One of the targets of the modern foundry business is to have a wide array of manufacturing process technologies. This enables it to cater to as many customers as possible. That includes logic, embedded memory, radio frequency, analog, high voltage, long life cycle, and now silicon photonics. As part of a partnership between GlobalFoundries and PsiQuantum announced today, new proprietary manufacturing tools have been installed into GlobalFoundries' most advanced facility in Malta, New York. This will enable GF to build silicon photonics to enable what could be a 1 million+ qubit quantum computer from PsiQuantum.

The nature of process node manufacturing is varied and the research can be highly targeted in many different directions. If all you care about is peak logic frequency, then a process can be designed to facilitate that, but if you need the densest memory, or embedded non-volatile memory, or you might have high voltage silicon sensors, then the process technology needs to vary and the IP portfolio has to be broad enough to build these parts. Over the last few years, we have seen GlobalFoundries develop its customized processes, such as 22FDX and 12FDX which are cost-effective FDSOI technologies for low power embedded applications, such as RF or mid-tier IoT, and even that has optimizations for higher performance modes (at the expense of efficiency) should a customer need it.


GlobalFoundries Fab 8, Malta New York

Today’s announcement is that GlobalFoundries is expanding its manufacturing portfolio in a very different way than it has previously. In partnership with PsiQuantum, part of the Fab 8 facility in Malta and the Fab 1 facility in Dresden have new proprietary equipment specifically for the production of core quantum computer components. PsiQuantum’s goal is its Q1 system, a 1 million+ photonic qubit quantum computer, which needs critical semiconductor components to function. PsiQuantum lists single-photon sources and single-photon detectors (the silicon photonics part), and control circuitry to scale into a quantum computing solution. The Photonics will be built in Fab 8, while the control chips will be built in Fab 1.

Part of the big issue with quantum computing is controlling qubit coherency. There’s no point building a qubit if it isn’t stable, and controlling the stability of one qubit is hard, let alone a million. In order to assist with this, quantum computing applies error correction algorithms, however the results of which mean that for one logical qubit, anywhere from 10 to 100 physical qubits are needed. This scales up the problem immensely. So while a 1000 logical qubit computer might be powerful for the algorithms that can run on it, there might actually be 100,000+ physical qubits due to the requirements of coherency and control. We’re nowhere near 1000 logical qubits today, and part of that is manufacturing the right technology to build such a system. This is why we’re getting an announcement out of GlobalFoundries and PsiQuantum today.

PsiQuantum’s goal is to be in a position to assemble a final photonic qubit quantum computer by mid-decade. The partnership between the companies is also highlighted in the announcement as an example of the US focusing on global technology leadership in quantum computing, and supporting domestic supply chains with feature-rich solutions.

GlobalFoundries is currently owned by Mubadala, the UAE sovereign wealth investment fund, but is looking at a potential IPO in the US in the next 12 months or so. PsiQuantum is a 2016 private startup based in Palo Alto, with over $500m in venture capital funding, ~100 employees, and investors such as Playground Global, BlackRock, M12, and C4 Ventures.

The title photo is of one of PsiQuantum's 300mm wafers.

Related Reading

 

Comments Locked

31 Comments

View All Comments

  • Unashamed_unoriginal_username_x86 - Wednesday, May 5, 2021 - link

    No one in the quantum computing field (at least not the scientists or engineers) thinks quantum computers can take over classical computing for most applications. The point is that it has a lot of specialised applications such as cryptography and simulation where the properties of a qubit gives it abilities I don't really understand (naturally), such as square rooting the strength of encryption with Grover's algorithm. This would mean 128-bit AES would be equivalent to 64-bit, so as far as I can understand a quantum computer could be 2^31 times slower than a classical computer but still beat it in brute force decryption.
  • Unashamed_unoriginal_username_x86 - Wednesday, May 5, 2021 - link

    2^63*
  • Oxford Guy - Wednesday, May 5, 2021 - link

    Is it true that the US government bans citizens from using 512-bit encryption? If so, it seems rather droll that there is so much news coverage about attempts by governments to ban the encryption that is already available (i.e. weak easy-to-break encryption). Clever tactic to control the narrative, though.
  • Lord of the Bored - Wednesday, May 5, 2021 - link

    I believe they ban the EXPORT of 512-bit encryption, but no, US citizens are not legally prohibited from utilizing any form of encryption.
  • Oxford Guy - Sunday, May 9, 2021 - link

    Thanks for the clarification. A bit pointless to put strong encryption atop an intentionally-insecure foundation, anyway. Until hardware, networks, and consumer-use operating systems are designed to be secure rather than honeypots – all the bits in the world aren't going to make quicksand into platinum.
  • GeoffreyA - Thursday, May 6, 2021 - link

    I believe classical computers can simulate quantum ones and vice versa: both are Turing complete/equivalent. A QC cannot compute anything a classical computer can't. Seems the only difference is drastically higher speed for certain applications. At any rate, speaking ignorantly as a layman, I wonder whether QCs are just being probabilistic and nothing more.
  • GeoffreyA - Thursday, May 6, 2021 - link

    I must admit, like the child in Emperor's New Clothes, I've been sceptical of this field from the word go. But it could be a success and create applications we haven't even thought of yet. Who would have guessed that Babbage's Analytical Engine would lead to fighting dragons in Tamriel or reading the "newspaper" on a small slab? And in the realms of sheer speculation, I've got a feeling our universe is being computed on a QC!
  • Santoval - Monday, May 10, 2021 - link

    Quantum computing is not just investor bait. There are inherent limits to classical computing (particularly to conventional, transistor based computing), both in terms of computation and efficiency (see Amdahl's law, Gustafson's law, Moore's law, Koomey's law etc). After the collapse of Dennard scaling in 2005/2006 clocks froze, Koomey's law ("number of computations per joule of energy") started slowing down from a doubling every 1.57 years to a current doubling every 2.8 - 3 years, Moore's law also slowed down and continues to slow down etc

    Now, it's entirely possible that quantum computing will never take off beyond the labs and very niche uses and conventional classical computing is replaced by something different, like neuromorphic computing, non quantum optical computing, molecular computing or spintronic computing. These are still classical but fundamentally different from the status quo of the last 50+ years. They either ditch silicon and MOSFET transistors and/or Von Neumann computing entirely, and are far less constrained by the above laws.

    In the last years optical computing has largely been employed in the quantum realm so it's arguably an unlikely option for classical computing - unless the low density problem of optics is resolved. Molecular computing is still kind of sci-fi. Still, neuromorphic computing is already a thing (as a co-processor) and we already have spintronics based memory (STT-MRAM).

    Nevertheless, I still think that quantum computing is the future. The question is if it will always remain in the future -like net energy from nuclear fusion- or if it can be mastered in some present..
  • Pinn - Wednesday, May 5, 2021 - link

    Quantum computing will be good for certain applications. I hear they managed to count to 4 or so. Ego is the problem, anyway.
  • Oxford Guy - Wednesday, May 5, 2021 - link

    How many lights are there? (You knew that was coming.)

Log in

Don't have an account? Sign up now