Quantum Connectivity Solutions for Computing Systems

Check out the latest from MIT EQuS and Lincoln Laboratory published in @NaturePhysics! In this work, we demonstrate a quantum interconnect using a waveguide to connect two superconducting, multi-qubit modules located in separate microwave packages. We emit and absorb microwave photons on demand and in a chosen direction between these modules using quantum entanglement and quantum interference. To optimize the emission and absorption protocol, we use a reinforcement learning algorithm to shape the photon for maximal absorption efficiency, exceeding 60% in both directions. By halting the emission process halfway through its duration, we generate remote entanglement between modules in the form of a four-qubit W state with concurrence exceeding 60%. This quantum network architecture enables all-to-all connectivity between non-local processors for modular, distributed, and extensible quantum computation. Read the full paper here: https://lnkd.in/eN4MagvU (paywall), view-only link https://rdcu.be/eeuBF, or arXiv https://lnkd.in/ez3Xz7KT. See also the related MIT News article: https://lnkd.in/e_4pv8cs. Congratulations Aziza Almanakly, Beatriz Yankelevich, and all co-authors with the MIT EQuS Group and MIT Lincoln Laboratory! Massachusetts Institute of Technology, MIT Center for Quantum Engineering, MIT EECS, MIT Department of Physics, MIT School of Engineering, MIT School of Science, Research Laboratory of Electronics at MIT, MIT Lincoln Laboratory, MIT xPRO, Will Oliver

IBM Successfully Links Two Quantum Chips to Operate as a Single Device Key Insights: • IBM has achieved a significant milestone by linking two quantum chips to function as a single, cohesive system, enabling them to perform calculations beyond the capability of either chip independently. • This accomplishment supports IBM’s modular approach to building scalable quantum computers, a strategy aimed at overcoming the limitations of single-chip architectures. • The linked chips demonstrated successful cooperation, marking a step closer to larger and more powerful quantum systems capable of addressing complex real-world problems. The Modular Quantum Computing Approach: • IBM employs superconducting quantum chips, manufactured using processes similar to traditional semiconductor technology, allowing scalability and integration with existing hardware infrastructure. • Modular quantum systems involve linking smaller quantum processors, rather than relying on a single massive chip, reducing fabrication challenges and improving scalability. • This architecture allows multiple chips to share quantum information seamlessly, paving the way for constructing larger quantum systems without exponentially increasing hardware complexity. Addressing Key Challenges in Quantum Computing: • Scalability: Connecting multiple chips is a critical step toward scaling quantum computers to thousands or even millions of qubits. • Error Reduction: Larger quantum systems increase susceptibility to errors. Modular architectures provide pathways for better error management and correction across linked processors. • Coherence Across Chips: Maintaining the delicate quantum states across separate chips is technically challenging, and IBM’s success suggests progress in solving this issue. Implications of IBM’s Achievement: • Enhanced Computational Power: Linked quantum chips unlock the potential for more complex simulations and problem-solving capabilities. • Practical Quantum Applications: Industries like pharmaceuticals, cryptography, and materials science may soon benefit from more robust and scalable quantum computing solutions. • Competitive Advantage: IBM’s progress underscores its leadership in modular quantum computing, positioning it strongly in the competitive quantum technology landscape. Future Outlook: IBM’s successful demonstration of inter-chip quantum communication validates the modular quantum computing strategy as a viable path to scaling up systems. Future advancements will likely focus on enhancing chip-to-chip communication fidelity, increasing the number of interconnected chips, and reducing overall error rates. This breakthrough brings us one step closer to practical, large-scale quantum computing systems capable of solving problems previously deemed unsolvable by classical computers.

From NVLink to NVQLink: Wiring Quantum Processors into AI Supercomputers NVIDIA just unveiled NVQLink - an open interconnect + software stack that tightly couples quantum processors (QPUs) with AI supercomputers for real-time hybrid workflows like calibration and quantum error correction (QEC). It's not a quantum computer from NVIDIA, it's the missing fast path between QPUs and today's accelerated systems so the two can work as one. ✅ What is NVQLink exactly? A hardware + software integration path that links QPUs to NVIDIA GPU/CPU systems with low-latency, high-throughput data movement and real-time control via CUDA-Q (formerly CUDA-Quantum). Performance (NVIDIA-stated): up to 400 Gb/s GPU↔QPU throughput and <4 μs minimum round-trip latency in a reference (FPGA→GPU→FPGA) loop, sized for fast feedback tasks like QEC decoders and calibration. ✅ Why do we need NVQLink? Quantum isn't standalone: to be useful, QPUs depend on classical compute for: 🔹 Calibration and drift tracking, 🔹 Real-time QEC decoding and control, 🔹 Logical program orchestration (dynamic routing, lattice surgery, just-in-time compilation). All three are latency-critical control loops. NVQLink provides the speed/scale so GPUs can run these loops in real time while QPUs stay coherent. NVIDIA's message is hybrid is the future: supercomputers + QPUs co-evolve. quantum doesn't replace GPU systems. ✅ How does NVQLink work? 🔹 A QPU (the quantum chip) is driven by nearby control electronics that send precise pulses and read measurements. 🔹 NVQLink is the fast lane between that controller and the GPU, so results from the QPU reach the GPU in microseconds and new commands go back just as fast. 🔹 CUDA-Q is the programming layer: you write one hybrid program where the QPU does the quantum steps, and the GPU does the heavy classical math (like error-correction and optimization). 🔹 Inside the AI node, NVLink/NVSwitch connects GPU↔GPU at very high bandwidth. NVQLink connects QPU↔GPU for tight, real-time control. ✅ Where does it fit inside today's GPU systems? In a Blackwell/NVLink-5 cluster (or CPU+GPU nodes), GPUs already share data over NVLink/NVSwitch at TB/s. NVQLink brings the QPU/control side into that world: measurement results flow quickly to GPUs. GPU decoders/control kernels send decisions back within microseconds, the rest of the AI stack (simulation, scheduling, ML-based decoders) runs on the same accelerated node. Think of NVQLink as the southbridge to quantum: it's the tight, deterministic path between the quantum device and the GPU side where the heavy classical algorithms live. Nvidia NVQLink: https://lnkd.in/gYr4xZk3

Is this the "Attention Is All You Need" moment for Quantum Computing? Oxford University scientists in Nature have demonstrated the first working example of a distributed quantum computing (DQC) architecture. It consists of two modules, two meters apart, which "act as a single, fully connected universal quantum processor." This architecture "provides a scalable approach to fault-tolerant quantum computing". Like how the famous "Attention Is All You Need" paper from Google scientists introduced the Transformer architecture as an alternative to classical neural networks, this paper introduces Quantum gate teleportation (QGT) as an alternative to the direct transfer of quantum information across quantum channels. The benefit? Lossless communication. But not only communication: computation also. This is the first execution of a distributed quantum algorithm (Grover’s search algorithm) comprising several non-local two-qubit gates. The paper contains many pointers to the future, which I am sure will be pored over by other labs, startups and VCs. I am excited to follow developments in: - Quantum repeaters to increase the distance between modules - Removal of channel noise through entanglement purification - Scaling up the number of qubits in the architecture Amid all the AI developments, this may be the most important innovation happening in computing now. https://lnkd.in/e8qwh9zp

Today we introduced a new reference architecture for quantum-centric supercomputing, outlining how quantum processing can be integrated directly alongside modern high-performance computing systems. With our partners, we are now seeing hybrid quantum-classical workflows reaching parity with leading classical methods on real problems. Preparing for this quantum-classical future means building infrastructure where quantum resources plug naturally into existing HPC environments, not as bolt-ons but as part of a unified, heterogeneous computing system. Our new architecture demonstrates how near-term integration can enable more seamless execution of hybrid workflows, while also establishing a forward-looking path for deeper co-design between quantum hardware, classical accelerators, and scientific applications as systems scale and new algorithms emerge. Read our blog and paper for more details. We invite collaborators across HPC, quantum computing, and system design to join us in shaping the standards, best practices, and use cases that will define the future of quantum-centric supercomputing. blog: https://lnkd.in/eNJqfwzX paper: https://lnkd.in/epv9XsQ7

For quantum computing to reach its full potential, it will need to become part of a broader computing fabric—working alongside classical HPC and AI systems to tackle problems that no single paradigm can address alone. This has been the idea behind quantum-centric supercomputing (QCSC): integrating quantum processors with classical compute, and orchestration layers so hybrid algorithms can run as coherent, end-to-end workflows rather than fragmented experiments. Today we’re sharing a concrete step in that direction: our Quantum-Centric Supercomputer Reference Architecture, which describes how quantum processors can integrate with classical HPC and AI infrastructure across the full stack—from applications and orchestration layers to how these systems may ultimately be deployed in data centers. Today’s hybrid workflows are still largely stitched together manually by experts. Our goal with this architecture is to outline the system components, software layers, and interconnects that will be needed to make quantum-classical workflows more natural and scalable as hardware and applications mature. Importantly, the framework is evolutionary. Early systems may operate with loosely coupled resources, but over time we expect progressively tighter integration between quantum processors, CPUs, and GPUs—enabling deeper co-design across hardware, software, and applications. References in comments.

Who says connectivity is only about chip design? One of the most striking insights I took away from my chat with Pedram Roushan (Google) a few weeks ago was about 𝗿𝗲𝘄𝗶𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗾𝘂𝗮𝗻𝘁𝘂𝗺 𝗴𝗿𝗶𝗱—𝗻𝗼𝘁 𝗶𝗻 𝗵𝗮𝗿𝗱𝘄𝗮𝗿𝗲, 𝗯𝘂𝘁 𝗶𝗻 𝗰𝗹𝗮𝘀𝘀𝗶𝗰𝗮𝗹 𝗰𝗼𝗻𝘁𝗿𝗼𝗹 𝗹𝗼𝗴𝗶𝗰. In superconducting systems, qubits sit on a 2D grid. Long-range couplers between distant qubits? Technically possible—but costly, complex, and challenging to scale. But here’s the twist: 𝗬𝗼𝘂 𝗱𝗼𝗻’𝘁 𝗮𝗹𝘄𝗮𝘆𝘀 𝗻𝗲𝗲𝗱 𝗵𝗮𝗿𝗱𝘄𝗮𝗿𝗲-𝗹𝗲𝘃𝗲𝗹 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝗶𝗼𝗻𝘀 𝗶𝗳 𝘆𝗼𝘂𝗿 𝗰𝗼𝗻𝘁𝗿𝗼𝗹 𝗲𝗹𝗲𝗰𝘁𝗿𝗼𝗻𝗶𝗰𝘀 𝗮𝗿𝗲 𝗳𝗮𝘀𝘁 𝗲𝗻𝗼𝘂𝗴𝗵. Measure one qubit → process that information immediately in classical hardware→ apply a conditional gate on another qubit anywhere on the chip. Suddenly, 𝘁𝗵𝗲 𝗴𝗿𝗶𝗱 𝗯𝗲𝗰𝗼𝗺𝗲𝘀 𝗳𝗹𝗲𝘅𝗶𝗯𝗹𝗲. Connectivity becomes programmable. “If your feedback loop takes 500 nanoseconds, the whole procedure becomes pointless. But if you can do it fast—really fast—you effectively stitch your sample together for logical operation.” This is where modern control systems (like Quantum Machines OPX series) come in—offering ultra-low latency feedforward and feedback that makes these strategies practical. It’s not just a clever trick for entanglement generation. It’s a paradigm shift: • Adaptive calibration during job execution • Fast conditional logic without reconfiguring the chip • Software-defined connectivity at scale    This feels like one of the most underrated, yet powerful, enablers for near-term quantum experiments. 📸 Image adapted from Google Quantum AI

🔴 Xanadu publishes a milestone in #Nature. The paper Scaling and networking a modular photonic quantum computer proves that the path to millions of #qubits isn't making a bigger chip. It's networking them together. Building a monolithic #QuantumProcessor is hitting a yield and size wall. To scale, we must go #Modular. This work demonstrates a programmable, distributed quantum system that connects distinct #QuantumModules via #OpticalFibers, effectively turning a room full of server racks into a single giant quantum processor. 🔴 1. The Aurora Architecture The team unveiled a system comprising three interconnected quantum modules. Unlike #SuperconductingQubits which require complex microwave-to-optical transducers to leave the fridge, #PhotonicQubits are light. This allows for native, low-loss communication between modules using standard optical fibers, enabling a true #DataCenterScale quantum system. 🔴 2. Beating the #PercolationThreshold Connecting chips is easy, maintaining #entanglement across them is hard. The crucial breakthrough here is achieving an inter-module connection quality that exceeds the Percolation Threshold for #FaultTolerance. This means the distributed #ClusterState is robust enough to support #QuantumErrorCorrection, proving that modularity does not compromise computational reliability. 🔴 3. Synthetic Dimensions via #TimeMultiplexing Instead of just printing more physical qubits, Xanadu leverages Time-Domain Multiplexing (#TDM). They generate streams of entangled #SqueezedLight pulses that form a 3D cluster state in time. This allows a compact hardware footprint to generate a massive, scalable resource state for Measurement-Based Quantum Computing (#MBQC). 👇 Link in the comments #QuantumTech #Photonics #SiliconPhotonics #QuantumNetwork #QuantumInformation #OpticalInterconnect #AdvancedPackaging #Chiplet #MooreLaw #MoreThanMoore #SignalIntegrity #HardwareArchitecture #Semiconductor #Optoelectronics #HeterogeneousIntegration #Telecommunications #DataCenter PsiQuantum IonQ Rigetti Computing IBM Quantum Google Quantinuum D-Wave Intel Corporation TSMC Samsung Electronics SK hynix NVIDIA AMD Broadcom Marvell Technology Cisco GlobalFoundries Applied Materials Corning Incorporated

Nvidia’s Quiet Shift Towards Quantum Computing Why would a tech juggernaut like Nvidia, a premier chipmaker at the forefront of today’s AI revolution, make an under‑the‑radar move toward quantum technology? Do they see something that we don’t? The company recently unveiled its NVQLink, a custom make architecture that pairs quantum processors with classical supercomputers to enable hybrid quantum/classical computing. Nvidia also introduced CUDA‑Q, a platform that also connects supercomputers with quantum devices. And this move isn’t about supercharging quantum machines directly, but about accelerating and facilitating the blending of these two technologies into a cohesive whole. So, why does this matter? Until recently, quantum computing was considered a pipe dream, hindered by fragile hardware and the need for complex error correction. Nvidia’s new technology changes that dynamic by acting like a traffic cop - keeping quantum processor qubits in sync by using classical computing power. Now, smart software can monitor and correct system errors in real time which radically increases the number of queries that researchers can run. So, instead of waiting years for breakthroughs that are slowed by processing limitations, this integration should accelerate discovery across a wide range of fields and industries. Nvidia isn’t trying to become another quantum hardware manufacturer. Rather, it’s positioning itself as an indispensable link between classical and quantum systems by offering technology that allows organizations and researchers worldwide to plug into its platform. And while Quantum computing may still be years away, Nvidia’s bold move means the industry will mature and grow faster. By owning the key interconnecting technology that bridges quantum and classical devices, Nvidia is hedging its bets—and that could significantly reshape the future of computing. Here’s more: For a related video, click here: https://lnkd.in/g4CKzZSi - (GoogleLM) For more tech insight, check out this LinkedIn newsletter, Scientia: https://lnkd.in/gCaxVdPu #NVQLink #CUDA‑Q, #Nvidia #aiandquantum #bridgeaiquantumgap