announced that the machines were now available in a public preview on Azure. This now makes these machine available via Amazon Braket, Microsoft Azure, Rigetti’s own QCS, Strangeworks, and the Oak Ridge National Laboratory (ORNL).
QPU Roadmap
Rigetti had previously provided a brief description of their QPU roadmap in May of this year, but they have now provided some additional information about it. In terms of qubit count, their next processor will sport 84 qubits in a single chip, codenamed Ankaa, with an expected availability in early 2023. This chip will use their fourth generation architecture that will use tunable couplers and a square lattice topology that will provide significant improvements in gate fidelity and qubit connectivity over their third generation QPUs. The tunable couplers will allow better isolation between neighboring qubits and will also enable them to improve the gate delays in the processors. The tunable coupler was first demonstrated by Google in their Sycamore processor. The company indicated it was targeting gate fidelities of 99.5% or better with this new processor, a significant improvement. The new square lattice architecture will improve the average qubit connectivity and allow each qubit to interact with 4 neighboring quibts, an improvement from the 2 or 3 that is in the current 3rd generation architecture. This will make it easier to embed a particular quantum program in the processor and reduce the number of SWAP gates needed to move the qubit information around. It may also make it easier to implement various error correction codes, like the surface code which work better with a square lattice topology.
After the introduction of the 84 qubit processor, Rigetti plans to introduce a 336 qubit processor, codenamed Lyra, later in 2023. The Lyra will consist of four of the Ankaa die which will be assembled on a carrier substrate. This technique will leverage key technology that Rigetti has been working on for a long time that allows entangled qubit information to be exchange between die with no loss in speed or fidelity versus connections made between different qubits on the same die. Following the 336 qubit Lyra introduction, they are planning a 1000+ QPU in roughly the 2025 timeframe and a 4000+ QPU in roughly the 2027 timeframe. Interestingly, those two QPU’s will still consist of multiple 84 qubit die, but they will assemble more die, perhaps dozens of them, on the carrier substrate and also utilize multiple fridges for the 4000+ QPU to achieve the higher number of qubits. Although Rigetti isn’t expected to increase the number of qubits on the Ankaa die during this period, it is likely they will continue refining it with subsequent revisions to provide continuous improvement in the gate fidelities, coherence times, and other parameters. They will also need to implement larger dilution refrigerators (covered below), improved flex I/O, and a next generation control system for these higher capacities.
Fab-1 Expansion
Rigetti has a fabrication strategy that is much different from most of their other quantum startup competitors. From their early days they decided to obtain their own wafer fab to build the chips rather than utilize a university clean room or try to work with a large semiconductor foundry like GlobalFoundries. To do this they opened up their own fabrication facility in Fremont, called Fab-1, in May of 2017. Part of their rationale was that they wanted to achieve as fast of a processing cycle time as possible in order to allow their engineers to quickly try out and iterate improved designs and felt that best way of doing this was to own the fab. Using their fab, they are able to build designs in 5-15 weeks versus an estimated 22-40+ weeks if they were using an external semiconductor foundry. More recently, Rigetti has not only been fabricating devices for themselves, but they are also fabricating devices for Fermilab, AFRL, and Purdue University. Their primary purpose for this is not to gain a load of new revenue by serving as a foundry, but rather obtain engineering feedback from their organizations that will help their own processing and processor development. Rigetti is the lead commercial partnership for the SQMS (Superconducting Quantum Materials and Systems Center) center being led by Fermilab.
In order to increase the capacity of Fab 1, Rigetti started a Fab-1 expansion project at the beginning of this year to add 5,000 square feet which will almost double its size. The company expects to have this project completed by the end of 2022 and it will allow them to process the larger volume of wafers needed to develop the larger QPUs mentioned above.
Partnership with Bluefors
The larger multichip processors will require bigger dilution refrigerators with more cooling power and a modular design that can facilitate multi-fridge implementations. To achieve this, Rigetti has entered into a partnership wtih Bluefors to develop a new refrigerator design called KIDE. The company expects to receive the first prototype of KIDE in early 2023 with subsequent deliveries in late 2023 to support the Lyra 336 qubit multi-chip design.
Partnership with Keysight for Error Mitigation Tools
An important requirement for achieving Quantum Advantage is to provide processors that have the lowest possible error rates. Although a lot of this has to do with the hardware architecture and implementation at the chip level and control electronic circuitry, software can play a big role too. In May of 2021, Keysight Technologies acquired a Canadian startup called Quantum Benchmark that had developed error mitigation software, which they call True-Q. As part of the investor day announcement, Rigetti announced they will be integrating the True-Q software into their QCS platform and they will provide their users with this error mitigation capability as a built-in function.
Partnerships with Ampere and NVIDIA
Rigetti has also been an early proponent of hybrid classical/quantum processing to provide effective solutions for difficult computation problems. In February of this year, they announced a partnership with classical microprocessor startup Ampere to integrate Ampere’s extremely high performance chips into QCS. These Ampere processors should provide much higher performance for the classical portion of the algorithms than any other company using more traditional microprocessor chips and provide an advantage to Rigetti for end users working with their QCS because it should speed up overall runtime.
Rigetti has now started another program with NVIDIA to develop a hybrid GPU-QPU workflow for climate modeling applications. The goal will be to apply quantum machine learning techniques in a high-performance hybrid workflow and build upon recent weather modeling application research by Rigetti.
For more on Rigetti’s announcements, you can access their press release about their expanded QCS availability on Microsoft Azure here. Their press release describing several of their announcement at their Investory Day is available here. And you can view a video of their 3 hour Investor Day presentations here.
September 16, 2022