IonQ: Don't Buy The Wrong Qubit Technology
Apr. 08, 2024 7:26 AM ETIonQ, Inc. (IONQ) Stock45 Comments
Stephen Tobin profile picture
Stephen Tobin
3.04K Followers
Summary
IONQ's trapped Ion qubits technology is not scalable enough for Quantum Supremacy.
IONQ is exploring markets where their computers may outperform classical computers and trying to reduce the number of qubits needed for specific problems.
IONQ must ensure access to qubits capable of scaling to quantum advantage to ensure the company's survival.
Logic Qubit with Error Protection vs Standard Qubits, Superposition, Animated Concept of Logic Qubit - 3D Rendering
atdigit/iStock via Getty Images
I first wrote about IONQ in August 2023; I gave them a strong buy rating, and the article was titled "Buy it Now Quantum Advantage is Close." I no longer believe quantum advantage is close for IONQ and am downgrading them to a hold rating.
I do not think the technology chosen by IonQ (NYSE:IONQ), trapped Ion qubits, is sufficiently scalable to deliver a computer capable of Quantum Advantage or Quantum Supremacy. I also believe IONQ has reached the same conclusion. Consequently, IONQ is exploring markets where their computers may perform better than existing classical computers and trying to reduce the qubits needed to solve specific problems.
This strategy may provide a commercial runway of several years and complement their current plan of providing quantum computers that people can use to prepare for the day when a computer with supremacy arrives.
To ensure the company's success, IONQ must ensure they have access to qubits capable of scaling to quantum advantage; if they cannot do this, the company's survival will be in doubt.
Quantum Research
This is my second article on IONQ and my sixth on quantum computing companies. In my previous IONQ article, I accepted the argument from IONQ that its AQ64 machine would deliver computational power beyond what is currently available, even from the largest supercomputers in the world. Things are changing quickly; Alphabet (GOOG) and International Business Machines (IBM) claimed to have delivered a quantum advantage in 2023. These claims were proven false within weeks (see my article on Microsoft (MSFT), where I covered this in detail).
The quantum articles I have published, and the quantum technology I attempted to explain are in this table. In this article, I will look at the progress of IONQ and consider a new qubit technology.
Date
Company
Technology Explained
June 2023
D-Wave
Qubits and the Adiabatic Algorithm
Jul 2023
Rigetti
Quantum Algorithms and Superposition
Aug 2023
IONQ
Entanglement and trapped Ion Qubits
Feb 2024
Microsoft
Topological Qubits and Majorana Zero Modes
Mar 2024
D-Wave
Quantum Annealing and Zapata AI
Quantum Supremacy the problem
A computer will have a quantum advantage when it can perform calculations far quicker than today's classical ones. It will also be able to solve specific problems that cannot be solved today.
A quantum computer will gain supremacy when it can run all of the existing quantum algorithms that will solve the more complex and critical problems we face. Quantum supremacy will be when a computer can solve problems that would take thousands of years on today's supercomputers. When it arrives, the world will be changed forever.
Companies have previously thought they had achieved quantum advantage; in all cases, it was later shown how a classical computer could solve the problem.
Basic idea
A quantum computer will develop an advantage in two strands. Firstly, it will run algorithms or computer programs based on linear algebra (now often called quantum mathematics). These algorithms manipulate the qubits, utilizing their superposition and entanglement properties before taking final measurements to answer the problem.
Algorithms and mathematics are vital parts of the improved performance of a quantum machine; however, these algorithms can run in simulated environments on today's classical machines and have proven to be very effective at solving many problems compared to other programming techniques.
The second strand is hardware; the computer's power will depend on its number of error-corrected qubits. The computer could run algorithms to solve some of the world's most intractable problems if it has enough. The number of qubits will take quantum computers beyond what can be simulated, and it has long been postulated that computers will require in the region of a million qubits to develop actual quantum supremacy.
A recent article in American Physics suggests that the algorithms used in classical computers are improving at a similar speed to the progress with quantum computers and that to develop a true quantum advantage (the ability to perform calculations beyond classical computers, regardless of the improvements being made) will require a machine with tens of thousands perhaps even millions of qubits.
Qubits, Scaling, and Error Correcting
I have followed at least ten technologies to produce the necessary qubits, each showing immense promise but failing when an intractable problem arrives. The company with the right qubit technology will be the first to make a working quantum computer with a quantum advantage, perhaps even quantum supremacy, and will be worth trillions of dollars; those pursuing what turns out to be the wrong technology will probably be worthless.
The technical field keeps changing; discoveries are constantly being made and new research published, and problems continue to arrive with existing approaches.
The Superconducting Brigade
This was the first technology out of the blocks, a superconducting qubit is a small loop of superconducting wire. It is controlled by more electrical wires and is connected to other qubit loops by even more superconducting circuitry.
This technology came to the fore when International Business Machines showcased their 5-qubit Canary computer in 2015; it took IBM another 8 years to get to the 100-qubit Eagle machine, and they released the 1,000-qubit Condor machine late last year.
Rigetti (RGTI), Alphabet (GOOG), and many others followed IBM into this technology, and it is easy to understand why. This type of circuitry is well known to these companies, and they have had great success with scaling similar ideas in the past; it is not that dissimilar to the silicon chip.
Superconducting qubits have proven to be highly problematic regarding error generation and correction. The slightest environmental change, and they forget the information they are meant to hold. They keep their state for less than a millisecond before they decohere (produce errors), so all calculations must be completed before that time. They are susceptible to heat, magnetism, energy, each other, and the quantum computer itself; they even forget their value when you measure them, so you cannot check their value without curtailing the calculation.
Two techniques have been developed to help with error correction: logical qubits (small groups of qubits that share the information rather than using a single qubit to hold it). The logical qubit must meet an error threshold of 99.5% to allow operations to be performed. The second technique is error-correcting code. The most widely used error correction technique is called "Surface Code." Using surface code, you would need over 300 qubits to error correct 1 working logical qubit. This redundant overhead means you will likely need a million qubit devices to produce meaningful computation.
In March of this year, IBM released a paper showing the potential for an error correction technique to reduce overhead. They call it the Gross Code (because it uses 288 qubits to correct 12 logical qubits). The new code is entirely theoretical and has not yet been applied, and will still require machines with tens of thousands of qubits to make meaningful progress.
Error correction now makes superconducting qubits look like a long shot; they were first, and several companies are still committed to them, but no one has yet solved the error correction problem, and there are significant doubts if it can ever be done.
Trapped Ion qubits
This was the next technology to go into commercial research; it is the technique of IONQ and Quantinuum (a merger between Honeywell Quantum Solutions and Cambridge Quantum Computing, currently privately funded by JP Morgan Chase).
The technology needed to control superconducting qubits was well understood; this was not true of laser optics needed to control trapped ions.
An ion is an atom with an electron removed, giving it a charge. That charge makes it relatively easy to trap in electrical fields (hence the name trapped ion computing). The ions are trapped in tiny vacuums at ultra-low temperatures; lasers are used to manipulate the information they hold.
IONQ has created a 36-qubit machine, the Forte; it is the only quantum gate computer available for sale and is consequently the world leader in quantum gate computing; it has the highest performance in error mitigation due to better coherence and higher gate fidelity.
IONQ machines are not bedeviled by the error correction problems of superconducting qubits, but they have another, perhaps equally intractable problem, scaling.
The problem is the number of qubits; 36 qubits is an order of magnitude less than what is needed to deliver quantum advantage.
Scaling trapped ion machines looked easier than superconducting qubits because you do not need the complex circuitry to connect all qubits as they can entangle naturally. However, they have limitations; firstly, they all have the same charge and repel each other. This limits the number of qubits you can put in any given space. They are usually arranged in either a straight line or a loop, making it very difficult to entangle one ion qubit with one 10 or 20 places down the row, and if you manage to scale up to the thousands of qubits, that becomes even more difficult.
IONQ Doing More With less
I think IONQ has accepted that its trapped Ion technology will not scale to deliver Quantum Advantage or Supremacy. As a result, they are concentrating on using their current technology to provide a commercial advantage while considering the possibility of buying a more scalable qubit technology.
In the IONQ Q4 earnings call (Feb 28, 2024), the CEO discussed the improvements in quantum algorithms, explicitly mentioning a quantum algorithm company that had "published research showing they could reduce a complex material simulation requiring 1.5 trillion gates down to requiring only 410,000. "
The CEO said in the same statement,
algorithmic work to find ways to do more with a smaller number of qubits is progressing at a much faster pace than the hardware itself.
The CEO also discussed how the existing IONQ technology could help in AI. He did not provide any concrete example of how IONQ is being used for AI but stated, "IONQ intends to pursue the artificial intelligence market." It is possible that quantum computers from IONQ and D-Wave could reduce the time and cost of training large language models for AI.
He said that IONQ will be able to produce machines with a commercial advantage in the next 2 to 3 years. I think it is telling that he talks of commercial advantage, not quantum advantage. This change in the company's language perhaps reflects the problems of IBM and Google when they thought they had achieved a quantum advantage.
The other major player in trapped ion quantum computing Quantinuum, is trying to solve the scaling problem by moving trapped ion qubits from a storage place to a compute zone where they entangle them and then move them back and bring in other ions. This is similar to classical computer architecture, where classical bits are transferred from memory to the cores to be processed before returning to memory. The technique brings in different problems, notably the shuffle time. The IONQ CEO said in a Morgan Stanley webinar
they do this because they can't do what we do because of patents…… they do shuttling, they take ions just like us, and they bring them into a compute zone and entangle them and then shovel them away and bring in other ions…..About 99% of the wall clock time is being used to shuttle back and forth.
He suggested that the time saved in moving to quantum computing was lost in the shuttling of ion qubits, making the IONQ system superior.
I think IONQ is already preparing for a new qubit technology. Their co-founders, Chris Monroe and Jungsang Kim, both world leaders in trapped ion technology, left the company towards the end of 2023 to return to their academic research.
During that discussion with Morgan Stanley, the CEO made some key remarks; he said,
if you ask me in 10 years or so I would actually choose a different modality…..You might see IONQ buys a different kind of qubit company. I don't see these as competition I just think they have different times when they will mature.
Putting these together, the founders and leaders in trapped ion qubit technology leaving, the discussion of commercial advantage, not quantum advantage, the search for new markets for their existing technology. A focus on doing more with fewer qubits. The CEO saying that he might in the future choose a different modality and IONQ might buy a different qubit technology makes me believe that IONQ has already accepted that their technology will not scale to the number of qubits needed for Quantum Advantage and they are looking toward life beyond trapped ion technology.
The New Technologies
Microsoft (MSFT) looked at all the technologies I have discussed and decided to pursue a wholly different and unique path. Microsoft is researching topological qubits based on Majorana Zero Modes. I covered the technology in detail in my Microsoft article. It has the advantage of being error-protected but the disadvantage of not being proven to exist. Microsoft is trying to invent something entirely new and has given itself a 20-year time frame.
In the Morgan Stanley webinar, the CEO of IONQ said he thought the Microsoft topological Qubits were "really elegant." he also said he hears about breakthroughs daily and noted it would take five years and a billion dollars to turn any breakthrough into a commercial product.
He did say he thinks IONQ is the only company in the market today and has several years to generate revenue before any competition arrives, which is probably true.
The Latest new tech: Neutral Atom Qubits
Neutral atoms seemed the least likely technology to make a clear difference a few years ago. Without a charge to hold onto, they seemed challenging to manipulate. However, recent breakthroughs from Harvard and MIT, plus the commercialization of private start-up companies like QuEra, make them appear the most likely technology currently being commercialized to deliver the required number of error-corrected qubits.
Trapping neutral atoms is difficult as there is no charge; they are held in fields of intense laser light, and atoms appear to stay wherever the light is brightest. Entangling neutral atoms is done by developing a Rydberg state, an enhanced energy state that allows atoms to entangle with others nearby. A laser delivers the energy and has been shown to enable atoms up to several micrometers apart to entangle. This level of interaction is easily achieved with optical lasers.
Neutral atoms have been shown to have long coherence times (up to 40 seconds) and have been scaled to a thousand qubits by a private company (Atom Computing). Researchers at Caltech have engineered a 6,100 qubit neutral atom array with a coherence time of 12.6 seconds; these results are far more than those achieved by the competing technologies of trapped ion or superconducting qubits.
In December 2023, three key researchers in this field Markus Greiner (Harvard), Vladan Vuletic (MIT) and Mikhail Lukin (Harvard) (Lukins 2020 paper first suggested Neutral atoms as qubits) published a paper in Nature. The paper is written in a way accessible to most people and includes several important points. They describe a quantum processor consisting of several distinct areas. Qubits are held in a reservoir as individual items and are combined in groups of 20 to make logical qubits. An entangling zone is used for parallel qubit encoding and gate operations, and a read-out zone for measuring qubits.
The qubits are Rubidium atoms trapped in optical tweezers, and these optical tweezers can manipulate the qubit while preserving coherence.
The developed machine had 280 qubits and 48 high-fidelity logical qubits, surpassing the current machine from IONQ. The 48 logical qubits had much higher connectivity than the IONQ machine, equivalent to a 7D hypercube rather than the single loop found with trapped ION qubits. The paper notes that the techniques described and the results found should allow the architecture to scale to 10,000 qubits. The Harvard Gazette suggests it is the first logical quantum processor and an essential step to a scalable quantum device.
The MIT/Harvard team behind this discovery has a spin-out called QuEra, which has made their 256 Qubit neutral atom computer ( Aquila) available on the Amazon (AMZN) Braket service. QuEra believes their new technology gives them a roadmap to a highly scalable architecture leading to quantum supremacy.
We have heard this many times, and the time will come when one of these technologies does deliver supremacy. I want to ensure I am invested in the company with the winning architecture, so I will continue researching developments.
IONQ Business Progress
As the CEO said, IONQ has this space for now. They are the only company selling quantum gate computers and have made notable progress.
Technical Road Map
The IONQ technical roadmap suggests an AQ number of 1024 by 2028. AQ stands for algorithmic Qubits and is IONQ's own measurement that others in the industry do not use. It is approximately equal to the number of error-corrected qubits a machine has.
A graph with numbers and text Description automatically generated
IONQ qubit roadmap (IONQ )
The roadmap suggests that the technology is now quite a way behind neutral atom technology, perhaps as much as 2 to 3 years. The road map includes commercial advantages from 2026 onwards but gives no idea of quantum advantages. IONQ has consistently hit the targets on its road map, usually early, and has had no significant setbacks.
They are building out the world's first quantum computer manufacturing facility and appear to be the only company selling quantum gate computers.
Financial Road Map
The growth of IONQ is impressive, and the following graph of IONQ bookings shows a very positive trend.
A graph of a number of people Description automatically generated with medium confidence
Bookings (Author Database)
Bookings for 2024 are forecast at $80 million, double that of 2023. Revenue is growing quickly, although not as fast as bookings. Revenue increased 100% in 2023, hitting $22 million; guidance was given of $39 million for 2024.
Costs are also increasing. Sales and marketing increased 118% in FY 2023 (189% Q4 year on year), and R&D increased 110%. Total operating expenses were up 86% over FY 2022.
There was an acceleration of costs in Q4 2023 over Q4 2022. Total operating expenses were 121% higher, suggesting that expenses are multiplying. Guidance was given for an adjusted EBITDA loss of $110.5 million in 2024, much worse than the 2023 figure of -$77.7 million. (all figures FY 2023 earnings call)
Using all of the guidance given and adjusting my mathematical model (presented in my previous IONQ article) for IONQ suggests that the company will not break even in my five-year forecast period.
A graph of a chart Description automatically generated with medium confidence
5 Year forecast (Author Mathematical Model)
There is little point in producing a fair value from this cash flow as it would not represent the value of IONQ (and likely be negative)
The Balance Sheet
At the end of 2023, IONQ reported shareholder equity of $485 million and cash of $456 million, sufficient to see them through 2027 (I expect total negative cash flow to be approximately $350 million 2024-2026).
In 2027, IONQ should have an AQ384 machine for sale, delivering a commercial advantage over existing classical computers and one more year of cash.
The Future
By 2027, which qubit technology will most likely produce a scalable quantum computer capable of achieving quantum advantage will hopefully be apparent. I do not think it will be the current trapped ion technology of IONQ; however, by 2027, IONQ will have built a significant user base and a good reputation for delivering working quantum devices. That reputation and sales channel could be worth a great deal of money if they had a quantum advantage to sell. The sales channel and IONQ might be worthless if someone else has the quantum advantage.
If IONQ intends to continue with its current plan of developing a commercial advantage and selling its current and forecast computers in the AI field, as suggested by the CEO, or to data centers and other major computer users, then I would need to see a significant reorganization in the company for it to achieve profitability.
IONQ growht metrics
Growth in the Income Statement (Author Database)
R&D costs are doubling yearly, and SG&A would have to increase further if IONQ looks to expand its sales operations. CAPEX must rise as they build out their first manufacturing facility.
The cost of goods sold is outpacing revenue, pushing gross profit negative in 2023. IONQ computers sell for around $18 million each (covered in my first article); with significant cost-cutting and an enormous reduction in the cost of goods sold, it might be possible to profit with unit sales in the hundreds. Of course, this has a huge caveat: if another manufacturer delivers a quantum advantage, selling significant numbers of small quantum machines will be near impossible.
Conclusion
The technology in this field is changing quickly; IONQ chose to work with trapped ion qubits, which looked like the best option back then. The founders have left, and it seems likely that the technology will not scale sufficiently to attain quantum advantage or supremacy.
Microsoft, QuEra, and Atom are researching new technologies, and neutral atoms benefit from the extensive research performed by world-leading physicists at Harvard and MIT.
At the moment, neutral atoms seem a better bet, but we have seen this stage before with both superconducting qubits and trapped ion qubits. In both cases, there were unforeseen and possibly insurmountable problems.
IONQ continues to build its reliable quantum devices and remains the only player in this field. They likely have several years where they will be the only supplier and are building a reputation, sales channel, and customer base. The CEO has already said that we may see IONQ buy a different qubit technology. If they do, they will have built a formidable sales channel that will likely make them one of the world's most influential companies.
If they cannot acquire a qubit technology that will scale in a timely and error-corrected manner, it isn't easy to see how they will prosper.
To upgrade IONQ to a buy rating, I would want to see a clear path to quantum advantage. This could be achieved by IONQ showing how it can scale its current technology to tens of thousands of qubits or announcing the purchase of scalable new qubit technology.
mitdiskutieren » 
USA
