Quantum networking and AI upgrades

Welcome back to the Circuit Breaker, where you can find the best recaps on the latest innovations in AI, quantum computing, semiconductors, and more, from across IBM Research and beyond.

Week of November 17 - 21

• IBM and Cisco's quantum networking plans
• Accelerating AI inference with IBM Storage Scale
• Making open infrastructure for AI a reality

The network effect: Scaling quantum computing with Cisco

The Quantum team at IBM has been on a tear recently, announcing major breakthroughs in quantum algorithms with partners, new hardware and software that will power the future of quantum computing, and how the world will achieve quantum advantage. But what comes after that?

🌐 Linking it all together. This week, we announced our partnership with Cisco to explore how to link quantum computers together for a future where distributed quantum computing is a reality, much like cloud computing today. The advantages of being able to run workloads across the world, based on capacity and necessity, are clear. But getting there will take a lot of work.

🗺️ On the road ahead. The first milestone to be reached within five years will be when we're able to successfully entangle a pair of cryogenically separated quantum processors — meaning two processors that are physically isolated from each other — which the team hopes to achieve in the next five years.

💡 Past knowledge will inform future breakthroughs. Much like the future of quantum-centric supercomputing will undoubtedly involve traditional CPUs and GPUs along with quantum processors, quantum networking will require a deep understanding of traditional network theory and how digital information moves. Entangled processors will act as a single entity, mathematically speaking, unlike distributed classical devices. With Cisco, IBM plans to explore how to build transducers and optical links between QPUs.

Learn more about the research here

Accelerating AI inference with IBM Storage Scale

GPUs may be the stars of AI infrastructure, but network and storage are the unsung heroes. Without the right ones, today’s AI workloads would be slow and costly. This week at Supercomputing 2025, IBM Research showed how high-performance storage can transform LLM inference at scale.

⚖️ IBM Storage Scale as a KV cache tier enables sharing across hundreds or even thousands of GPU servers, cutting redundant computation and accelerating inference.

📊 Experiments with Llama3-70B, served by vLLM on four H100 GPUs, showed up to 12x faster time-to-first-token compared to recomputing, with latency reduced from 19 seconds to less than 2.

🔗 Seamless integration with frameworks like vLLM and llm-d delivers enterprise-ready performance, scalability, and reliability—right out of the box. Storage isn’t just supporting AI anymore, it’s accelerating it.

Click here to read the full story

Recap: AI Hardware Forum 2025

The future of AI won’t be built by one company alone, it will be forged through collaboration. This was the theme of IBM Research’s 6th Annual AI Hardware Forum, where researchers came together with partners in academia and industry to showcase how open software that supports a heterogeneous hardware ecosystem is reshaping the AI landscape.

This year the Spyre AI accelerator launched for IBM z17 and Power11 systems, delivering secure, reliable performance for generative AI workloads. Supporting this hardware is an open software stack that's opening new possibilities for a multi-accelerator world.

Multiple discussions focused on the power requirements of AI inference, including the massive cost of moving data. To this end, IBM's breakthrough in co-packaged optics brings densely packed high-speed fibers to the edge of the chip, and is poised to offer an 80x bandwidth boost, revolutionizing datacenter efficiency and reducing energy demands.- 

We also announced the next evolution of its partnerships with the University at Albany and the National University of Singapore , fueling new projects in NLP, cancer research, weather modeling, and sustainable urban development.

➡️ Learn more about AI Hardware here

Smart Talks with IBM Research

Malcolm Gladwell and Jay Gambetta discuss IBM’s plans to scale quantum computing power, the groundbreaking experiments already underway and what impact these new computers could have on chemistry, medicine and even finance.

🎙️ Watch the full conversation here:

Pushing the boundaries of semiconductor technology with the University of Dayton

Deploying IBM Generalist Agent in Enterprise Production

Highlighting new publications from IBM researchers that we liked the sound of:

• Quantum resources in resource management systems
• Classifier-Augmented Generation for Structured Workflow Prediction
• BenchmarkCards: Standardized Documentation for Large Language Model Benchmarks

For more on the latest news, be sure to follow IBM Research on LinkedIn. And if you want to go even deeper, subscribe to our monthly newsletter, Future Forward for more on the latest news on breakthroughs in AI, quantum computing and hybrid cloud.