IonQ CEO, Peter Chapman, has written a letter to stockholders today which we would like to share with our community of customers, developers, and followers. His eight-point plan describes how we plan to leverage our technology, finances, and team in order to power the next industrial revolution through quantum computing.
The letter can be found in our 2021 Annual Report, and below for convenience. We invite you to read the letter and to follow along in the coming months and years as we demonstrate concrete action on our roadmap.
We believe the 21st century will be powered by quantum computers. We expect these machines will open the doors to an amazing new era where “smart devices” are actually smart, computers can program themselves, batteries are light and powerful enough to enable flying cars, society is able to pause or even reverse climate change, we have cures for diseases such as cancer, Alzheimer's and diabetes, and with luck, reverse aging itself (which at my age, can’t come soon enough).
We see a massive opportunity in the large markets we are targeting. We believe the Total Addressable Market (TAM) for quantum computing is likely larger than the existing classical hardware and software markets combined. As a test of this theory, what would general artificial intelligence1 be worth alone?
IonQ is not alone in this belief; we compete against some of the largest technical companies and even the largest countries in the world. And despite that--amazingly--our little company leads the world today with the best2 commercially available quantum computers. We have a window of opportunity to establish IonQ as the industry commercial standard in this space. When you hear, “the promise of quantum computing is 10 or more years away,” we do not agree, as we, fortunately do not believe that to be true for IonQ. Our intent is to move quickly to solidify and extend our current leadership position in the marketplace. This strategy is not without challenges; it requires significant investment and focused execution.
Some readers may not be aware of our plans for quantum and how we plan to power the next industrial revolution with IonQ at the forefront. Here is our eight-point plan to get there:
1. Build a quantum computer with 35 algorithmic qubits3
As you have heard us say, the power of a quantum computer is not a function of the number of physical qubits, but instead, the number of “useful qubits,” or what we call “algorithmic qubits.” We recently announced our latest IonQ Aria quantum computer with a record-breaking 20 algorithmic qubits. Our goal is to get to 35 algorithmic qubits in a single quantum computer without error correction. The 35 algorithmic qubit milestone is key for unlocking many of the most important near-term applications of quantum computing, such as quantum machine learning. We believe the way to get there is by reducing native error rates.
Every quantum computation goes through three steps where errors can occur:
Quantum State Preparation, where we initialize the qubits;
Computation, where we run an algorithm made up of a series of quantum gates; and
Measurement, where we read out the final state of each qubit after the calculation (also known as “state detection” or “readout”).
We recently announced4 that our new barium-based quantum computer achieved an amazingly low SPAM error of 0.04%. SPAM errors, or State Preparation And Measurement errors, refer to sources of error #1 and #3, described above. This SPAM error rate could achieve more than 2,000 algorithmic qubits if we could ignore errors related to the computation or quantum gates (#2 above). In other words, we have already fixed two sources of error necessary to achieve our 35 algorithmic qubit goal.
As for the error related to quantum gates, we recently announced5 that, by using barium qubits in our systems, the fundamental limit for two-qubit gate errors is closer to 0.02% (as compared to 0.1% for ytterbium). Although this improvement might seem small, this meaningful difference should allow us to achieve 35 algorithmic qubits and beyond. It might not seem like much, but as you approach an error of 0%, small increases like this have an outsized impact.
2. Supersede Moore’s Law and create a new price-performance ratio of quantum computers
In 1965, Gordon Moore published a paper in which he observed that the number of transistors in computer chips was doubling roughly every one to two years. Related to Moore’s Law is the price-performance ratio of computers. Loosely stated, a given amount of money will buy twice as much computing power two or three years from now. Or more simply stated, when you buy a new computer, you expect it to be two times better for roughly the same price.
In quantum, every time you add a single algorithmic qubit, you double the computational power of the machine. Similar in spirit to Moore's law, if we added a single qubit every year, you should expect to pay the same amount as today.
Our roadmap reflects qubit growth at a much faster rate with 35 algorithmic qubits expected in 2024 to 64 algorithmic qubits expected in 2025, the latter of which is a machine that is 229 or 536,870,912 times more powerful.
What would you expect to pay for this machine? Clearly the price of the quantum computer cannot scale to the same degree as the computational power it delivers.
As much as we are focused on increasing the computational capabilities, we are similarly focused on reducing the price per algorithmic qubit. This year, we are building up our production engineering team, an effort that we hope will put us on a trajectory to reduce the size and weight of our future quantum computers. We are also making technical investments in photonics technologies used in the telecom industry to further shrink and integrate our system components.
As we put more of the system on a chip and shrink the quantum computer, we expect our cost per qubit to shrink and the performance to rise. This is also why we believe that building a quantum computer that can operate without the need for near absolute-zero temperatures is so important. We will continue to invest in innovation that realizes the economic balance that kept Moore’s law going: exponentially more powerful quantum computers available at modestly increasing prices.
3. Network multiple quantum computers together to create incredibly powerful quantum computers
No matter what technology you use for a qubit, sooner or later you can’t get more qubits on a chip to get to scale. At a certain point, every quantum computing manufacturer will need to network several systems together.
In classical computing, networking is enabled by communication.
In quantum computing, networking allows entanglement of qubits across different quantum computers. Qubits don’t differentiate between being entangled together on a chip and being entangled across chips. The qubits can be next to each other—or miles apart. And once they are all networked, they work together as a single quantum computer.
As I mentioned, we recently announced a new IonQ system using barium qubits. Barium allows us to leverage the same lasers and silicon photonics technology used by telecom and data center networking devices. We are now working on adding networking to our quantum computers, and we will know we are successful when we have a network that operates at room temperature using a simple fiber optic cable. This is work we first envisioned in 2007, when our co-founder, Chris Monroe, demonstrated how to network two ion trap systems together with a technology called photonic interconnects6 . Today, we are productizing that experiment and expect to incorporate it at commercial scale to build increasingly powerful quantum computers.
Our quantum computers allow any qubit to entangle with any other qubits; we call this “all-to-all” connectivity. And with an optical switch between the quantum computers, we believe the “all-to-all” connectivity of IonQ’s qubits can be maintained across all the machines in the quantum network.
4. Implement low overhead error correction
As you increase the number of algorithmic qubits in a quantum computer, you need to decrease their error rates correspondingly. You can do that either by reducing the native error rate (described in #1, “Quantum State Preparation,” above) or by introducing error correcting codes to make up for the loss in fidelity.
The overhead of the error correcting code is a function of the error you are trying to overcome. In 2020, IonQ staff, Duke University and the University of Maryland demonstrated the world's first error corrected qubit with an outstandingly low overhead of only 13:17 . This innovation overturns the conventional wisdom that error-corrected qubits will require thousands, if not millions, of physical qubits to realize.
By networking multiple systems together, we hope to increase the number of physical qubits so we can start to introduce error correction and increase the resulting number of algorithmic qubits. We have proven our ability to increase the number of physical qubits on a chip within a single system and, in the future, anticipate increasing that number to maybe hundreds, or even thousands of qubits. We expect that the combination of networking and the number of qubits on a single chip will allow us to scale to much more powerful quantum computers. You can see why integration, and the benefits it creates in scaling, fidelity and cost, is so important.
5. Increase gate speeds
We expect that “all-to-all” connectivity, low error-correction overhead and likely new N-Qubit gates8 will dramatically hasten the speed of an entire quantum computation. But it also helps to improve the speed of individual gates. Ion trap gate speeds are currently slower compared to other technologies, but that's not as large of an issue because ion qubits last (coherence time) far longer than the manufactured qubits used by many of our competitors. Furthermore, gate speeds on trapped ion systems are a function of the laser power one applies to the qubits.
We recently announced9 a new quantum computer based on qubits that use barium. Barium qubits use a laser in visible light that is much easier on optics and, as a result, should allow us to increase our laser power, significantly improving gate speeds.
In the future, improved gate speeds will be required to allow large quantum circuits to be run, but current customer needs are not complex enough or long enough for gate speed to be a blocker to commercialization.
6. Create the next generation of machine learning and chemistry applications
We are working to increasingly invest in applications that are stepping stones to our larger goals. For instance, our recent announcement10 with Hyundai on battery chemistry that works toward better and lighter batteries. Our primary focuses this year will be in machine learning, chemistry and finance applications.
We expect to release libraries and web service endpoint APIs that allow application developers to access early quantum computers for inclusion in their applications.
This year, we also expect to significantly increase our investment in the second generation of our operating system, compilers and language tools to support application developers.
7. Invest for the long term
We believe that a fundamental measure of our success will be the stockholder value we create over the long term. Becoming a self-sustaining business is a priority in order to capture the ever expanding TAM.
In the short term, we will focus our investments in a way that we hope will bring a quantum computer with quantum advantage to market. At this stage, we are a front-runner in a space race for long-term leadership in the quantum computing industry and speed is critical to achieving the potential of our business model. However, we will maintain our emphasis on long-term profitability and capital management.
As with many fast-growing technology companies, we expect that almost all of our free cash flow will be invested back into research and development in order to improve performance, increase integration, drive down cost and bring follow-on products to market as fast as possible. When someone buys time on one of our quantum computers, they are helping pay for development of the next generation of quantum computers, which we expect will be more powerful and also ultimately benefit from cost efficiencies relative to current quantum computers.
We will work hard to spend wisely and maintain a lean culture that preserves capital. We understand the importance of continually reinforcing a cost-conscious, ROIC-centric culture.
8. Attract, develop and retain a world-class team
We know our success will be greatly affected by our ability to attract and retain a versatile, talented, diverse, inclusive, and motivated employee base, each of whom must think and act like owners, which is why we plan to prioritize compensating employees primarily with equity. We are pleased with our momentum in growing our talent base since becoming a public company.
You may ask if, in the future, IonQ will be known as the company that makes the best quantum computers? I prefer to envision us as the company that is building the technology that makes everything else possible. Look for the “Powered by IonQ” sticker on batteries, self-driving cars, and vaccines of the future.
Lastly, on behalf of myself and the IonQ team, I want to express my gratitude for your support last year as we became the world’s first public quantum computing company. We are proud to be trailblazing for our industry, and would not be here without your enthusiasm for IonQ.
President and Chief Executive Officer