- IBM and AMD teamed to try to create a software orchestration layer for hybrid quantum-classical compute systems
- The trick is sending the right workloads to the right kind of compute
- But the task is even trickier because GPUs and QPUs don't speak the same language
Quantum processing units (QPUs) and GPUs have the potential to be the all-star tag team in the supercomputing arena. But there’s one big problem: they don’t speak the same language. So, engineers from IBM and AMD are stepping in to create a system that lets each know when to tag in for different workloads. The goal? Combine the superpowers of QPUs and GPUs to solve some of the world’s toughest computational problems.
As part of a newly announced collaboration to solve this problem, IBM is bringing its quantum hardware to the table while AMD is bringing its GPU and CPU chips and software engineering know-how.
Ralph Wittig, who runs AMD’s advanced research team, told Fierce that while GPUs are good at linear algebra problems, quantum computing is great at solving correlation problems. Certain workloads can contain both types of problems. The trick is breaking those workloads into smaller chunks and sending those pieces to the best type of compute for each task.
“You need software that orchestrates that optimally,” Antonio Corcoles, Principal Research Scientist at IBM Quantum, explained. “We don’t need to invent new tools, we need to give the existing tools in the community enough context that they can deal with this new computing resource.”
If that sounds like it’ll involve quite a bit of refactoring, you’d be right. That’s especially true for AI and ML models which will need to be reworked to account for the quantum element, Wittig said.
There’s also the hurdle of translating between the language GPUs and CPUs in the digital domain and the qubit-based tongue of QPUs.
“It’s not a simple evolution, there’s different math,” Corcoles said.
Quantum for the masses
IBM and AMD plan to work through all of these issues together and are aiming to debut a hybrid quantum-classical computing demo before the end of this year.
Asked when the market can expect to see commercial deployments of quantum computers in data centers, Corcoles said those already exist – just not at scale.
Why? The technology has yet to hit an inflection point. Though much of 2024 was focused on reducing error rates and noise to enable quantum tech to scale, there’s still one other big ingredient missing: efficiency.
“What we don’t have is quantum advantage. What we are exploring is how to get to the point where using the QPU will give you some tangible difference in cost or energy or in time to solution or in accuracy,” Corcoles said. “This is what we are anticipating over the next year or so to get to.”
Investors have noticed the market is on the cusp of a turning point. Investment in quantum startups grew to $2 billion in 2024, up from $1.3 billion in 2023, according to McKinsey & Co. Among government entities, Japan, the U.K., Germany and the U.S. are leading quantum investments.
McKinsey predicted quantum computing revenue will grow from $4 billion in revenue in 2024 to between $28 billion and $72 billion in 2035.
“The inflection point will be when, with the use of QPU in the mix, your algorithms become more efficient,” Wittig concluded. “Then very quickly you’re going to see that you’re going to be able to scale to solve problems that you couldn’t solve before…that’s exactly the same inflection that also happened when the GPU was discovered as a computing fabric for machine learning.”
AI in telecom isn’t tomorrow’s problem — it’s today’s imperative. If you’re waiting for it to catch up, you’re already behind. Join Fierce Network, Supermicro and Nvidia for an exclusive Accelerating AI Solutions for Telecom event at the Dallas Cowboys HQ. Click here to check your eligibility and register today.