It’s been a busy week for IonQ, the quantum computing start-up focused on developing trapped-ion-based systems. At the Quantum World Congress today, the company announced two new systems (Forte Enterprise and Tempo) intended to be rack-mountable and deployable in a traditional data center. Yesterday, speaking at Tabor Communications (HPCwire parent organization) HPC and AI on Wall Street conference, the company made a strong pitch for reaching quantum advantage in 2-3 years, using the new systems.

If you’ve been following quantum computing, you probably know that deploying quantum computers in the datacenter is a rare occurrence. Access to the vast majority NISQ era computers has been through web portals. The latest announcement from IonQ, along with somewhat similar announcement from neutral atom specialist QuEra in August, and increased IBM efforts (Cleveland Clinic and PINQ²) to selectively place on-premise quantum systems suggest change is coming to the market.

IonQ’s two rack-mounted solutions are designed for businesses and governments wanting to integrate quantum capabilities within their existing infrastructure. “Businesses will be able to harness the power of quantum directly from their own data centers, making the technology significantly more accessible and easy to apply to key workflows and business processes,” reported the company. IonQ is calling the new systems enterprise-grade. (see the official announcement.)

Snapshot of the new systems:

• “IonQ Forte Enterprise brings quantum computing to modern data centers: With a target performance of #AQ 35, IonQ Forte Enterprise is expected to further IonQ’s lead as the provider of the most powerful, commercially available quantum computer in the world. IonQ Forte Enterprise is designed for complex computational problems, including process optimization, quantum machine learning, correlation analysis, and pattern recognition. With today’s announcement, IonQ is streamlining these capabilities into a compact form factor that can be easily deployed across existing data center infrastructures.
• “IonQ Tempo enables commercial advantage capabilities for the most demanding use-cases: IonQ has revealed for the first time new details of its highly anticipated #AQ 64 enterprise-grade system, IonQ Tempo. Tempo is anticipated to be a commercial advantage system capable of delivering substantial business value for today’s use cases. An #AQ 64-based Tempo system would far exceed what can be simulated with classical computers and GPUs, and provide a computational space 536 million times larger than even IonQ Forte Enterprise, an astonishing leap in computational power.”

IonQ reported its Forte Enterprise system will be introduced in 2024 and the Temp on 2025.

Speaking at Tabor’s HPC and AI on Wall Street conference, Philip Farah, VP strategic partnerships, and Bob Fletcher, director, enterprise sales, jointly reviewed IonQ’s progress and plans for delivering quantum advantage. It’s thought that financial services will be among the first industry sectors to deploy quantum computers.

“Our systems enable enterprises to start targeting use cases, to test them and figured out which ones have more potential. We’re not yet at the point where you can say, all right, we have quantum advantage in these use cases, and that we can start scaling them. But we believe we’re about two years away. The areas that seemed the most promising, machine learning, optimization, simulation cryptography all come to mind. And in terms of industries, financial services is one of the key sectors – banking, insurance, and capital markets – and is probably the top industry that we’ll see [quantum] advantage coming from [first],” said Farah.

There are, of course, many qubit modalities in development. Trapped ions are among the more mature approaches and have several strengths including long coherence times, fewer manufacturing issues because the ions themselves are naturally identical, and don’t require the extreme cold, for example, that superconducting qubits do. The latter require expensive dilution refrigerators to operate, which is one factor that complicates placing such systems in datacenter. That said IBM’s System One is a fully enclosed system and has been located at several sites.

(Brief ion trap technology primer at the end of article)

No one is arguing quantum computers are ready for production environments now – they remain too error-prone, often difficult maintain, typically aren’t larger enough (number of qubits) for practical applications. Many think it will take a decade to produce fault-tolerant machines sufficiently large (1000s to millions of qubits) to be practical for production use.

IonQ characterizes the number of qubits in its devices by a metric called algorithmic qubits – a metric based on work by the Quantum Economic Development Consortium. IonQ’s metric bakes in several elements, including performance on some standard tasks, as well as circuit depth and width. The forthcoming systems will have 35 and 64 algorithmic qubits, respectively. (Link to IonQ explanation of algorithmic qubits)

Even then, the consensus is quantum computers will be used for select problems and act as accelerators working with classical computers. Pieces of the problem will be solved by quantum computers and pieces of the problem will be solved by classical computer.

IonQ embraces this hybrid model, but contends that sufficiently reliable quantum computers, even with a relatively modest number of ‘algorithmic’ qubits, will be able to deliver quantum advantage for some applications. Farah’s pitch to the Fintech audience was “don’t wait” to start exploring quantum computing.

“There’s a study from BCG (Boston Consulting Group) looking at the potential of early adopters; it [argues] the 10% early adopters capture 90% of the value created [by a new technology]. Why is that? That three main reasons why the early adopters will benefit,” said Farah.

• #1. Building quantum muscle memory. “What I mean by this is, number one, getting people on your teams to know how to build quantum algorithms. Quantum algorithms are different from classical; you have to rethink the problem whether it’s quantum Fourier transformations or whether you’re trying to find a minimal energy state that gives you the answer. It takes time to learn how to build a quantum algorithms. The good news is, you can take your existing AI and machine learning experts and take them through this program through this training. We do it on a regular basis with our clients,” according to Farah.
• #2 Find compelling use cases. “The second part is building and investigating what use cases are going to take most advantage of quantum. It’s hard to say from the get-go. You can’t just build the matrix, and say, based on these 10 criteria, those are top use cases. You have to go build them, and while you’re building them, figure out which ones have the most advantage potential. Then decide which ones you want to start integrate into a hybrid application. So the journey from use case identification, prioritization, building quantum algorithms, integrating into your application, and then from an IT perspective, as you go closer to production optimization, in making them production ready, this journey, takes between 18 to 36 months. I think the clock is ticking now, because within 18 to 36 months, we’re going to be at the position where some of these algorithms will have proven quantum advantage,” he said.
• #3 Hardware access. “Finally, the third leg of the stool is access to hardware. Today, we take it for granted that there is hardware accessible when we need it. My sense is when we hit the inflection curve, the demand will outstrip the ability of hardware providers to provide the capacity needed. And to add to that, it’s not just because of [hardware availability, but also, it’s important to have people who understand the hardware. That’s the interesting part of it. If you’re writing a quantum algorithm today, I would argue you don’t care who’s going to win on the hardware machines because 80% to 90% of the logic of the algorithm, let’s say it’s a gate-based, will apply to any of the hardware providers. The other 10 to 20% [is] optimization of your algorithm to work on the specific hardware machine. So, worst case scenario, you throw the 10%, you keep the logic, and then you optimize it for a different type of machine. The 80% stays with you,” said Farah.

Currently, most quantum algorithms development is done on GPU-accelerated classical systems which work reasonably well for low-qubit counts. There are even a few companies – Terra Quantum is one – who say they can effectively simulate full physical qubits on classical system today and deliver meaningful results.

Rising qubit-counts, said Fletcher, will soon overwhelm these simulators’ capabilities.

“Remember, I said each qubit on our system can encode 10,000 different values [i.e. states]. As we add more and more qubits, when we go to 35 to 64, there’s going to be a big watershed. So, an Nvidia DGX box can simulate somewhere in the order of 30 or so qubits today. If you take 256 GPUs, maybe, A100s maybe H100s, they can go up to 35 qubits. That’s great. You got the simulator that’s doing a fabulous job. And those aren’t free,” said Fletcher.

Sometime in the 2024-2025 timeframe, says Fletcher, the simulators will no longer be able to keep up with IonQ’s qubit- count.

“There’s this big dichotomy between 2024 and 2025, where the simulators can’t simulate the full-scale of the quantum computers. What does this mean? Well today, when you look at algorithmic development environment, we’re seeing about 40%-45% done on simulators with no noise introduced, and another 40%-45% using a simulator that has a noise footprint of the QPU that you’re going to use. Then having got your algorithm developed using both of those and the error mitigation, you finally run it on the QPU and verify that everything is working fine,” said Fletcher.

“Two years from now, it’s going be a completely different ballgame, because you will not be able to use a simulator for that. So, we’re going to see much, much smaller percentage of time being used in the simulation domain for [algorithm development], and a much higher percentage of time being used on actual QPU use,” he said.

Stay tuned.

Ion Q Trapped Ion Technology Primer (excerpted from earlier HPCwire reporting)

Start with atoms of the right size and valence electron structure, strip off an electron to create the ion, and load ions into an evacuated chamber. IonQ presently uses ytterbium (Yb⁺) ions but plans to switch to barium (Ba+). Magnetic fields are used to confine the ions in a line. Lasers are used to cool the ions and limit their jiggling. The ions act as qubits. Lasers are also used to excite the ions into the desired state. Accomplishing all of this, as you might imagine, requires precision engineering and optics but steers clear of the big dilution refrigerators required by, for example, superconducting qubits.

The most current IonQ QPU (Forte) has 32 physical qubits and uses an acousto-optic deflector to precisely direct laser beams to target individual ions. This is instead of an earlier fixed path laser addressing mechanism. IonQ says, “With the aid of precision lasers, these trapped ions can then be entangled using high-quality two-qubit quantum gates between any pair of qubits in the chain. This approach is the core technology behind IonQ’s quantum computers. We believe that the all-to-all connectivity within a core, enabled by the long-range Coulomb interaction (the force between two stationary, electrically charged particles) is a significant and unique advantage of IonQ’s processor architecture. It enables highly efficient realizations of quantum algorithms with arbitrary internal structure.”