Call for Intel to Focus on Components for Others to Easily Build Their Own Quantum Computers

Rather than see Intel build its own quantum computers and compete with the other quantum computing vendors, imagine what it would be like to see Intel focus on designing, producing, and selling components which enable others to easily design and build their own quantum computers. Intel could do for quantum computing what they did for personal computing and servers — focus on the key circuitry — processors and supporting chipsets, and some other tricky components for qubits and their control and readout. Enable other vendors to easily design and build quantum computers. This informal paper will imagine some of the possibilities.

• To be clear, the proposal in this informal paper is purely speculative on my part — Intel has not suggested any such intentions and I have no specific knowledge of their intentions other than their own statements in the cited press releases.

Topics covered in this paper:

1. In a nutshell
2. Goals and benefits
3. Areas of focus for Intel in quantum computing
4. Areas for Intel to stay away from or at least keep at arm’s length with outside partners
5. Other areas for Intel to focus on
6. Intel should be agnostic relative to all quantum computing tools, frameworks, and SDKs, such as Qiskit and Cirq
7. Intel’s Horse Ridge cryogenic qubit control chip inspired this proposal
8. Intel milestones in quantum computing
9. q86 architecture for quantum computer systems
10. q86 is a name I contrived, to parallel x86
11. Quantum computer as a coprocessor
12. Tight integration of q86 with x86
13. q86-compatible quantum computer systems
14. q86 architecture — a big question mark at present
15. Publish fully documented reference designs
16. Publish Principles of Operation and Implementation Specifications
17. Produce limited machines as reference platforms
18. Test bed (or testbed)
19. Original Equipment Manufacturers (OEMs)
20. Not all q86 customers would be OEMs
21. A rich and thriving ecosystem
22. Distributors
23. Value-added resellers (VARs)
24. Turn-key solution providers
25. Open source is essential
26. Facilitate academic research in quantum computing hardware
27. Academic consortium for customized designs for quantum computers for academic institutions
28. Contract engineering and manufacturing firms
29. Analog components and RF components
30. Timing
31. Historical timing
32. How much might a quantum computer system cost?
33. Pricing for service — leasing and usage
34. Service level agreement (SLA)
35. Access: Experimental, testing, and production
36. Siting
37. Government labs might wish to build their own quantum computer systems
38. Architectures tailored for high-end quantum simulators
39. Subsystems
40. Co-design of dilution refrigerator, wiring, and electronics to match each other?
41. Industry standards — potential but premature
42. Might AMD or NVIDIA beat Intel to the punch?
43. Intel’s key advantages
44. Other potential competitors
45. Might some scrappy startup eat Intel’s lunch?
46. Upgrade from FPGA to full-custom logic to boost performance and capacity and reduce cost
47. Universal quantum computer — hybrid quantum/classical integrated processor with QRAM
48. QRAM would offer dramatic improvement in SPAM — State Preparation And Measurement
49. Let Intel handle all of your intellectual property (IP) issues
50. Intel venture capital investments to spur progress in the quantum computing ecosystem
51. Academic research grants
52. What needs to happen next?
53. My original proposal for this topic
54. When might this proposal come to fruition? Six months to three years, or longer
55. Alternative titles
56. Summary and conclusions

In a nutshell

1. To be clear, the proposal in this informal paper is purely speculative on my part — Intel has not suggested any such intentions.
2. Intel could do for quantum computing what they did for personal computing and servers. Focus on the key circuitry — processors and supporting chipsets — to enable other vendors to easily design and build quantum computers.
3. Traditional computer hardware systems vendors. Also known as OEMs (Original Equipment Manufacturers.) Such as HP and Dell. Or even IBM, to diversify and address commodity and entry level markets. IBM could do for quantum computing what they did for… the IBM PC.
4. Non-traditional hardware vendors. Amazon. Oracle. Accenture.
5. Larger customers could develop or contract for their own custom systems. Big car companies. Aerospace firms. Big drug firms. Big financial firms.
6. Contract engineering and manufacturing firms. Develop quantum computers to specs of larger and more sophisticated customers. Or even medium-sized customers. Using Intel components.
7. Opportunities for distributors. Most customers would buy components through distributors rather than directly from Intel.
8. Value-added resellers (VARs). Adding applications and packaged solutions to basic quantum computing hardware.
9. Turn-key solution providers. A step beyond VARs, providing everything that the user needs to use the system. Everything is in place. Nothing more to add. Plug it in, turn it on, and away you go.
10. Make it easier for academic researchers to build their own research systems. Enable them to focus on their research, not all of the supporting hardware.
11. Academic consortium for customized designs for quantum computers for academic institutions. For those who want customized designs for quantum computer systems but not do all of the engineering and manufacturing work themselves.
12. Maybe license designs to AMD, et al to diversify vendor supply chains.
13. Carefully and fully document reference designs. Publish all of the details to produce a fully functional quantum computer using Intel components.
14. Publish Principles of Operation and Implementation Specifications. The Principles of Operation document tells a programmer everything they need to know about a computer system to write code for the system. Typically one document for a whole family of processors. The Implementation Specification is a document for each processor which tells a programmer all of the details they might want to know to code well for that particular processor which might vary between members of the processor family.
15. Produce limited machines as reference platforms. According to published reference designs. Even hardware vendors would want these for testing and reference.
16. Maybe licensing for white-label production. And maybe even for Amazon, et al.
17. Intellectual property (IP). Intel can control all of the IP they want. As well as likely have to license some IP from others. Hardware vendors would pay for the IP through component prices as they do for x86. No need to deal with the hassles and headaches of patents and patent disputes and licensing.
18. Focus on massive electronics integration. Play to Intel’s strengths. More electronics and connectivity at the lowest levels of the frig. Less complex wiring. Less communication with the outside world.
19. Modular design. Intel designs and produces the basic modules. Vendors decide how many modules to support. Unclear whether 32 or 48 or 64 or 128 qubits is the sweet spot. Vendors could offer small, medium, large, and very large quantum computing systems.
20. Subsystems. It’s premature to judge whether Intel’s components for quantum computers would simply be discrete chips and chip sets or might constitute entire subsystems — circuits boards with more than a few chips or even collections of circuit boards. That said, it just seems inevitable that some degree of subsystems would be likely. The only point here is as a placeholder to allow for the possibility of subsystems.
21. Co-design of dilution refrigerator, wiring, and electronics to match each other? At present, the dilution refrigerator is a distinct component from another vendor, but conceptually Intel could design all of the components to work better together.
22. Open source is essential. Everything should be open source. All of the software for sure. Generally firmware as well. Any diagnostics, configuration tools, and support software. And operating systems. And even hardware designs when possible. Facilitate customization and extension by customers and academic and government researchers. Make the Intel architecture the place where quantum computing innovation occurs.
23. Call it the q86 architecture to parallel the x86.
24. Quantum computer as a coprocessor. Generally speaking, a quantum computer is not a complete computer system capable of running full applications as a classical computer system does, but to function as a coprocessor or adjunct processor to the classical processor, running quantum algorithms on behalf of the quantum application running on a classical processor.
25. A rich and thriving ecosystem. Many vendors, partners, and suppliers each doing their niche part in a much larger ecosystem. Just as we have today with the x86 architecture.
26. Also look at tighter integration with x86.
27. Maybe analog and RF components as well. Reduce uncertainty and complexity due to component mismatches. If it makes life substantially easier for customers, do it.
28. Maybe even wiring and connectors. They are so critical and so difficult to do quality control successfully.
29. Maybe everything but the refrigerator and box it all goes in. At least all of the critical electronics.
30. Support for high-end simulation as well. Classical simulation of quantum circuits is still needed since real quantum hardware doesn’t support debugging. Much higher performance and capacity is needed than is found on typical commodity servers.
31. Upgrade from FPGA to full-custom logic to boost performance and capacity and reduce cost.
32. Eventually a universal hybrid classical/quantum processor with QRAM. But that’s beyond the scope of this paper.
33. QRAM would offer dramatic improvement in SPAM — State Preparation And Measurement. Handle larger amounts of input and output data and do it more efficiently.
34. Consider diversification to other qubit technologies. Eventually, but not required for the initial thrust of this proposal.
35. Support for industry standards. Hardware standards as well as software standards. Enable easy portability of quantum algorithms and quantum applications between hardware vendors.
36. Validation. Test to confirm adherence to standards and compatibility.
37. Benchmarking. Validate and document performance and capacity.
38. Leverage quality control across vendors. Quality control of complex electronics is difficult to get right. Each vendor wouldn’t have to replicate the work of Intel.
39. Someone else could do this. It doesn’t have to be Intel. Somebody else could be the new Intel, the Intel of quantum computing. AMD or NVIDIA might smell the blood in the water and decide to eat Intel’s lunch. Or some scrappy startup — who understands that focusing on the high-end Cadillac segment of the market misses the bigger picture, and that the real money is to be made with the low-end and mid-range commodity and premium-commodity mass markets.
40. Some scrappy startup could eat Intel’s lunch. Maybe less likely, but still possible. And certainly an exciting prospect.
41. Siting. Beyond the scope of this paper. In-house on-premise vs. vendor or cloud provider remote access.
42. Timing. Maybe this proposal doesn’t take root with the current generation of qubit hardware. May have to wait for a future generation of hardware, possibly more amenable to commoditization. Current focus is pre-commercialization — research, prototyping, and experimentation — may take quite some time until hardware is ready for commercialization. And it would all seem pointless until Intel has fully-functional hardware ready for mass production and delivery to customers.
43. Not the purpose of this paper to detail all aspects of Intel’s quantum computing work. Focus is on the big picture.
44. Doesn’t matter what their current offerings look like. Focus is on future potential.
45. Horse Ridge inspired this proposal. Reading the announcement of Intel’s Horse Ridge cryogenic control chip inspired the proposal espoused in this paper. It’s exactly the kind of complexity which Intel can master, and which can dramatically simplify the efforts of others to design and build their own quantum computer systems.
46. Intel venture capital investments to spur progress in the quantum computing ecosystem.
47. Academic research grants. Continue and accelerate to fuel future growth.
48. What needs to happen next? Just lots of research by Intel. Let’s see where they are in two years or so.
49. When might this proposal come to fruition? Six months to three years, or longer.

Goals and benefits

• Massive compatibility of algorithms and applications from a wide range of vendors. Ala the x86 architecture.
• Amortize research costs over a larger number of systems.
• Amortize capital investment over a larger number of systems.
• Dramatic increase in availability of quantum computing hardware.
• Facilitate entry of new vendors.
• Supply chain diversification.
• Ready and rapid supply of hardware for academic research.

Areas of focus for Intel in quantum computing

Short term:

1. Quantum computing processor components.
2. Include analog and RF components along with digital components.
3. Maybe even wiring and connectors. They are so critical and so difficult to do quality control successfully.
4. Architectures tailored for high-end quantum simulators.
5. Integration and testing of all components.

Medium term:

1. Modular and networked processor architectures.

Longer term:

1. Universal hybrid quantum/classical integrated processor with QRAM.

Areas for Intel to stay away from or at least keep at arm’s length with outside partners

These are all areas that are very tempting and Intel could handle them but they would be significant distractions. Intel would benefit more greatly by encouraging a software ecosystem and focus on working with outside partners (third parties) in these areas:

1. Operating systems.
2. Infrastructure software.
3. Support software.
4. Beyond basic minimal system management software.
5. Software tools.
6. Beyond basic minimal software development kit (SDK). Expect SDKs such as Qiskit and Cirq to be provided by third parties.
7. Compilers.
8. Component libraries.
9. Programming models.
10. Applications.
11. Application frameworks.
12. Quantum solutions.

Hopefully many of these areas will be fully covered by third parties — in open source.

Other areas for Intel to focus on

Although it would be best for Intel to stay away from software in general, some areas where they can deliver value would be:

1. Reference configuration for overall system control. Customers can add significant value, but basic system functions should be available right out of the box. But it would still be best for Intel to partner and outsource for this. And it should be open source — Intel could start it or partner for it and then turn it into open source.
2. Encouraging and developing software industry standards.
3. Firmware. Such as qubit control. FPGA firmware and the low level software for implementing unitary transforms and qubit control.
4. Calibration and lower-level hardware and firmware configuration.
5. Hardware diagnostics.
6. Basic minimal system management software. Just the bare, minimal basics. Including an API for remote network access.
7. Basic minimal software development kit (SDK). Just the bare, minimal basics. Nothing comparable to Qiskit or Cirq.

And it should all be open source.

Intel should be agnostic relative to all quantum computing tools, frameworks, and SDKs, such as Qiskit and Cirq

Intel should focus on only the most basic and minimal tools, frameworks, and SDKs. Just enough for a quantum computing system built from Intel components to be barely functional.

Customers, partners, and third parties can supply any and all tools, frameworks, and SDKs, such as Qiskit and Cirq.

Intel’s Horse Ridge cryogenic qubit control chip inspired this proposal

Reading the various announcements about Intel’s Horse Ridge cryogenic control chip inspired the proposal espoused in this paper.

It’s exactly the kind of complexity which Intel can master, and which can dramatically simplify the efforts of others to design and build their own quantum computer systems.

Reading the original press release planted the seed, although I didn’t fully realize it at the time. It was just a vague possibility.

But while reading the press release for the second generation of Horse Ridge it really sunk in as a firm idea with great potential — Intel as the key component provider for those who wish to design and build their own quantum computers.

Granted, there’s a lot more to this proposal than a single control chip, but that’s the seed that was planted which grew into this proposal.

Intel’s announcements…

2019 announcement of Horse Ridge chip for cryogenic qubit control:

• Intel Introduces ‘Horse Ridge’ to Enable Commercially Viable Quantum Computers
• December 9, 2019
• https://newsroom.intel.com/news/intel-introduces-horse-ridge-enable-commercially-viable-quantum-computers/

2020 announcement of second generation of Horse Ridge chip for cryogenic qubit control:

• Intel Debuts 2nd-Gen Horse Ridge Cryogenic Quantum Control Chip
• December 3, 2020
• https://www.intel.com/content/www/us/en/newsroom/news/2nd-gen-horse-ridge-cryogenic-quantum-control-chip.html

2021 announcement of QuTech using Horse Ridge for cryogenic qubit control:

• Intel and QuTech Demonstrate Advances in Solving Quantum Interconnect Bottlenecks
• May 12, 2021
• https://www.intc.com/news-events/press-releases/detail/1466/intel-and-qutech-demonstrate-advances-in-solving-quantum

Intel milestones in quantum computing

This is a brief summary of Intel’s history in quantum computing. It is not designed to be comprehensive, but just to give an overview.

For Intel’s overall history in quantum computing, see their quantum press room:

• Intel Newsroom Press Kit: Quantum Computing
• https://www.intel.com/content/www/us/en/newsroom/resources/press-kits-quantum-computing.html

A brief summary of some milestones with some details following the list:

1. 2015 announcement of collaboration between Intel and QuTech at TU Delft.
2. 2016 press on Intel in quantum computing in Technology Review.
3. 2017 announcement of 17-qubit quantum computer chip.
4. 2018 announcement of 49-qubit quantum computer chip.
5. 2019 announcement of Horse Ridge chip for cryogenic qubit control.
6. 2020 announcement of second generation of Horse Ridge chip for cryogenic qubit control.
7. 2021 announcement of QuTech using Horse Ridge for cryogenic qubit control.
8. March 2022 announcement that Intel will present 14 quantum computing papers at the APS March 2022 Meeting.
9. April 2022 announcement that Intel will deliver its first quantum computing test bed to Argonne National Laboratory this year.

2015 announcement of collaboration between Intel and QuTech at TU Delft:

• QuTech quantum institute enters into collaboration with Intel
• American chip manufacturer Intel and QuTech, the quantum institute of TU Delft and TNO, have finalised plans for a ten-year intensive collaboration. Alongside financial support for QuTech totalling approximately $50 million, Intel will also contribute expertise, manpower and facilities. “QuTech is delighted to welcome Intel as their new partner. The major challenge facing quantum technology development in the coming decades, such as creating a quantum computer, is set to be upscaling: being able to create complex structures with an enormous number of quantum bits. This partnership will enable us to combine our scientific expertise with the best engineering expertise in the computer industry”, says lead scientist Lieven Vandersypen of QuTech.
• September 3, 2015
• https://qutech.nl/2015/09/03/qutechentersintocollaborationwithintel/

2016 press on Intel in quantum computing in Technology Review:

• Intel Bets It Can Turn Everyday Silicon into Quantum Computing’s Wonder Material
• The world’s largest chip company sees a novel path toward computers of immense power.
• Sometimes the solution to a problem is staring you in the face all along. Chip maker Intel is betting that will be true in the race to build quantum computers — machines that should offer immense processing power by exploiting the oddities of quantum mechanics.
• Competitors IBM, Microsoft, and Google are all developing quantum components that are different from the ones crunching data in today’s computers. But Intel is trying to adapt the workhorse of existing computers, the silicon transistor, for the task.
• Intel has a team of quantum hardware engineers in Portland, Oregon, who collaborate with researchers in the Netherlands, at TU Delft’s QuTech quantum research institute, under a $50 million grant established last year. Earlier this month Intel’s group reported that they can now layer the ultra-pure silicon needed for a quantum computer onto the standard wafers used in chip factories.
• December 21, 2016
• https://www.technologyreview.com/2016/12/21/243959/intel-bets-it-can-turn-everyday-silicon-into-quantum-computings-wonder-material/

2017 announcement of 17-qubit quantum computer chip:

• Intel Delivers 17-Qubit Superconducting Chip with Advanced Packaging to QuTech
• Today, Intel announced the delivery of a 17-qubit superconducting test chip for quantum computing to QuTech, Intel’s quantum research partner in the Netherlands. The new chip was fabricated by Intel and features a unique design to achieve improved yield and performance.
• The delivery of this chip demonstrates the fast progress Intel and QuTech are making in researching and developing a working quantum computing system. It also underscores the importance of material science and semiconductor manufacturing in realizing the promise of quantum computing.
• October 10, 2017
• https://newsroom.intel.com/news/intel-delivers-17-qubit-superconducting-chip-advanced-packaging-qutech/

2018 announcement of 49-qubit quantum computer chip:

• During his keynote address, Intel CEO Brian Krzanich announced the successful design, fabrication and delivery of a 49-qubit superconducting quantum test chip
• January 8, 2018
• https://newsroom.intel.com/news/intel-advances-quantum-neuromorphic-computing-research/

2019 announcement of Horse Ridge chip for cryogenic qubit control:

• Intel Introduces ‘Horse Ridge’ to Enable Commercially Viable Quantum Computers
• December 9, 2019
• https://newsroom.intel.com/news/intel-introduces-horse-ridge-enable-commercially-viable-quantum-computers/

2020 announcement of second generation of Horse Ridge chip for cryogenic qubit control:

• Intel Debuts 2nd-Gen Horse Ridge Cryogenic Quantum Control Chip
• At an Intel Labs virtual event today, Intel unveiled Horse Ridge II, its second-generation cryogenic control chip, marking another milestone in the company’s progress toward overcoming scalability, one of quantum computing’s biggest hurdles.
• December 3, 2020
• https://www.intel.com/content/www/us/en/newsroom/news/2nd-gen-horse-ridge-cryogenic-quantum-control-chip.html

2021 announcement of QuTech using Horse Ridge for cryogenic qubit control:

• Intel and QuTech Demonstrate Advances in Solving Quantum Interconnect Bottlenecks
• New research published in Nature highlights progress toward quantum scalability with high-fidelity control and programmability of silicon qubits using the Horse Ridge cryogenic controller.
• Intel and QuTech — a collaboration between Delft University of Technology and the Netherlands Organisation for Applied Scientific Research — published key findings in quantum research to address the “interconnect bottleneck” that exists between quantum chips that sit in cryogenic dilution refrigerators and the complex room-temperature electronics that control the qubits. The innovations were covered in Nature, the industry-leading science journal of peer-reviewed research, and mark an important milestone in addressing one of the biggest challenges to quantum scalability with Intel’s cryogenic controller chip Horse Ridge.
• May 12, 2021
• https://www.intc.com/news-events/press-releases/detail/1466/intel-and-qutech-demonstrate-advances-in-solving-quantum
• https://www.businesswire.com/news/home/20210512005757/en/

The Nature paper referenced in that press release:

• CMOS-based cryogenic control of silicon quantum circuits
• May 12, 2021
• https://www.nature.com/articles/s41586-021-03469-4

The arXiv preprint for that paper:

• CMOS-based cryogenic control of silicon quantum circuits
• September 29, 2020
• https://arxiv.org/abs/2009.14185

March 2022 announcement that Intel will present 14 quantum computing papers at the APS March 2022 Meeting:

• Intel Presents 14 Quantum Computing Papers at APS March Meeting 2022
• March 14–18, 2022
• https://www.intel.com/content/www/us/en/research/blogs/march-meeting-2022.html

APS is the American Physical Society, the premier professional organization, conference, and publication for physics.

Most of those papers are not yet available online, but here’s one of them that is — and is Open Access — in Nature:

• Qubits made by advanced semiconductor manufacturing
• March 29, 2022
• https://www.nature.com/articles/s41928-022-00727-9

The arXiv preprint of that paper is available online:

• Qubits made by advanced semiconductor manufacturing
• January 29, 2021
• https://arxiv.org/abs/2101.12650

April 2022 announcement that Intel will deliver its first quantum computing test bed to Argonne National Laboratory this year:

• Intel to install quantum computing test bed for Q-NEXT
• The machine will be the first major component installed in Argonne’s quantum foundry, which will serve as a factory for creating and testing new quantum materials and devices. It is expected to be completed this year.
• Q-NEXT scientists will use Intel’s machine to run quantum algorithms on a real, brick-and-mortar quantum computing test bed rather than in a simulated quantum environment.
• April 11, 2022
• https://www.anl.gov/article/intel-to-install-quantum-computing-test-bed-for-qnext

But the question arises as to what exactly Intel or Argonne considers a “test bed”. See my section entitled Test bed for a discussion of the possibilities.

q86 architecture for quantum computer systems

All of Intel’s quantum computing components would collectively be designed to function together to implement a complete quantum computer system. There may be other system components beyond those supplied by Intel, but Intel would supply the core, critical components which are not easily replicated by an organization wishing to design and build a quantum computer system and not easily and readily acquired on the open market as commodity components.

Intel would provide both the hardware components as well as technical specifications and documentation to guide customers on the path to designing and building complete quantum computer systems.

The goal of this paper is not to detail all of those system components or any of the details of any of those components but to refer to the entire architecture as if it were a monolithic entity, which from a brand, business, and technology perspective it is intended to be.

q86 is a name I contrived, to parallel x86

q86 is a name I contrived, to parallel x86, the recognized standard architecture for personal computer and server computer systems. I have no idea what name Intel would actually use if they adopt this proposal and one similar to it.

Quantum computer as a coprocessor

Generally speaking, a quantum computer is not a complete computer system capable of running full applications as a classical computer system does, but functions as a coprocessor or adjunct processor to the classical processor, running quantum algorithms on behalf of the quantum application running on a classical processor.

Relatively narrow computations will be extracted from classical software and reimplemented as quantum algorithms. That’s a tedious manual process.

The quantum algorithm would deliver functionally-compatible results to the classical computation but in a tiny fraction of the time.

And in some cases the quantum computation will deliver results which the comparable classical code would never be able to deliver in any acceptable amount of time — it might take weeks, months, years, decades, even centuries or many centuries for even the largest classical supercomputers or distributed computer systems to deliver those results.

In any case, the quantum algorithm is a black box which is implemented by the quantum computer system acting as a coprocessor.

The application would combine classical code running on a classical computer system with one or more quantum algorithms running on the quantum computer system acting as a coprocessor.

Technically, the quantum computer executes quantum circuits, but the distinction between a quantum algorithm and a quantum circuit is beyond the scope of this paper. For our purposes here, they are functionally equivalent.

Tight integration of q86 with x86

Although the q86 architecture should be independent of any classical computing architecture which is using it, the general expectation is that the average q86 customer would tightly integrate it with the x86 architecture.

Details are beyond the scope of this paper. All that matters here is to highlight the opportunity for tighter integration.

q86-compatible quantum computer systems

Any quantum computer system designed and built with the q86 components from Intel would be known as q86-compatible — a q86-compatible quantum computer system.

Being q86-compatible would imply that the quantum computer system would support quantum algorithms and quantum applications which are designed to run on q86-compatible quantum computer systems.

The commitment is intended to be that if your quantum algorithms and quantum applications run on one q86-compatible quantum computer system then they should run on any other q86-compatible quantum computer system — provided that it has comparable technical specs such as number of qubits, qubit fidelity, qubit connectivity, and phase granularity.

q86 architecture — a big question mark at present

Unfortunately, the notion of a q86 architecture for quantum computer systems as espoused in this paper is purely speculative on my part and may never come to fruition.

I do think it is possible and even likely, but that doesn’t assure that it really will happen.

All that I can personally do is propose possibilities.

Publish fully documented reference designs

Although customers are free to design and build quantum computer systems according to their own proprietary designs using Intel components, it would still be helpful for Intel to publish very detailed designs for how to build a fully functional quantum computer using Intel components.

Some customers may only wish to modify the designs in limited areas, but would benefit from using the published Intel designs in all other areas.

Publish Principles of Operation and Implementation Specifications

The Principles of Operation document tells a programmer everything they need to know about a computer system to write code for the system. Typically there would be one document for a whole family of processors.

This would exclude processor-specific details such as performance and capacity.

There would be a separate document, the Implementation Specification, for each processor of the family which tells a programmer all of the details they might want to know to code well for that particular processor, such as performance and capacity, and any details which might vary between members of the processor family.

Intel must produce these documents for their reference designs.

Individual vendors and customers designing and building their own quantum computer systems using Intel components may need to customize these documents to the extent that they customize the design of their quantum computer systems in ways which differ from the Intel reference designs.

For more discussion of Principles of Operation and Implementation Specifications for quantum computers, see my paper:

• Framework for Principles of Operation for a Quantum Computer
• https://jackkrupansky.medium.com/framework-for-principles-of-operation-for-a-quantum-computer-652ead10bc48

Produce limited machines as reference platforms

Even if customers fully intend to design and build their own quantum computer systems using Intel components, it would still be helpful to acquire or at least gain access to quantum computer systems produced directly by Intel — which I am calling reference platforms in this proposal.

At a minimum such systems from Intel would be useful for benchmark comparisons.

And such systems would be useful for a lot of activities which wouldn’t otherwise be possible until the customer finally has their own quantum computer systems fully operational. This can include testing of software components as well as quantum algorithms and quantum applications.

And maybe the Intel-supplied reference platforms can be used as testbeds for testing proprietary hardware development efforts.

And firms focused on software, algorithms, and applications could of course utilize reference platforms from Intel for both development and testing.

And academic researchers could use such reference platforms as well, especially if they wish to test out hardware innovations short of designing and building entire quantum computer systems.

Test bed (or testbed)

People throw around the vague term testbed (or test bed) in a cavalier and ambiguous manner so that it’s not clear what is really meant. It could simply refer to the reference platform concept which I just described. Or, it might be any of the following based on who is using it and the specific context:

1. An early version of a new product. Such as an alpha test or beta test version. It may (or may not) be fully functional, but simply needs testing and some polishing. Documentation may be incomplete.
2. Incomplete version of a new product. Known not to be fully functional, but sufficient for experimentation for some purposes.
3. Release candidate. It actually is fully functional and ready for production use, but may simply need some final testing and bureaucratic approval. All ready, but not a truly official product.
4. One or more portions of the full product. Not enough to constitute a full and complete product, but enough to test and validate carefully identified subsets of the full product.
5. A prototype for a product. Not intended to be the real, final product, but simply to test out concepts and approaches.
6. A preliminary version of a product. More than just a mere prototype, it is intended to represent the full product but with too much uncertainty to presume that it really will become the full product. Maybe it will work, or maybe it will encounter fatal flaws requiring a substantial redesign or even going back to the drawing board.
7. Fully functional subsystems of the full product. Intended to test and evaluate those specific subsystems rather than the full product.
8. Full (or partial) product plus extra hardware and software to facilitate testing. Focus is on facilitating testing, not production function per se. May in fact have limited performance and limited capacity if that facilitates testing. Facilitate examination of internal state which might not be possible in production usage.
9. Fully functional product intended only for testing, not production. There may or may not be technical limits on usage, but contractually production usage is not permitted. And maybe constraints on disclosure to other parties, such as NDAs — Non-Disclosure Agreements.

Original Equipment Manufacturers (OEMs)

The traditional computer hardware systems vendors for personal computers and servers are also known as OEMs — Original Equipment Manufacturers. They design, build, and sell complete hardware systems based on high-value components which they purchase from hardware component vendors such as Intel.

HP and Dell are two prime examples of OEMs.

Or even IBM, even though they have their own quantum computer systems, they may seek to diversify and address commodity and entry level markets. IBM could do for quantum computing what they did for… the IBM PC.

These and other OEMs would purchase q86 components from Intel (directly or through a distributor) and package them into complete quantum computer systems capable of supporting quantum algorithms and quantum applications which are q86-compatible.

Not all q86 customers would be OEMs

Although not all Intel q86 customers would be OEMs. Customers designing and building their own quantum computer systems or researchers in academia would purchase components from Intel, but not be OEMs since they are not selling their quantum computer systems to others.

Technically, distributors are customers of Intel and q86, and they are not OEMs. They are in fact reselling Intel q86 components to OEMs (or other organizations who are not OEMs.)

A rich and thriving ecosystem

A key goal is to generate a large and vibrant ecosystem with many vendors, partners, and suppliers, each doing their niche part in a much larger ecosystem. Just as we have today with the x86 architecture.

Each player in the ecosystem has their own share of a much larger pie than the quantum computing components supplied by Intel.

Intel may indeed be the dominant player, but it is vital and urgent that Intel does not so overshadow the rest of the ecosystem so as to stifle competition and innovation.

Some of the areas of a thriving ecosystem for quantum computing would include:

1. Hardware component vendors. OEMs.
2. Operating systems.
3. Infrastructure software.
4. Support software.
5. Software tools.
6. System management software.
7. Compilers.
8. Component libraries.
9. Application frameworks.
10. Programming models.
11. Algorithm designers.
12. Application developers.
13. Software developers.
14. Solution developers.
15. Solution providers.
16. System integrators.
17. Distributors.
18. Value-Added Resellers (VARs).
19. Turn-key solution providers.
20. Consultants.
21. Trainers.
22. Conferences, trade shows, and seminars.
23. Researchers. Both academic and commercial.
24. Journalists and media. Both tech-oriented and mainstream. Both offline and online. And books — both offline and electronic.

Distributors

Not everyone who purchases q86 quantum computer components would necessarily purchase them directly from Intel. As with all electronic components, customers, even large firms, tend to make their purchases through distributors.

Value-added resellers (VARs)

Value-added resellers (VARs) purchase complete hardware systems from OEMs and then add applications, packaged solutions, and other software or even hardware to the basic hardware systems. Their customers get more complete systems.

Actually, there can be a range of levels of value added, with different vendors offering different and complementary value(s). There can be a chain of value added, additional value being added at each step of the chain.

Turn-key solution providers

A turn-key solution provider is a step beyond VARs, providing everything that the user needs to use the system. Everything is in place. Nothing more to add. Plug it in, turn it on, and away you go.

There may be additional configuration and setup required, but all of the hardware and software components would be in place. Generally any additional configuration and setup would be performed by a specialist from the solution provider, a solution specialist.

Open source is essential

It is no exaggeration to suggest that open source is essential. It is absolutely vital.

Everything should be open source. This includes:

1. All of the software for sure.
2. Generally firmware as well.
3. Any diagnostics, configuration tools, and support software.
4. And even hardware designs when possible.

Open source facilitates customization and extension by customers and academic and government researchers.

Open source is the single best way to make the Intel architecture the place where quantum computing innovation occurs.

Facilitate academic research in quantum computing hardware

Sure, there are a number of larger academic institutions which have the expertise and resources to develop complete quantum computing systems from scratch, but that does place a rather severe limit on academic research in quantum computing hardware.

The proposal of this paper dramatically lowers the bar to entry for an academic institution wishing to do research in selected areas of quantum computing hardware.

In particular, a research team might prefer to be focussed on some narrow, niche aspect of the hardware, in which case having to replicate all of the rest of the hardware for a quantum computer system is a rather onerous burden, not to mention a significant capital expense.

The proposal of this paper would enable lean research teams to focus on their particular area of interest, leveraging Intel for the remainder of the hardware required for the complete quantum computer system.

Academic consortium for customized designs for quantum computers for academic institutions

Some academic institutions will have the skills and resources to design and build their own customized quantum computer systems, while others will not. For those academic institutions who need customization, when standard commercial configurations are insufficient for their needs and interests, and customization is beyond their own skills and resources, an academic consortium to design and build customized quantum computer systems would be helpful and beneficial.

There could be any number of such consortia, each focused on the specific needs and interests of its academic members. The scope of a consortium could be regional or geographic, or functional, such as for a specific or general application area.

The details are beyond the scope of this informal paper and can vary from consortium to consortium.

A consortium may in fact simply contract out all work to a contract engineering and manufacturing firm, or the consortium might acquire components directly from Intel or a distributor and design and build quantum computer systems to the specifications of the consortium member.

In some cases academic institutions may opt to do all of the same work as some of the consortiums are doing but without any of the overhead. The consortiums are a great place to start when nobody has any expertise, and then some consortium members may go their own way as they gain knowledge, experience, and expertise.

Contract manufacturers may have limited abilities to do customization work themselves, so that the consortium may merely supervise the overall process but not do any of the technical work itself.

Contract engineering and manufacturing firms

A contract engineering and manufacturing firm may do all of the technical work of a hardware system vendor but according to technical specifications supplied by the customer.

There are two basic business models:

1. White label. Simply a generic commodity. No branding. No customization. Distinct from custom designs. Customer simply needs a significant volume of generic boxes.
2. Customized designs. No longer simply a generic commodity. Customers must supply a technical specification detailing the extent and details of the customization.

And there can be distinctions for customized designs:

1. System size. No change in configuration other than variations in module count and number of qubits
2. Configuration. Selection and arrangement of components, but still using off the shelf Intel components.
3. Some custom components. With or without Intel’s support and blessing.

Several modalities for custom components:

1. Firm may work closely with Intel for custom components.
2. Firm may do its own design without Intel’s approval.
3. Firm may even do its own design in the face of outright disapproval by Intel.

Analog components and RF components

Although Intel’s core and historical strength and competency is in digital electronic components, quantum computing is a bit odd. There is a much heavier dependence on analog components and radio frequency components in particular. Intel could try sticking to digital components alone, but there are critical benefits to including critical analog components in their portfolio as well.

The key, critical criteria for inclusion is simply whether the customer and user of Intel quantum computing components would struggle and have difficulty if they had to acquire and integrate analog and RF components on their own. If it were easy for the customer to do it, then there would be no merit to Intel doing it.

Given the very nature of quantum computing, with quantum effects and cryogenic temperatures, it’s a virtual slam dunk that it makes sense for Intel to worry about the analog and RF integration so that the customer of Intel quantum computing components does not need to worry about them.

Timing

Exactly when the proposal of this paper could or should be put into effect is beyond the scope of this paper. Some factors and considerations include:

1. Maybe this proposal doesn’t take root with the current generation of qubit hardware.
2. May have to wait for a future generation of hardware, possibly more amenable to commoditization.
3. Current focus is pre-commercialization — research, prototyping, and experimentation. It may take quite some time until quantum computing hardware is ready for commercialization.
4. And it would all seem pointless until Intel has fully-functional hardware ready for mass production and delivery to customers.

Historical timing

A quick look back at the timing for Intel with the personal computer and servers, and microprocessors in general.

Details here are from the WIkipedia:

• Wikipedia article for List of Intel processors
• https://en.wikipedia.org/wiki/List_of_Intel_processors

1. 4-bit 4004. 1971.
2. 8-bit 8008. 1972.
3. 4-bit 4040. 1974.
4. 8-bit 8080. 1974.
5. MITS Altair 8800. 1974. Used the Intel 8080.
6. 16-bit 8086. 1978. Used by some early low-end workstations and personal computers, such as the IWS (Integrated Workstation) by Convergent Technologies in 1980. Limited to modest success.
7. 8/16-bit 8088. 1978. Used by the original IBM PC
8. IBM PC. 1981. Used the Intel 8088. Wildly successful.
9. 16-bit 80286. 1982. Used by the IBM PC AT.
10. IBM PC AT. 1984. Used the 80286. Very wildly successful. Significantly more capable than the original IBM PC.
11. 32-bit 80386DX. 1985. Finally, some real horsepower. No longer a toy. Supported virtual memory and memory beyond the original PC 640K limit.
12. Compaq 386 Deskpro. 1986. Wildly successful. Very capable.
13. 16/32-bit 80386SX. 1988. Functionally-compatible with the full 32-bit 80386DX, but lower performance to achieve lower cost and lower power. Enabled low-cost mobile computing — laptops.
14. 32-bit 80486DX. 1989. Desktop workstations. Servers. Math coprocessor built in.
15. 32-bit 80486SX. 1991. Enabled full-power mobile computing. 80486DX without the math coprocessor.
16. Pentium II Xeon. 1998. First high-end server chip.
17. Pentium III Xeon. 1999. Higher performance server.
18. Xeon DP and Xeon MP. 2001. Higher performance server.
19. 64-bit Itanium. 2001. Incompatible 64-bit architecture. A flop as AMD introduced an x86-compatible 64-bit processor.
20. 64-bit Xeon DP and Xeon MP. 2004.
21. Dual-Core Xeon. 2005.
22. 64-bit Core 2. 2006. Desktop 64-bit processing.
23. And so much more. Many successful, some not so successful.

The point here, for this paper, is to examine what points along that historical timeline were more appropriate to push for Intel as the premier vendor of components for personal computers and servers.

Too early:

1. The early microprocessors such as the 4004 and 8008.
2. Even the 8080 was too early. But getting close.
3. The 8086 was almost there.
4. The 8088 may have been the ideal spot.

Too late:

1. The 80286 may have been a little late.
2. The 80386 would have been too late. The party was in full swing. Needed to get started well before then.
3. The 80486 would have been too late. Too late fo desktops, but not for laptops and servers.
4. Xeon too late. For personal computers, but not for servers.
5. Core 2 too late. The x86 architecture and ecosystem were well underway before this.

Personally, I would have ramped up with the 8086 since the basic technology was in place even if the 8088 was the big market winner. And the 80486 was the sweet spot for the start of both servers and mobile computing. The 80386SX was a start for mobile computing, but a very weak start, in my opinion.

How much might a quantum computer system cost?

Generally this question is beyond the scope of this paper at this time, but it’s worth raising the question on principle. So, how much might a quantum computer system cost?

Some aspects of pricing:

1. Retail price. Total cost to buy a single system.
2. Volume pricing. Discount from retail price for multiple or many systems.
3. Hardware component cost. Money to Intel and/or distributors. Cost before the components are assembled into a complete quantum computing system.
4. Cost of dilution refrigerator and packaging. Beyond the cost of electronic components from Intel, what does a dilution refrigerator and its associated equipment cost, as well as the overall packaging for the quantum computing system.
5. Cost to assemble and test. How much does it cost to assemble and test a full quantum computer system?

Here are some wild guesses for possible total quantum computing system costs:

1. $100,000. Bare bones simple system. Possibly second-hand components.
2. $250,000. A little better than a bare bones simple system. Possibly second-hand components.
3. $500,000. Moderately better than a bare bones simple system. Possibly second-hand components.
4. $1,000,000. Entry level system. More than bare bones.
5. $2,500,000. Above entry level system.
6. $5,000,000. Mid-range system.
7. $7,500,000. Mid-range system. Larger configuration.
8. $10,000,000. Super-mid-range system. More than mid-range.
9. $15,000,000. Super-mid-range system. More than mid-range. Larger configuration.
10. $20,000,000. Super-mid-range system. More than mid-range. Larger configuration.
11. $25,000,000. High-end system. More than super-mid-range. Multiple processors.
12. $50,000,000. Super-high-end system. Lots of processors.

Pricing for service — leasing and usage

Not everyone will wish to purchase a quantum computer system to own it all for themselves. Alternatives include:

1. Leasing. Interesting questions about the useful service life of a computer system. How long before it becomes obsolete? How long before it is ready to be discarded in favor of newer systems? How long can a lease be before it is effectively ownership for the useful life of the system?
2. Dedicated. May be similar to leasing, but focused on 100% availability and no term of lease per se.
3. Metered usage. Pay as you go. Simply pay for whatever service you use on an hourly basis.
4. Guaranteed availability. Beyond paying for actual usage, pay a premium to be guaranteed that the system will be available when needed. At the extreme, 24/7 availability. Or for x hours within a specified daily time interval — may not need real-time access, just guarantee that the system will be available somewhere between a specified start time of day and a specified end time of day.

But this is pricing for service of the final system. This paper is focused on production of components which can be used to construct usable systems, so the issue of pricing of final systems is beyond the scope of this paper.

Service level agreement (SLA)

A service level agreement (SLA) lays out a contractual commitment of availability and quality of service for a system. This is more relevant for a usage-based service. The service provider will generally have to maintain an excess of equipment to assure committed service in the face of potential equipment downtime.

This paper is focused on production of components which can be used to construct usable systems, so the issue of pricing of service for final systems is beyond the scope of this paper.

Access: Experimental, testing, and production

There are three levels of access or needs for quantum computer systems:

1. Experimental. For research and prototyping. Nothing critical.
2. Testing. Preparation for production. Validating that the system is suitable for production.
3. Production. Fully operational in a mission-critical mode. Everything has to work, all of the time. Performance is required. Capacity is required. Quality is required. Availability is required.

Siting

There are a wide variety of choices for where a customer might place a quantum computer system.

1. Vendor cloud access. A service provider owns and maintains the system at a site of their choosing.
2. In-house. The customer sites the system at their own location.
3. Corporate data center. The customer has their own data center at which they can site the system.
4. External data center. The customer chooses an external data center to site the system.
5. Cloud vendor. A cloud service provider such as Amazon offers remote access to systems sited at Amazon data centers.
6. Hardware vendor. The hardware vendor maintains the quantum computer systems at its own site and offers remote access to customers. Such as Intel for reference platforms.

Siting is separate from pricing and availability commitments, such as:

1. Service pricing.
2. Availability commitment.
3. Access.

This paper is focused on production of components which can be used to construct usable systems, so the issue of siting for final systems is beyond the scope of this paper.

Government labs might wish to build their own quantum computer systems

Some government agencies will actually have the skills and resources to design and build their own quantum computer systems. In some cases even down to the qubit level, but the availability of a rich collection of quantum computer system components from Intel will greatly facilitate the design and development of relatively custom quantum computing systems by a number of government agencies, such as:

1. DOE national labs. Such as Argonne, Oak Ridge, Lawrence Livermore, Sandia, and Los Alamos. Research and modeling for nuclear weapons, nuclear energy, high-energy physics, and quantum computing itself, etc.
2. NIST. A fair amount of quantum computing research originated at NIST. General emphasis on quantum information science overall, including quantum metrology, quantum sensing, quantum communication, quantum networking, post-quantum cryptography, and of course quantum computing.
3. NSA. Secretive but plenty of skills, money, and resources available for classified but high value compute-intensive tasks.
4. NOAA. Weather and climate research, modeling, and forecasting,
5. NASA. Aerodynamic research, astronomy, spaceflight, climate, and other research.

Architectures tailored for high-end quantum simulators

Although the general focus of this paper is on actual, real quantum computers, classical simulators are also of great interest, and Intel can play a significant role, especially considering their role in classical computing and the x86 architecture.

Even with high-end quantum computers available, simulators are still needed for debugging since quantum circuits cannot be debugged in a classical sense — quantum state cannot be read nondestructively.

As the qubit count grows, the classical resource requirements grow exponentially.

Intel can’t fully bypass that exponential growth, but can make classical resource usage more efficient.

The three main goals are:

1. Increase qubit count.
2. Increase circuit depth.
3. Reduce resource requirements. The data and processing required for each gate execution.

Details are beyond the scope of this current paper, but opportunities for Intel include:

1. Optimal use of x86. Anybody can code a simulator, but Intel has special expertise both in quantum and the use of the x86 architecture. Start using strictly off the shelf commodity processors. Focus on offering reference configurations and documentation.
2. Enhanced use of x86. Going beyond the basic, standard, commodity x86. Upgraded bus, memory, storage, and networking architecture, including tightly-coupled processors and memory, etc. but all still based on off the shelf x86. Offer reference designs as well as reference platform implementations.
3. Enhancements to x86. Intel could introduce new microprocessors which extend the current x86. Such as adding GPU capabilities. Simulation-specific capabilities.
4. Simulation-specific processors. A separate line of custom microprocessors focused on simulation of the execution of quantum circuits. Smaller and simpler processors — and lots of them. Custom data format for state vectors and transformation matrices. State compression for more efficient storage and faster execution.
5. Very large scale simulators. Try to push above simulation of 50 qubits, maybe even to 60 or even 64 qubits. Up to even my idea of a simulator the size of a full data center or a football field. Room-size at a minimum. Could either base it on strictly commodity x86 processors or custom simulator processors — benchmark both approaches on a smaller scale first.

Subsystems

It’s premature to judge whether Intel’s components for quantum computers would simply be discrete chips and chip sets or might constitute entire subsystems — circuits boards with more than a few chips or even collections of circuit boards. That said, it just seems inevitable that some degree of subsystems would be likely.

Maybe even to the extent of packaging all of the electronic components for a full quantum computer system, everything but the physical box and the dilution refrigerator.

Or maybe all of the circuitry that goes in the dilution refrigerator might be an integrated subsystem.

The only point here is as a placeholder to allow for the possibility of subsystems.

Co-design of dilution refrigerator, wiring, and electronics to match each other?

At present, the dilution refrigerator is a distinct component from another vendor, but conceptually Intel could design all of the components to work better together.

I don’t want to prejudge Intel’s design decisions, but simply to allow for possibilities.

As things stand now, the dilution refrigerator stands out as a singular component (or subsystem) which doesn’t fit in with Intel’s standard and traditional line of capabilities and products.

But since the dilution refrigerator is a special challenge for any organization wishing to design and build their own quantum computer, it makes sense for Intel to at least contemplate incorporating it under the umbrella of components which they design, build, and test so that customers don’t have to do so on their own or to rely on an additional vendor with all of its complexities, uncertainties, and testing.

Industry standards — potential but premature

A key part of the proposal of this paper is heavy support for industry standards. And that is indeed a significant opportunity for Intel. Unfortunately, it’s premature to pursue any of that. Until we have working and proven quantum computing technologies it’s somewhat pointless to try to develop and promote standards for technologies which don’t exist.

Only in rare situations does it make sense to pursue standards for future technological advances. Quantum computing is not one of those rare situations, at least not yet.

Once Intel and its customers actually build and test real live quantum computer systems for real applications for a couple of years, then and only then will standardization begin to make sense.

Once Intel has a stable enough engineering base to call it a q86 architecture (or whatever Intel decides to call it), then standardization will begin to make sense.

Might AMD or NVIDIA beat Intel to the punch?

Although my focus — and belief — in the proposal of this paper is that Intel is ideally suited to play this role as the primary supplier of components to organizations wishing to build their own quantum computer systems, there are other possible vendors who could play this role, particularly AMD and NVIDIA.

1. AMD. Already an alternative to Intel for personal computer and server microprocessors.
2. NVIDIA. Focused on the high-end GPU market, close to the market for quantum computer systems.

Intel has key advantages, but AMD and NVIDIA are still potential contenders, especially if Intel doesn’t decide to play the role envisioned by this paper.

Intel’s key advantages

As much as other competitors have potential to compete with Intel for the role envisioned by this paper, Intel still has key advantages:

1. Already has a head start. Research in quantum computing is well underway at Intel and its partners.
2. Experience at leading the market. Their experience leading the market for personal computers and servers.
3. Finance. Intel has deeper pockets. More money available to throw at speculative and risky projects.
4. First mover advantage. Okay, technically, they aren’t actually in the market yet. But they are ahead in at least some of the relevant research.
5. They’re Intel. Their reputation with advanced technology.

Other potential competitors

Besides AMD and NVIDIA, there are plenty of other deep-pocketed tech companies who could enter this market. I won’t do any deep analysis, just to mention some names:

1. IBM. Sure, they could do it, but do they have that kind of focus or interest? They gave us the IBM PC and ThinkPad, but then walked away from their own success.
2. Micron. It’s possible although not so likely. They do high-technology but low-value commodity components (memory chips), and not high-value components. Still…
3. Amazon. Just build the quantum computer systems they wish to deploy on AWS.
4. Oracle. Has some hardware expertise (acquired Sun Microsystems), lots of data experience, and ambitions for a lot more.
5. Infineon. Announced deal for trapped-ion technology with Oxford Ionics.
6. MITRE. Government and defense and security orientation. Although more focused on research and government contracts than commercial business.
7. Accenture. A real long shot. They could contract out for all of the hardware expertise, but focus on system design for high-end corporate applications. Maybe they would simply be a turn-key solution provider.
8. Apple. Kind of a long shot.
9. Exxon. They owned Zilog and Exxon Office Systems at one point. An application-focused architecture might be appealing.
10. Lenovo.
11. Samsung.
12. Japan. Various Japanese hardware companies.
13. India. Who knows!
14. China. Who knows!
15. UK. Unspecified, but possible.
16. Europe. Unspecified, but possible.

Also, all of these firms also have potential to be system integrators who could build quantum computer systems of their own using Intel quantum processor technology and market as their own products.

Might some scrappy startup eat Intel’s lunch?

Besides large, established tech companies, it’s quite possible that some scrappy startup could come along and eat Intel’s lunch. They would need to understand that focusing on the high-end Cadillac segment of the market misses the bigger picture, and that the real money is to be made with the low-end and mid-range commodity and premium-commodity mass markets. Exactly were Intel has thrived.

Upgrade from FPGA to full-custom logic to boost performance and capacity and reduce cost

For low-volume production, an FPGA (Field-Programmable Gate Array) is a relatively cheap and easy way to implement hardware logic. But there are trade-offs when compared to full-custom logic implemented directly on the main integrated circuit, namely performance and capacity.

Intel, producing chips at a much higher volume, could amortize the increased up-front cost of full-custom logic over a much higher unit volume. In return, Intel could offer a significant boost in both performance and capacity.

Full-custom logic is less flexible than an FPGA — it can be dynamically changed under software control, but that flexibility won’t be needed as much once we get more experience with high-performance and high-capacity quantum computing so that the logic won’t change so often.

At present, FPGA usage is less about the flexibility of the chip (dynamic update) than the lower upfront design cost. The ease of handling changes to the logic is more of an added benefit rather than the main benefit — reduced upfront design cost.

The bottom line is the bottom line for users — lower system cost, higher performance, and higher capacity.

Universal quantum computer — hybrid quantum/classical integrated processor with QRAM

Although this is a much longer-term proposition, a universal hybrid quantum/classical integrated processor with QRAM AKA a universal quantum computer would deliver a multitude of benefits and be much easier to use to deliver the benefits of quantum computing to classical applications, or technically to hybrid quantum/classical applications.

From a proposal I wrote almost four years ago:

• The essence of a universal quantum computer is that it combines the full power of a classical computer with the power of a quantum computer, and enables simulation of physics, including and especially quantum mechanics.

The notion of QRAM is vague and ambiguous, and different people use it differently, but as used in my proposal it enables classical code to read and write directly into a quantum memory — since classical binary 0 and 1 are a strict subset of quantum information — as well as limited quantum operations without the need to resort to full quantum circuits, but as an adjunct to the full power of full quantum computing. This would allow a much larger amount of classical data to be fed into quantum algorithms and make it easier to get larger volumes of data out.

I also suggest the possibility that this new hybrid processor and QRAM should work with 128 bits at a time rather than 32 or 64 bits. Again, the goal being to move a lot more data in and out of quantum processing at a time. This is commonly referred to as SPAM — State Preparation And Measurement. Input data is state preparation and output data is measurement.

Again, this is a much longer-term proposition, and is only mentioned here to demonstrate how the basic proposal of this paper can head in that ultimate direction. And, most relevantly, it’s a natural fit for Intel.

Read my proposal for a universal quantum computer including QRAM, here:

• What Is a Universal Quantum Computer?
• https://jackkrupansky.medium.com/what-is-a-universal-quantum-computer-db183fd1f15a

QRAM would offer dramatic improvement in SPAM — State Preparation And Measurement

The preceding section on a universal hybrid quantum/classical processor with QRAM and supporting direct access to quantum memory from classical code with a bandwidth of 128 bits (or qubits) noted that SPAM — State Preparation And Measurement — would be greatly facilitated. The effect would be to offer a dramatic boost in both performance and capacity.

It still wouldn’t be enough to support classic Big Data, but would enable moderately larger computations than are possible with current quantum computer architectures. This could enable input and output of hundreds or even thousands of bits or qubits.

Let Intel handle all of your intellectual property (IP) issues

Intellectual property (IP), mainly patents, is a vast and looming issue for the emerging quantum computing sector. I’m not sure how much longer competing hardware vendors with competing patents can progress before many of them run into patent infringement claims, limits on their hardware design freedom, or onerous licensing fees. This will eventually limit new entrants and/or raise the cost of entry as a result of patent licensing fees.

Intel is an old hand at intellectual property and patents. If you want to develop your own chips you will quickly run into patent licensing and patent infringement issues. So if you can acquire chips from Intel you can “outsource” all of your intellectual property headaches, at least those related to the functions of those chips, but that can be substantial given the complexity of a quantum computer system.

Paying Intel a premium price for their chips could actually be quite economical and greatly limit intellectual property risk. And eliminating all of those intellectual property headaches can be… priceless.

And the more components of a complete quantum computer system that Intel provides, the more your intellectual property hassles and risks will be fully covered by your chip prices.

So let Intel control all of the intellectual property that they want.

Intel will also likely have to license some IP from others. In some cases Intel can outright buy the firm owning the patents.

Hardware vendors would pay for the intellectual property through component prices as they do for x86.

There’s no need to encounter the hassles of patents and patent disputes when all you want to do is build your own quantum computer system.

Intel venture capital investments to spur progress in the quantum computing ecosystem

Intel has a venture capital arm, Intel Capital, which makes venture capital investments in tech startups. It is my expectation and proposal that Intel Capital will make investments in promising startups which will spur the growth of the quantum computing ecosystem.

There should probably be a special arm of Intel Capital which is exclusively focused on quantum technologies.

Or even an arm more narrowly focused on the q86 ecosystem since there are many quantum technologies which aren’t strictly quantum computing per se.

Such investments would become partners in promoting and supporting Intel quantum computing efforts, such as designing, distributing, and supporting or exploiting components for producing and using quantum computing systems.

Some possible types of investments:

1. Quantum applications. Horizontal or industry specific.
2. Quantum application development consulting.
3. Quantum solutions. Including holding the customers’ hands.
4. Industry-specific consulting services. Industrial. Finance. Transportation. Logistics. Federal.
5. Quantum application framework tools.
6. Quantum system management tools.
7. Debugging tools.
8. Add-on hardware.
9. Hardware debugging add-ons.
10. Ruggedized quantum hardware.
11. Quantum supercomputers. Networked quantum computer systems. Work closely with customers to meet their needs.
12. Quantum training and education services.

The possibilities are endless.

Academic research grants

A key aspect of this proposal would be for Intel to continue and accelerate financial grants to academic research efforts. The essential goal of such efforts is to fuel future growth. The exact form of such grants can vary greatly, but one example is their relationship with QuTech at TU Delft:

2015 announcement of collaboration between Intel and QuTech at TU Delft:

• QuTech quantum institute enters into collaboration with Intel
• American chip manufacturer Intel and QuTech, the quantum institute of TU Delft and TNO, have finalised plans for a ten-year intensive collaboration. Alongside financial support for QuTech totalling approximately $50 million, Intel will also contribute expertise, manpower and facilities. “QuTech is delighted to welcome Intel as their new partner. The major challenge facing quantum technology development in the coming decades, such as creating a quantum computer, is set to be upscaling: being able to create complex structures with an enormous number of quantum bits. This partnership will enable us to combine our scientific expertise with the best engineering expertise in the computer industry”, says lead scientist Lieven Vandersypen of QuTech.
• September 3, 2015
• https://qutech.nl/2015/09/03/qutechentersintocollaborationwithintel/

The range of such efforts has the prospect of broadening widely once something like the proposal of this paper becomes a reality, with a diverse range of research teams being able to engage in a wide range of quantum computing hardware research at a relative modest expense compared to the $50 million required for the QuTech at TuDelft collaboration.

It is precisely such research efforts which will fuel future growth. Product engineers can innovate, but research results fuel product engineering.

What needs to happen next?

What’s next? Much of this proposal is moot until Intel actually has a working quantum computer. So, lots of research is needed by Intel, first.

Let’s see where they are in two years or so.

But, of course, deciding to pursue a path of building components for others to build their own quantum computer systems could come at any time. The sooner, the better.

When might this proposal come to fruition? Six months to three years, or longer

My general expectation is that two years might be the sweet spot for the proposal of this paper to finally come to fruition. But there’s a wide range of uncertainty around that. Some possibilities:

1. Six months. This would be about the timeframe in which Intel is scheduled to deliver a testbed system to Argonne National Laboratory.
2. One year. It’s possible, but no slam dunk.
3. Eighteen months. A little more likely.
4. Two years. A much more likely timeframe.
5. Two and a half years. Allow a little more time for unforeseen delays.
6. Three years. Definitely should have happened by now. If not, we’re facing a Quantum Winter.
7. Four years. Yeah, it could take this long since so much in quantum is… uncertain. We may be in the middle of a Quantum Winter by then, but sometimes that’s the best time to start a massive new enterprise, when you’re the only notable show in town and resources and talent are cheaper and readily available.
8. Five years. It really could take this long. Some competitors may do moderately well over the next few years, but then run into a wall or get bloated or complacent, and then this might be the ideal time for Intel to take over the sector.

My original proposal for this topic

For reference, here is the original proposal I had for this topic. It may have some value for some people wanting a more concise summary of this paper.

• What I want to see from Intel in quantum computing is a focus on components which enable others to easily build quantum computers. Do for quantum computing what they did for personal computing and servers. Focus on the key circuitry — processors and supporting chipsets, and some other tricky components for qubits and their control and readout. Enable other vendors to easily design and build quantum computers. Like HP, Dell, even IBM (to diversify and address commodity and entry level markets). Maybe Amazon and Oracle as well. Maybe license designs to AMD, et al to diversify vendor supply chains. Produce limited machines as reference platforms. Maybe licensing for white-label production, and maybe even for Amazon, et al. Call it the q86 architecture to parallel the x86. Also look at tighter integration with x86 and eventually a universal hybrid classical/quantum processor. Maybe analog components as well. Maybe even wiring and connectors since they are so critical. Maybe everything but the refrigerator and box it all goes in — at least all of the electronics.

Alternative titles

I spent a fair amount of time coming up with a variety of alternative titles for this paper before settling on the current title. There were some I liked better, but they were just too long. I list them all here for background and because they may be useful as summaries of this paper for others to use.

1. What I Want to See From Intel in Quantum Computing Is a Focus on Components Which Enable Others to Easily Build Quantum Computers
2. We need Intel to Focus on Components Which Enable Others to Easily Build Quantum Computers
3. What if Intel Focuses on Components Which Enable Others to Easily Build Quantum Computers?
4. What If Intel Focuses on Components Which Enable Others to Easily Build Their Own Quantum Computers?
5. What If Intel Were to Focus on Components Which Enable Others to Easily Build Their Own Quantum Computers?
6. What If Intel Focuses on Components Which Enable Others to Easily Build Their Own Quantum Computers?
7. Awesome If Intel Focuses on Components Which Enable Others to Build Their Own Quantum Computers
8. Could Intel Focus on Components Which Enable Others to Build Their Own Quantum Computers?
9. It would be great if Intel Focuses on Components Which Enable Others to Build Their Own Quantum Computers
10. Modest Speculative Proposal: Let’s Have Intel Focus on Components Which Enable Others to Build Their Own Quantum Computers
11. Modest Speculative Proposal: Let’s Have Intel Focus on Components for Others to Build Their Own Quantum Computers
12. The Potential for Intel to Focus on Components for Others to Build Their Own Quantum Computers
13. Intel Quantum Computing Manifesto.
14. The Potential for Intel to Focus on Components for Others to Easily Build Their Own Quantum Computers
15. Call for Intel to Focus on Components for Others to Easily Build Their Own Quantum Computers
16. Intel Should Focus on Components for Others to Easily Build Their Own Quantum Computers
17. q86 — An architecture for Intel Components for Building Quantum Computers
18. q86 — An architecture for Intel Components So Anyone Can Build a Quantum Computer

Summary and conclusions

1. To be clear, the proposal in this informal paper is purely speculative on my part — Intel has not suggested any such intentions.
2. Intel could do for quantum computing what they did for personal computing and servers. Focus on the key circuitry — processors and supporting chipsets — to enable other vendors to easily design and build quantum computers.
3. Traditional computer hardware systems vendors. Also known as OEMs (Original Equipment Manufacturers.) Such as HP and Dell. Or even IBM, to diversify and address commodity and entry level markets. IBM could do for quantum computing what they did for… the IBM PC.
4. Non-traditional hardware vendors. Amazon. Oracle. Accenture.
5. Larger customers could develop or contract for their own custom systems. Big car companies. Aerospace firms. Big drug firms. Big financial firms.
6. Contract engineering and manufacturing firms. Develop quantum computers to specs of larger and more sophisticated customers. Or even medium-sized customers. Using Intel components.
7. Opportunities for distributors.
8. Value-added resellers (VARs). Adding applications and packaged solutions to basic quantum computing hardware.
9. Turn-key solution providers. A step beyond VARs, providing everything that the user needs to use the system. Everything is in place. Nothing more to add. Plug it in, turn it on, and away you go.
10. Make it easier for academic researchers to build their own research systems. Enable them to focus on their research, not all of the supporting hardware.
11. Academic consortium for customized designs for quantum computers for academic institutions. For those who want customized designs for quantum computer systems but not do all of the engineering and manufacturing work themselves.
12. Maybe license designs to AMD, et al to diversify vendor supply chains.
13. Carefully and fully document reference designs. Publish all of the details to produce a fully functional quantum computer using Intel components.
14. Publish Principles of Operation and Implementation Specifications. The Principles of Operation document tells a programmer everything they need to know about a computer system to write code for the system. Typically one document for a whole family of processors. The Implementation Specification is a document for each processor which tells a programmer all of the details they might want to know to code well for that particular processor which might vary between members of the processor family.
15. Produce limited machines as reference platforms. According to published reference designs.
16. Maybe licensing for white-label production. And maybe even for Amazon, et al.
17. Intellectual property (IP). Intel can control all of the IP they want. As well as likely have to license some IP from others. Hardware vendors would pay for the IP through component prices as they do for x86. No need to deal with the hassles and headaches of patents and patent disputes and licensing.
18. Focus on massive electronics integration. Play to Intel’s strengths. More electronics and connectivity at the lowest levels of the frig. Less complex wiring. Less communication with the outside world.
19. Modular design. Intel designs and produces the basic modules. Vendors decide how many modules to support. Unclear whether 32 or 48 or 64 or 128 qubits is the sweet spot. Vendors could offer small, medium, large, and very large quantum computing systems.
20. Subsystems. It’s premature to judge whether Intel’s components for quantum computers would simply be discrete chips and chip sets or might constitute entire subsystems — circuits boards with more than a few chips or even collections of circuit boards. That said, it just seems inevitable that some degree of subsystems would be likely. The only point here is as a placeholder to allow for the possibility of subsystems.
21. Co-design of dilution refrigerator, wiring, and electronics to match each other? At present, the dilution refrigerator is a distinct component from another vendor, but conceptually Intel could design all of the components to work better together.
22. Open source is essential. Everything should be open source. All of the software for sure. Generally firmware as well. Any diagnostics, configuration tools, and support software. And even hardware designs when possible. Facilitate customization and extension by customers and academic and government researchers. Make the Intel architecture the place where quantum computing innovation occurs.
23. Call it the q86 architecture to parallel the x86.
24. Quantum computer as a coprocessor. Generally speaking, a quantum computer is not a complete computer system capable of running full applications as a classical computer system does, but to function as a coprocessor or adjunct processor to the classical processor, running quantum algorithms on behalf of the quantum application running on a classical processor.
25. A rich and thriving ecosystem. Many vendors, partners, and suppliers each doing their niche part in a much larger ecosystem. Just as we have today with the x86 architecture.
26. Also look at tighter integration with x86.
27. Maybe analog and RF components as well. Reduce uncertainty and complexity due to component mismatches. If it makes life substantially easier for customers, do it.
28. Maybe even wiring and connectors. They are so critical and so difficult to do quality control successfully.
29. Maybe everything but the refrigerator and box it all goes in. At least all of the critical electronics.
30. Support for high-end simulation as well. Needed since real quantum hardware doesn’t support debugging.
31. Upgrade from FPGA to full-custom logic to boost performance and capacity and reduce cost.
32. Eventually a universal hybrid classical/quantum processor with QRAM. But that’s beyond the scope of this paper.
33. QRAM would offer dramatic improvement in SPAM — State Preparation And Measurement. Handle larger amounts of input and output data and do it more efficiently.
34. Consider diversification to other qubit technologies. Eventually, but not required for the initial thrust of this proposal.
35. Support for industry standards. Hardware standards as well as software standards. Enable easy portability of quantum algorithms and quantum applications between hardware vendors.
36. Validation. Test to confirm adherence to standards and compatibility.
37. Benchmarking. Validate and document performance and capacity.
38. Leverage quality control across vendors. Quality control of complex electronics is difficult to get right. Each vendor wouldn’t have to replicate the work of Intel.
39. Someone else could do this. It doesn’t have to be Intel. Somebody else could be the new Intel, the Intel of quantum computing. AMD or NVIDIA might smell the blood in the water and decide to eat Intel’s lunch. Or some scrappy startup — who understands that focusing on the high-end Cadillac segment of the market misses the bigger picture, and that the real money is to be made with the low-end and mid-range commodity and premium-commodity mass markets.
40. Some scrappy startup could eat Intel’s lunch. Maybe less likely, but still possible. And certainly an exciting prospect.
41. Siting. Beyond the scope of this paper. In-house on-premise vs. vendor or cloud provider remote access.
42. Timing. Maybe this proposal doesn’t take root with the current generation of qubit hardware. May have to wait for a future generation of hardware, possibly more amenable to commoditization. Current focus is pre-commercialization — research, prototyping, and experimentation — may take quite some time until hardware is ready for commercialization. And it would all seem pointless until Intel has fully-functional hardware ready for mass production and delivery to customers.
43. Not the purpose of this paper to detail all aspects of Intel’s quantum computing work. Focus is on the big picture.
44. Doesn’t matter what their current offerings look like. Focus is on future potential.
45. Horse Ridge inspired this proposal. Reading the announcement of Intel’s Horse Ridge cryogenic control chip inspired the proposal espoused in this paper. It’s exactly the kind of complexity which Intel can master, and which can dramatically simplify the efforts of others to design and build their own quantum computer systems.
46. Intel venture capital investments to spur progress in the quantum computing ecosystem.
47. Academic research grants. Continue and accelerate to fuel future growth.
48. What needs to happen next? Just lots of research by Intel. Let’s see where they are in two years or so.
49. When might this proposal come to fruition? Six months to three years, or longer.

For more of my writing: List of My Papers on Quantum Computing.