For Earth Day, let’s update our thinking: Scaling AI does NOT have to mean scaling emissions. Improving utilization by 33% translates into 200 tons of CO₂ saving per rack annually. Now think about the 17 gigawatts of wasted energy in 2025 worldwide data centers that could have been saved and used to build more business outcomes on top of it. Moshe Tanach, NeuReality’s CEO: "When you fix system-level inefficiencies, what’s good for the business is good for the planet." Read more: https://lnkd.in/dnx9aaeC For a deeper dive on improving AI scale-out efficiency, see our new white paper in the first comment 👇
NeuReality
Semiconductor Manufacturing
Tel Aviv, Tel-Aviv District 7,336 followers
Transforming AI Infrastructure into a Unified Inference Platform.
About us
AI infrastructure has a hidden problem: the network and orchestration layer. As models scale to trillions of parameters and inference demand explodes, two bottlenecks emerge: how data moves between GPUs and how workloads are managed across them. The industry added more GPUs, scaled clusters, optimized models. But utilization still hovers around 50-70%. The compute is there, idle, burning watts. The bottleneck isn't the silicon. It's how data moves and how work gets distributed. Traditional networking was built for general-purpose workloads, not AI's east-west traffic and microsecond-sensitive synchronization. Traditional orchestration treats GPUs as generic compute, blind to the demands of prefill, decode, and model synchronization. Every GPU cycle wasted waiting is money and energy lost. We asked: What if the network wasn't just faster, but intelligent? What if orchestration understood AI workloads natively? NR-NEXUS is an inference operating system for large-scale inference. Hardware-agnostic, it unifies fragmented open-source frameworks into a single production platform, running across hyperscale clouds, GPU clusters, and emerging XPUs. NR2 AI-SuperNIC eliminates data-movement bottlenecks limiting GPU utilization. It executes the networking data path in hardware with no CPUs in the critical path, integrates in-network compute to offload communication operations, and supports open Ethernet-based networking. Together, they transform distributed GPU and XPU clusters into high-throughput token factories. The result: GPUs at near-100% utilization. Inference scales without adding racks. Energy consumption drops. This isn't incremental optimization. It's rethinking the data path and control plane so AI infrastructure matches AI ambition. For our customers: maximum performance from existing hardware. Lower cost, lower power, lower latency, higher throughput. NeuReality is headquartered in Tel Aviv with offices across North America and Europe.
- Website
-
http://www.neureality.ai
External link for NeuReality
- Industry
- Semiconductor Manufacturing
- Company size
- 51-200 employees
- Headquarters
- Tel Aviv, Tel-Aviv District
- Type
- Privately Held
- Founded
- 2019
- Specialties
- Machine Learning, Artificial Intelligence, Semiconductors, AI Inference, Data Centers, AI Infrastructure, Generative AI, Large Language Models, AI Deployments, AI Systems Engineering, AI Inference Chips, Computer Vision, AI Software, NR1 AI Inference Solution, AI Networking, AI Orchestration, AI Factories, AI at scale, Scale-out, and Scale-up
Locations
-
Primary
Get directions
10 Kremenetski
Tel Aviv, Tel-Aviv District 6789910, IL
-
Get directions
14 Tarshish Street
Caesarea, 3079559, IL
-
Get directions
2880 Zanker Rd
203
San Jose, California 95134, US
-
Get directions
Kamienna 21
Krowodrza, Małopolskie 31-403, PL
Employees at NeuReality
Updates
-
Your AI infrastructure is sitting idle. 𝟰𝟬–𝟳𝟱% 𝗼𝗳 𝗿𝘂𝗻𝘁𝗶𝗺𝗲 𝗶𝘀 𝘀𝗽𝗲𝗻𝘁 𝗼𝗻 𝗱𝗮𝘁𝗮 𝗺𝗼𝘃𝗲𝗺𝗲𝗻𝘁. See what’s limiting performance at scale-out and how to unlock more throughput from existing infrastructure. New white paper: https://lnkd.in/daW5rq7V
-
-
We’re excited to welcome Rotem H. as a Strategic Advisor, joining Moshe Tanach to help shape how enterprises bring AI into real-world production🚀 With leadership experience from Amazon, Rotem brings deep expertise in scaling global platforms at a pivotal moment for enterprise AI. “With NR-NEXUS, NeuReality is building a more practical foundation that enables AI to operate as a true infrastructure.” Welcome aboard, Rotem! https://lnkd.in/dGf4XSrb
-
-
Just as spring brings a fresh start, we’re celebrating the spirit of renewal and the new ideas driving the future of technology. While the world hunts for eggs 🪺, we’re busy hunting for the next breakthrough in AI infrastructure 💡. Whether you’re spending the day with family or taking a moment to recharge, we wish our partners and community a wonderful holiday filled with inspiration and new beginnings. 🌸 #Easter2026 #Innovation #AIInfrastructure #DeepTech
-
-
NeuReality reposted this
Classic computing and software approaches just don’t work anymore leaving expensive AI factories Idle while supply is limited. It’s time to think differently. Everywhere. Last week Rene Haas said “AI has fundamentally redefined how computing is built and deployed” and that’s what we at NeuReality together with ARM been working on when we’ve built the first NR1 AI-CPU and now when we take NR-Nexus to Neoclouds and GenAI companies and NR2 AI-SuperNIC to hyperscalers and sovereign AI projects.
AI inference at scale depends on more than accelerators alone. It requires a balanced approach to infrastructure - from compute to orchestration - to deliver consistent performance, efficiency, and cost control in production environments.⚡️ In this video, Moshe Tanach, CEO of NeuReality shares how our partnership is helping address system-level bottlenecks using Arm Neoverse, enabling more scalable and efficient AI inference across cloud deployments. NeuReality’s NR-NEXUS platform addresses these infrastructure challenges in real-world AI inference systems. 👀➡️: https://okt.to/GixEJH
-
NeuReality reposted this
AI inference at scale depends on more than accelerators alone. It requires a balanced approach to infrastructure - from compute to orchestration - to deliver consistent performance, efficiency, and cost control in production environments.⚡️ In this video, Moshe Tanach, CEO of NeuReality shares how our partnership is helping address system-level bottlenecks using Arm Neoverse, enabling more scalable and efficient AI inference across cloud deployments. NeuReality’s NR-NEXUS platform addresses these infrastructure challenges in real-world AI inference systems. 👀➡️: https://okt.to/GixEJH
-
In a world where Artificial Intelligence is rewriting the rules every day, we at NeuReality continue to push boundaries and redefine how compute power meets reality. As we approach Passover, a holiday symbolizing renewal, growth, and freedom, we want to take a moment to wish all our partners, customers, and colleagues a very Happy Passover! May this spring bring with it fresh opportunities, boundless creativity, and shared technological milestones. Wishing you all a joyful and inspiring holiday from the entire NeuReality team! 🍷✨
-
-
Great discussion from Ian Cutress and Sally Ward-Foxton. "Anything you do to increase GPU utilization is probably a good thing." That was the starting point behind NR1: rethinking the head node. But the real challenge was never just the head node. It's the system around it. With NR2 AI-SuperNIC and NR-NEXUS, we are now addressing scale-out across AI factories and the orchestration layer above them.
https://lnkd.in/eTXjPcCR Latest podcast of the AI Hardware Show S2 is now online! Joined by Sally Ward-Foxton from EE Times | Electronic Engineering Times, we talk about the mainline CPUs in AI hosts, as well as a few up and coming: Intel, AMD, IBM, NeuReality, Tenstorrent. Hyperscaler chips in a future episode :) Link to the full playlist: https://lnkd.in/e6J_PuBf
-
-
NeuReality reposted this
The vast majority of AI users are still chasing more compute to meet the demand. Some don't even monitor their utilization but others are just staying away from the complex and time consuming task of optimizing LLM production. That's why we developed NR-Nexus Token Factory OS to solve both issues and recover the hidden supply under our customers' hands.
Why does serving an LLM feel like systems hell when classical ML felt almost boring? Traditional AI models are simple: send in a request, get back an answer. Done. LLMs work completely differently: they generate one word at a time, loop back on themselves, and hold onto memory that grows with every token. The longer the conversation, the heavier the load. Most infrastructure treats it like a compute problem when it's not - it's a memory and scheduling problem, and that distinction is what breaks production AI at scale. This is exactly the infrastructure problem NeuReality is built around. Or Zipori from our team breaks it all down in part 2 of his series. Link in the first comment 👇
-