Arrcus, Inc.

Open-weight AI and open-source AI aren’t just technical choices. There are economic and geopolitical implications. Great piece from Mitch Wagner at @FierceNetwork_ digging into how enterprises and telcos should actually think about open vs. closed models. Our CEO @shekar_ayyar weighs in on two things every infrastructure leader should be tracking: - Cost as leverage: Completely closed AI models are too expensive for large enterprises at scale. Open weights and open-source can keep AI labs more honest on pricing. But “open” has many flavors and doesn’t equate to free. - The China risk: standardizing on an “open” model with China control risks a rerun of the Huawei problem for telcos, an abrupt, total loss of access. Ayyar also shares where he sees telcos playing offense, not defense: as infrastructure providers for distributed inference. fierce-network.com/cloud/open-wei…

@arrcusinc and @UfiSpace new Ethernet solution for @AMD Instinct clusters shows where AI networking is headed: co-validated, open systems combining AMD compute, @Broadcom silicon and @Arcos without an open-ended integration project. My latest @650Group blog: 650group.com/blog/arrcus-et… #AINetworking #Ethernet

The future of networking begins when the network understands the application. But moving packets is no longer enough. AI inference has requirements no packet-level network can satisfy alone: → Which model is being requested? → Which site has the lowest latency right now? → Does this request cross a data sovereignty boundary? → Is there GPU headroom, or do we need to shift load? Answering those questions at routing time, automatically, in real time, is what AINF does. The network needs to think. Arrcus built the one that does. ☑️Which of these questions is your network answering today? Learn more at arrcus.com/inferencing-ne… #AINF #AIInferencing #NetworkDifferent #PolicyRouting #AIInfrastructure #InferenceNetworkFabric

👏 It's been a big week for @arrcusinc . Press Release: ✅ Today, @arrcusinc and @UfiSpace unveiled a production-ready AI networking solution for AMD-powered infrastructure. ArcOS® is deeply integrated with UfiSpace's Broadcom®-based Tomahawk® & Trident® open switch platforms, delivering an open, high-performance alternative to proprietary AI fabrics. The solution is showcased alongside GIGABYTE's GIGAPOD at AMD Advancing AI Day 2026 in San Francisco this week. [Read the release : arrcus.com/news/arrcus-an…) Industry Recognition: 🎉 We're honored that Arrcus Inference Network Fabric (AINF) has been named a finalist for @Light_Reading 's 2026 Leading Lights Award, in the "Most Innovative New Networking Product (Cloud or Data Center)" category. (lightreading.com/optical-networ…) Featured Interview: ✅ Our Chairman & CEO @shekar_ayyar spoke with @Analyticsindiam about why smarter data center networking, not just more GPUs, may be the key to solving the compute crunch, and how AINF is helping enterprises modernize inference infrastructure at scale. [Read the interview - analyticsindiamag.com/ai-features/in…) A new production deployment, industry recognition, and a conversation shaping where the market is headed all point to the same thing: the network is becoming the deciding factor in AI infrastructure performance. #AI #NetworkDifferent #DataCenter #ArcOS #Arrcus #AMD #AIInfrastructure #UFISpace #AIM

India's GPU shortage isn't just a hardware problem. It's a networking problem. In a new feature with @Analyticsindiam Magazine, our CEO @shekar_ayyar breaks down why smarter data centre networking could be India's edge in scaling AI infrastructure, from inference delivery to optimizing GPU utilization. Key takeaways: ✅ AI/ML workloads can push GPU utilization to 30–40%, just by fixing network bottlenecks ✅ Arrcus' ACE Platform helps enterprises and providers build and manage large-scale networks ✅ The shift from Content Delivery Networks to Inference Delivery Networks is already underway ✅ Open, flexible, programmable networking is what will separate AI leaders from the rest Read the full story to see how #Arrcus is helping shape the next generation of AI-ready data centres in India and beyond. analyticsindiamag.com/ai-features/in… #AI #DataCenters #NetworkDifferent #GPU #India

In Part 2 of his RESTCONF on ArcOS series, Mahesh Jethanandani explains how understanding the underlying YANG data tree transforms the way engineers interact with network APIs. Rather than memorizing RESTCONF paths, you'll learn a repeatable methodology for discovering them yourself. Read here: arrcus.com/blog/restconf-… Inside this article: ✅The three categories of YANG modules in ArcOS ✅How OpenConfig and Arrcus extensions work together ✅A repeatable method for discovering RESTCONF paths ✅Practical techniques for exploring and navigating live data efficiently Whether you're building automation pipelines or integrating network infrastructure into your applications, understanding the data model is what turns RESTCONF from a collection of endpoints into a powerful operational tool. #RESTCONF #YANG #OpenConfig #NetworkAutomation #NetDevOps #InfrastructureAutomation #Networking #APIs #ArcOS #Arrcus #NetworkDifferent #ArrcusBlog

Congratulations to our partners at @1FinityInc on their one-year anniversary! 🎉 One year ago, #1Finity launched with a bold vision, Networks Without Limits, carrying forward more than 90 years of @Fujitsu_Global telecommunications innovation into a focused, agile new form. In just twelve months, they've advanced next-gen AI networking, Open RAN, and intelligent automation, and we're proud to partner with them to deliver scalable networking solutions across data center, cloud, and edge environments for AI workloads. Here's to CEO Masaaki Moribayashi and the entire 1Finity team on an impressive first year. We're excited for what's ahead. Read their first-anniversary blog here: blog.1finity.com/2026/07/01/one… #NetworksWithoutLimits #AInetworking #Partnership #Networking

@arrcusinc and @TELUS Unveil AI Networking PoC Arrcus, has announced a PoC initiative with TELUS to evaluate the deployment of the AINF. Read the full article on themachinemaker.com/news/arrcus-an…

Today's read: Arrcus, TELUS Advance Secure AI Infrastructure with Sovereign AI Networking 📣 ift.tt/E0iY4co Read the full story by clicking on the link below ⬇️ Never miss a beat in telecoms. Catch the latest news on @TheFastMode 🚀 #tech #telecoms #technews

Great read from @ConvergeDigest on our PoC with @TELUS: bringing sovereign, distributed AI inference to Canada's network for public safety & government, with data kept in-country. As AI moves to inference, the network becomes the control plane. convergedigest.com/arrcus-telus-t… #SovereignAI #AIInference

AI, mobile networks, and distributed workloads drive the need for @TELUS to adopt more dynamic networking with @arrcusinc futuriom.com/articles/news/…

AI is shifting from centralized training to distributed inference and the network is becoming the control plane for performance, sovereignty and economics. @arrcusinc is proud to announce a proof-of-concept with @TELUS , one of Canada's leading communications technology companies, to explore the Arrcus Inference Network Fabric (AINF) as the networking foundation for sovereign, distributed AI inferencing at national scale. The goal: deliver secure, low-latency AI to mission-critical applications, from public safety and emergency response to government and enterprise services, while keeping sensitive data and workloads within Canadian borders. AINF is purpose-built for this moment. It's model-aware, intent-driven, and enforces sovereignty, security and policy at the moment a request enters the network. The results point to: ☑️ Up to 60% faster Time to First Token ☑️ 40% lower end-to-end latency ☑️ Up to 30% lower cost per inference ☑️ Line-rate encryption with zero CPU overhead "Public safety and mission-critical services demand AI that is fast, reliable and sovereign by design," said Tim Fell, VP Wireline Technology & Services, TELUS. ⚡Read the full announcement here: arrcus.com/news/arrcus-an… #SovereignAI #AIInfrastructure #AIInferencing #Networking #Telecom #EdgeAI #Arrcus #TELUS #NetworkDifferent

RESTCONF makes automating an Arrcus device as simple as a curl command. Distinguished Engineer Mahesh Jethanandani has published part one of a new series: "How to Use RESTCONF with Arrcus Devices" ➡️ Here's why it matters: ArcOS exposes its entire configuration and operational state over RESTCONF - the same YANG data-model foundation as NETCONF, but over a familiar HTTP/REST interface. That means configuring or querying an Arrcus device is as approachable as a curl command or a Postman request, with no NETCONF expertise required. This first post covers the essentials: enabling the RESTCONF server, the URI structure, HTTP methods (GET/PATCH/PUT/POST/DELETE), and real ArcOS examples using OpenConfig paths. ⚡ Read Part #1 arrcus.com/blog/how-to-us… 💡 What’s next Stay tuned for the next post goes deeper, with more hands-on examples. The timing is intentional as Mahesh Jethanandani will run a hands-on workshop on exactly this at #NANOG98 in Miami this October 2026, where engineers get direct experience using RESTCONF against Arrcus devices.

The hidden backbone of #AI: Why the network decides your GPU ROI - bit.ly/4fCiGDs | #technews #technology #futuretech

Is your AI really scaling intelligence, or inefficiency? In a recent @ExpComputer article, our Chairman & CEO @shekar_ayyar makes the case that the real bottleneck in AI isn't the GPU anymore. It's the network. The numbers are hard to ignore. GPU utilization in large clusters often sits at just 50–70%. In multimodal workloads, up to 84% of GPU power goes to waste. The moment the network can't keep up, those expensive GPUs sit idle, burning power and returning a fraction of their value. The fix: treat the network as a first-class layer of the AI stack, not an afterthought. As AI scales, the winners won't be whoever deploys the most GPUs. They'll be whoever keeps them productive. Read Shekar's full piece 👇 🔗 expresscomputer.in/guest-blogs/th… #AI #Networking #AIInfrastructure #GPU #Arrcus #NetworkDifferent

Congratulations to @Fujitsu_Global on the launch of Management Vision 2035! @arrcusinc is proud to be named a Sovereign Platform partner in this bold, ambitious roadmap. We share Fujitsu's vision to deliver superior value to customers along with other key infrastructure partners, including @nvidia, @AMD, @Oracle, and @Supermicro (see slide 10 in the linked presentation). This underscores our commitment to building the sovereign, high-performance infrastructure the world will rely on over the next decade - particularly in concert with the FUJITSU-MONAKA chip. We're honored to be part of the journey. Explore the full announcement here: okt.to/MpPe2m #ManagementVision2035 #Strategy #Innovation #Arrcus #Fujitsu

AI infrastructure is no longer centralized. Inference workloads are now distributed across clouds, edge locations, sovereign environments, and regional GPU clusters. That changes everything about how requests must be routed, managed, and optimized. Every inference decision now depends on real-time capacity, latency targets, model availability, KV-cache affinity, residency requirements, and inter-site network conditions. Traditional infrastructure was never designed for this level of orchestration. Read Part 1 of the AINF blog series by Nalin Pai, exploring the challenges shaping the inference fabric era: arrcus.com/blog/ainf-part… The next era of AI will not be defined by compute alone. It will be defined by how intelligently infrastructure coordinates distributed inference at scale. #Arrcus #NetworkDifferent #AIInfrastructure #AIInference #DistributedSystems #AINF

🎉 Arrcus is growing, and our Bangalore team now has the space to match! Our new India office was recently inaugurated by Derek Yeung (Co-Founder and Chief Architect), Abhay Roy (VP of Engineering), and Jennifer Smith (Sr. Director of HR & People Operations), marking a significant milestone in our continued global expansion of engineering and operational capabilities. India is central to what we build at Arrcus, where some of our most talented engineers are shaping the future of networking technology. Our new Bangalore office reflects that commitment: a bigger, better space built for the bold ideas and deep collaboration that our team brings every day. Hear directly from our Head of Engineering in Bangalore, Ashutosh Sharma, on what this milestone means 👇 arrcus.com/blog/arrcus-mo… #Arrcus #Bangalore #India #Networking #Engineering #NewOffice #NetworkDifferent

Exciting #Webinar Alert - Sanjay Kumar, VP Product Management and Marketing at Arrcus, joins Naomi Chen from @EdgecoreNetwork for #ECTalk to examine what actually delivers at scale: cloud-native routing on open networking platforms. @arrcusinc brings a disaggregated approach that removes dependency on legacy vendors and replaces it with programmable control, elastic scale, and real performance gains. ArcOS®, ACE, and AINF are built to support AI-driven, distributed infrastructure across data centers, 5G, edge, and multi-cloud without the usual cost penalties. Join us on May 13th to see how @arrcusinc and @EdgecoreNetwork enable a full programmable open networking environment - Part of Edgecore's Ecosystem of Choice. 📅 May 13, 2026 | 10:00 AM PST Register now at na2.hubs.ly/H0539bF0 #Arrcus #NetworkDifferent #OpenNetworking #CloudNative #AIInfrastructure #EdgeComputing #DataCenters #Edgecore

@arrcusinc is honored to be featured in @AvidThink's 2026 Data Center Networking Report, authored by @WireRoy. 📊 The report highlights a fundamental shift in how AI infrastructure must be built, and validates what we've been saying for over a year: AI is a distributed problem, and the network has to evolve with it. @AvidThink specifically calls out @arrcusinc for: ➡️ ArcOS delivering sub-second convergence and CDN-like routing intelligence for GPU resource scheduling across distributed data centers ➡️ Our Arrcus Inference Network Fabric (AINF) — purpose-built to route inference requests across sites with awareness of XPU loading, model state, and KV cache locations ➡️ Deep integration with NVIDIA BlueField-4 DPUs and NVIDIA Dynamo for hardware-accelerated, AI-policy-aware networking ➡️ 3x bookings growth in 2025, driven by demand for scale-across networking As the report notes, power constraints are making the single mega-data-center model increasingly unsustainable. Scale-across networking isn't a future concept. In fact, it's becoming mandatory. @arrcusinc has built ArcOS and AINF for exactly this moment. Operators who invest in a unified, hardware-independent fabric today will be best positioned to ride these transitions without re-architecting every cycle. Read the full @AvidThink report to dive into the trends shaping data center networking in 2026 - nextgeninfra.io/2026-dcnetwork/ #Arrcus #ArcOS #AIInfrastructure #DataCenterNetworking #AvidThink #ScaleAcross #InferenceNetworking #Ethernet #NetworkDifferent