It’s Monday March 17, 2025, and the streets of San Jose, Calif. are covered in green. You might assume that, on St. Patrick’s Day, the flashes of green on buildings, windows, cars, and buses celebrated the renowned Irish holiday. But you’d be wrong. San Jose was painted green for the annual NVIDIA GPU Technology Conference, and as Gen Z might say, the vibe “was lit!” The week of GTC was not unlike previous conferences in the sense that it had a similar format of show-floor exhibits and breakout workshops. However, the atmosphere had a different energy. Exhibits showcased primarily established technology organizations, and the wares peddled were practical and solution-oriented, central to data management and infrastructure needs. This pivot from the exhaustive “what’s possible with AI” fringe discussion to the “practical implementation of AI” was surprising, reassuring, and reflective of the maturation of the industry and the company. Here are five key takeaways from my time at GTC 2025: Software development tools are as essential as advances in hardware acceleration As expected, NVIDIA highlighted the leaps in processing power they’d achieved this year, boasting the capabilities of their latest AI chips, Blackwell Ultra (2025), Vera Rubin (2026) and Vera Rubin Ultra (2027). However, they also highlighted NVIDIA Dynamo, an open-source inference software designed to efficiently deploy and scale AI models. For the industry, this gives us the conviction that we are beyond the early stages of training models and ready to reap the benefits en masse (and ROI) via the outcomes and decisions these models infer, or inference. More importantly, being an open-source platform, we should expect accelerated adoption. The need for dual operations During his keynote, NVIDIA’s CEO Jensen Huang said, “Every industry, every company that has factories will have two factories in the future… the factory for what they build and the factory for the mathematics, the factory for the AI.” He gave the auto manufacturing industry as an example: “A factory for cars” and a “Factory for AI for the cars”. What does this mean? It indicates we should expect a boon in a new type of infrastructure central to proprietary data management and data protection. If you thought the Coca-Cola recipe was heavily guarded, imagine how companies will need to protect the intellectual property of their deployed AI models after investing in them. A unification process will need to occur downstream between these two factories. The infrastructure deployed to support inferencing of AI models at the edge will require sophisticated hardware and software integration to manage, maintain, and govern these solutions. Robots, robots everywhere… The idea of robots certainly isn’t new. Humans have been leveraging robotics to perform daily, repetitive tasks for decades. However, sprinkle in some AI and autonomy and suddenly it feels like we’re deep into Hollywood sci-fi territory. NVIDIA emphasized the significance of AI and robotic integration to key industries like industrial automation and med-tech, where these advances can be lifesaving. To further accelerate the adoption of robotics, NVIDIA introduced two open-source models. NVIDIA Isaac Gr00t N1, an open-source humanoid robotics foundation model, aimed to accelerate the adoption of humanoid robots and their Newton Physics Engine, an open-source physics engine designed to simulate training for robots to interact with the physical environment. We’re about to put AI to work, literally. Tokens, Trillions and Trillions of Tokens As with any industry show, you can bet that you will hear the emergence of this year’s bingo card of techie buzzwords. Through a plethora of “north stars,” “blue oceans,” “circle backs,” and “low hanging fruit,” two specific words stuck out to me more than the rest. Tokens and Trillions. Individually, tokens and trillions may not sound so impressive. But together? Now you’ve caught my ear. Let me explain why. In AI terms, a token is how the AI thinks. AI models analyze words, phrases, and individual characters to understand the context and fulfill requests. These models rely on tokens, which comprise vast datasets of words, phrases, and characters. While some models train on billions of tokens, more advanced models like Gemini 1.5, Claude 3, and GPT-4 likely draw from trillions of tokens. Of course, to process these trillions of tokens being used to train models, you need a heck of a lot of horsepower. Something like NVIDIA’s GB10 Grace Blackwell Super-Chip, capable of delivering up to 1,000 TRILLION operations per second of AI compute. To comprehend how much a TRILLION is, think of it this way. If I paid you $100,000 per day in 10 days, I would have given you $1M. In 27.4 YEARS, I would have paid you $1B. But $1T… a trillion dollars? It would take me 27,397 YEARS to pay you that $1T. Relationships matter Lastly, the aspect of GTC25 that resonated most with me was the emphasis on partnerships. You’ve likely heard an African proverb: “If you want to go fast, go alone. If you want to go far, go together.” This proverb highlights the need for collaboration, support, and partnerships to sustain success over the long run. NVIDIA has always fostered partnerships with its developer community to drive design, adoption, and technical advancements. But this year was FULL of announcements beyond their development and startup community. Relationships with major companies signify the importance “going together,” even for large technology companies. As I made my way home after the show, my brain still overstimulated from the obscene amount of green before my eyes, I found myself blurring the lines between “what WILL be and what already IS.” So many of yesterday’s impossibilities have been achieved or even exceeded today. Many of them were on display last week in a San Jose convention center. The one abundantly clear thing is that we’re nowhere close to understanding the many ways AI will transform our lives. It may seem like we’re all knee-deep in the next space race and on the precipice of ‘groundbreaking’ every single day. That’s because we are. One can only imagine where our industry will be at next year’s GTC. From Prototype to Market: Ready to bring your AI innovations to the world? Arrow’s end-to-end support ensures your solutions are optimized, integrated, and market-ready. Partner with us for seamless execution. Contact one of our experts today to learn more.  

Gone are the days of hefty upfront costs for IT assets. Many companies now prioritize flexibility and affordability, driving a significant shift towards OpEx- (operational expenditure) and subscription-based procurement models. Arrow Electronics collaborated with NVIDIA to deliver an edge AI Platform with an as-a-service business model in response to this trend. This offering combines the NVIDIA IGX industrial- and medical-grade edge platforms with the recently announced NVIDIA AI Enterprise Edge for IGX. The edge AI Platform-as-a-Service (PaaS) program from Arrow aligns business models across the supply chain and will accelerate the adoption and deployment of edge AI platforms in both the industrial and medical markets.

The landscape of data processing is undergoing a massive shift. As artificial intelligence (AI) and particularly Generative (Gen) AI continue the relentless march forward, original equipment manufacturers (OEMs) face a critical question: where will the data be processed and stored? Let’s delve into the evolving world of far edge, near edge, and on-premise computing, exploring the considerations for OEMs seeking to build Gen AI solutions. Gen AI, while powerful at the edge, introduces some unique challenges to edge infrastructure, including: Massive datasets: Gen AI algorithms often require vast amounts of training data, which can strain far edge and near edge resources. Limited processing power: Gen AI models are computationally intensive. Edge devices, often with limited processing power compared to on-premise servers, can struggle to run complex Gen AI models efficiently. Real-time or near real-time needs: Certain Gen AI applications, like personalized recommendations or anomaly detection, might necessitate real-time or near real-time processing, pushing the boundaries of latency tolerance. The traditional centralized data processing model in the cloud is giving way to a more distributed approach. Leveraging a combination of far edge, near edge, and cloud resources can distribute the processing load. Far edge devices can perform initial data collection and pre-processing, while near edge nodes handle preliminary training and filtering. Complex Gen AI computations can then be offloaded to on-premise data centers or the cloud. Figure 1: Distributed computing: From the Far Edge to the Core/Cloud Far Edge Computing: Extending the Boundaries Far edge computing refers to processing data at the outermost edge of the network, close to where data is generated, which is particularly beneficial in scenarios where low latency and real-time processing are crucial. Applications in Gen AI Advantages of Far-Edge Processing Smart Surveillance and Security: Gen AI models at the far edge can analyze video feeds in real time to detect threats and recognize faces, reducing the need for data transmission to central servers. Autonomous Robots or Vehicles Robots and vehicles equipped with edge devices can instantaneously process sensor and camera data, enabling safer navigation and obstacle avoidance. Reduced Latency: 5 to 20 ms. Immediate data processing near the source. Bandwidth Efficiency: Less data goes to centralized cloud systems. Enhanced Privacy: Local data processing mitigates the risk of data breaches.   Near Edge Computing: Bridging the Gap Acting as a bridge between far edge devices and the cloud, near edge devices offer more processing power than their far edge counterparts. Gateways, local servers, and fog computing nodes typically reside at the near edge, handling tasks like data aggregation, filtering, and preliminary analytics. Applications in Gen AI Advantages of Near-Edge Processing Retail: Near edge devices can analyze customer behavior in real time to provide personalized shopping experiences and manage inventory efficiently. Smart Cities: Environmental monitoring systems at the near edge can swiftly process and respond to data on air quality, noise levels, and traffic patterns. Balanced Latency: Up to 50 ms. Quicker than centralized processing, suitable for applications that require timely responses but are not instantaneous. Scalability: Easier to scale across multiple locations compared to far edge devices. Cost-Effective: Reduces the need for extensive infrastructure at the far edge.   On-Premise/Cloud Computing: Control and Customization On-premise computing involves deploying and managing computing resources within the organization’s physical location. This approach offers high levels of control and security but can be less scalable than cloud-based solutions. Applications in Gen AI Advantages of On-Premise/Cloud Processing Healthcare: Hospitals can use on-premise Gen AI systems to analyze medical imaging and patient data, ensuring compliance with stringent data privacy regulations. Industrial IoT (IIoT): Manufacturing plants can deploy AI models to monitor machinery and predict maintenance needs, reducing downtime and improving productivity. Complete Control: Full data and infrastructure ownership allows for tailored solutions. Enhanced Security: Data remains within the organization's premises, reducing exposure to external threats. Customization: Flexibility to customize hardware and software to specific requirements.   6 Considerations for OEMs: Selecting the Right Infrastructure The choice between far edge, near edge, and on-premise processing depends on several factors for OEMs: Data volume and complexity: For applications generating large volumes of complex data, near edge or on-premise solutions might be necessary. Conversely, low-volume sensor data might be well-suited for far edge processing. Latency requirements: Real-time decision-making often necessitates on-site processing. Applications requiring minimal delay might favor near edge or on-premise solutions over the cloud. Security and privacy: Far or near edge processing can offer greater control and security for sensitive data. Regulatory compliance might also dictate processing location. Cost and scalability: Cloud-based solutions provide scalability and eliminate the need for upfront hardware investment. However, on-premise solutions can provide greater cost control over time. Power availability: Far edge devices often operate on limited power, necessitating efficient processing with minimal resource consumption. Operating environments: Temperature variations, dust, noise, and humidity are factors when determining where to deploy the infrastructure. Summary: The Future is Distributed The future of data processing lies in a distributed approach, with far edge, near edge, and on-premise solutions working in concert with the cloud. As Gen AI evolves, OEMs must adopt flexible and scalable processing strategies. By understanding the strengths and limitations of each processing platform and implementing a hybrid approach when necessary, OEMs can ensure their solutions effectively leverage the power of AI at the edge. Companies looking to leverage Gen AI at the edge can benefit from the combined expertise of Arrow and Dell OEM Solutions. Dell's robust portfolio provides the ideal building blocks for far edge, near edge, and on-premise deployments. Arrow can guide companies through the process, from selecting the optimal hardware for their Gen AI application to helping ensure a smooth supply chain and offering implementation support. Their expertise in solution orchestration enables companies to deploy and manage their Gen AI solutions at the edge with greater efficiency and reduced risk. Related Resources: eBook: Infrastructure Considerations for Rugged Edge Applications Product brief: Storage Portfolio At-a-Glance Infographic    

From revolutionizing patient care in healthcare to powering smart cities and streamlining industrial operations, AI at the edge is driving transformative outcomes. The value of moving compute closer to the edge lies in its ability to deliver real-time insights, secure sensitive data, and lower the costs and latency of cloud-based solutions. In this blog, we’ll visit how edge AI enables breakthroughs in three key verticals and explore how Arrow Electronics can help deliver accelerated compute in edge applications. Improving Patient Outcomes at the Edge Accelerating pharmaceutical research and enhancing precision imaging are just a couple of ways AI is assisting healthcare professionals to make faster, more accurate decisions and ultimately help improve patient outcomes. Edge computing amplifies AI's benefits by bringing processing power closer to where data is generated - within hospitals, clinics, and even portable medical devices. For example, AI-powered systems at the edge can analyze X-rays or body scans in real time, supplementing healthcare professionals’ diagnostic capabilities while keeping sensitive patient data secure. Deploying AI at the edge in medical applications requires robust infrastructure, including purpose-built systems with significant compute capabilities. Compact servers or micro data centers near medical devices can help ensure real-time insights without latency, while dedicated server rooms can support larger-scale installations. Providing rapid results is essential in an industry where time can be the difference between life and death. By removing the need to transport data back and forth from the cloud, edge solutions deliver fast insights with enhanced data security, a key advantage in this industry where data privacy is essential. Enabling Smart Cities at the Edge In cities across the globe, engineers are finding new ways to leverage technology to advance sustainability efforts, improve urban life, and create more livable cities. From traffic monitoring and public safety to resource management, accurate local data processing is helping to reduce latency in these systems and lower costs. For example, video analytics at the edge can detect congestion patterns in transportation hubs or enhance security through intelligent surveillance without relying on continuous cloud connectivity. In retail applications, edge AI enhances customer experience and operational efficiency. Retailers can optimize inventory, improve store layouts, and deliver personalized marketing by analyzing video feeds and other real-time data on-site. Processing this data close to where it is collected reduces costs, improves response times, and helps ensure compliance with data privacy regulations. Reducing Downtime and Boosting Productivity AI and edge computing greatly benefit industrial environments, from manufacturing facilities to energy plants. Predictive maintenance solutions powered by AI help organizations anticipate equipment failures before they occur, reducing costly downtime, creating safer work environments for employees, and increasing productivity. Edge deployments in these settings often must overcome significant physical challenges. Hardware must operate reliably in harsh conditions, such as extreme temperatures, vibration, and high humidity, while delivering continuous performance. Hardware and Beyond Selecting the proper hardware is critical for deploying effective edge AI solutions. While powerful and impressive GPUs often dominate the conversation around AI, there may be better choices for some use cases. The diversity of workloads in healthcare, smart cities, and industrial settings requires a broader spectrum of hardware options tailored to specific use cases. Arrow’s Intelligent Solutions team takes a consultative approach to hardware selection, helping ensure organizations can balance performance, power consumption, scalability, and cost. Let’s consider the hardware categories for deploying edge AI solutions, their applications, and benefits: General-Purpose CPUs Use Case: Ideal for applications that need a balance of compute power, flexibility, and cost efficiency. Key Benefits: General-purpose CPUs are versatile workhorses that handle various edge workloads effortlessly. They integrate seamlessly with existing systems, making them a great choice if compatibility with your software ecosystem is essential. Example Applications: Think data analysis for smaller medical devices, lightweight analytics in bustling retail environments, and foundational IoT edge deployments that keep things running smoothly. GPUs for High-Performance Tasks Use Case: Suited for demanding computational tasks such as AI model training, real-time video analytics, and image processing. Key Benefits: GPUs excel at parallel processing that fuels real-time decision-making and advanced analytics. Example Applications: Imagine advanced diagnostics in healthcare that save lives, traffic monitoring systems that enhance smart cities, and predictive maintenance in industrial systems that keep operations running smoothly. ARM-Based Multi-Core Processors Use Case: GPU-free AI inference where energy efficiency is a priority. Key Benefits: These processors boast lower power consumption yet pack powerful compute capabilities, making them perfect for compact systems that need to perform. Example Applications: Computer vision and other edge AI workloads with price performance advantages over GPU-based systems. In many applications, deploying edge AI solutions isn’t just about selecting hardware components, it includes designing purpose-built systems capable of operating 24/7 in demanding environments. Edge deployments in healthcare require systems compact enough to fit in medical facilities yet powerful enough to process high-resolution imaging data. Industrial environments demand hardware that can withstand vibration, dust, and temperature extremes while delivering reliable performance. Arrow's commitment to delivering edge AI solutions shows in our relationships with industry leaders like NVIDIA and Ampere. These collaborations enable us to offer a diverse range of hardware options, helping ensure we meet the unique demands of various applications in healthcare, smart cities, and industrial settings. By integrating NVIDIA's advanced GPUs, we address the needs of high-performance tasks, while Ampere's ARM-based processors offer energy efficiency for AI inference. A Holistic Approach Deploying edge AI solutions involves navigating a web of hardware and software considerations. How do you ensure systems run 24/7? What’s the best way to balance performance and power consumption? Arrow’s holistic approach to guiding organizations through these challenges helps by tailoring recommendations based on the unique demands of each use case. It also helps achieve the right compute power and scalability balance and enables customers to deploy efficient and effective solutions. As hardware specialists, we support OEMs and ISVs by guiding them through the technical challenges of turning ideas into reality and achieving their goals: Recommending and sourcing the appropriate hardware for each application. Solving architectural and software challenges for reliable performance. Offering expertise in scaling and supporting edge deployments across verticals. Bringing the Edge to Life AI and edge computing are reshaping industries, driving innovation in healthcare, smart cities, retail, industrial applications, and more. By bringing compute power closer to the source, these technologies enable faster insights, improved efficiency, and better outcomes. With Arrow's expertise in purpose-built hardware and scalable edge solutions, organizations can confidently tackle the challenges of accelerated computing at the edge and bring their AI solutions to life.

Adopting Artificial Intelligence (AI) is an essential product feature and an operational asset. For many organizations, using AI becomes a reality when deployed at the edge – limiting data loss risk, lowering costs, and helping accelerate results. But where do you begin? For many original equipment manufacturers (OEMs), the journey starts by seeking clear answers to a few big questions: What data do you have now? What problems are you trying to solve with AI? And ultimately, how can you deliver and monetize those solutions? To compete in sectors like healthcare, manufacturing, and retail, OEMs need to approach AI from both sides—integrating it into their products and building it into their business processes. This article covers key elements in crafting an edge strategy that lets AI work for your products and your operations. A Purposeful Approach to AI for OEMs Deploying AI at the edge isn’t about seeking quick fixes for your problems. AI is about creating long-term, focused solutions that drive actual results. This purposeful approach to AI is what sets successful AI adoption apart. It starts with understanding what problem you’re solving and why it matters to your market. OEMs must identify their goals, knowing that successful AI strategies evolve and adapt over time. Arrow’s Intelligent Solutions team provides OEMs with a roadmap to success, offering purpose-built hardware and an end-to-end approach for deploying AI at the edge in real-world applications. From design and validation to supply chain management, Arrow helps align each step with your strategic goals. But even with the proper hardware, OEMs benefit most when they develop a holistic edge strategy that reduces complexity, manages costs, and accelerates time-to-market. Picking the Right Technology and Platforms The best AI platform does not necessarily mean the one with the latest hardware (although hardware performance matters). You also want to make sure you avoid obsolescence and have access to long-term support. Every AI solution requires a solid foundation—robust hardware and the right software ecosystem. Arrow’s role as an expert in providing purpose-built hardware means we can help organizations navigate a tricky marketplace and make the right decision for their specific needs. Leveraging relationships with leading technology suppliers is essential and allows you to stay current with hardware updates and shifts in the tech landscape. Arrow’s Intelligent Solutions team has spent decades building this network, helping OEMs create reliable connections with the right hardware suppliers and aligning them with the latest industry developments. What Is Needed to Deliver an Effective AI Solution? When delivering AI solutions, it’s important to remember that most products need careful hardware and software integration before they are customer-ready. With companies like Amazon and Roku leading the way in usability, today’s customers have higher expectations than ever before. For OEMs, a solution that delivers a consistent, user-friendly experience is the key to success. Arrow supports OEMs through every integration phase so complex AI systems function smoothly and are optimized to create value. This consistency and post-sales support are critical: if a system isn’t performing correctly, it’s not paying for itself. Our approach provides the infrastructure for OEMs to meet these expectations reliably, helping them build brand trust and operational efficiency. Capabilities Required for an OEM-Focused Strategy To maximize the value of AI, OEMs must focus on three essential capabilities: 1. Software Deployment and Integration Effective AI solutions aren’t just "plug and play"; they require thoughtful integration into existing systems. Software deployment must align with industry standards, optimizing performance without adding unnecessary complexity. For example, by handling deployment internally, OEMs can avoid the delays and costs of piecemeal integration and help launch immediate operational products. 2. Post-Sales Support and Risk Management The job continues once a product reaches the customer. OEMs need a dedicated support framework to ensure customers have a seamless experience. Post-sales support should extend throughout the lifecycle and offer troubleshooting, updates, and guidance when challenges arise, thus enhancing stability for OEM operations. 3. Customer Interfaces From order tracking to budget reviews, OEMs need straightforward tools to manage business processes efficiently. Leveraging user-friendly digital interfaces can empower OEMs, helping them focus on growth instead of navigating complex customer systems. These capabilities are foundational for any OEM looking to maximize the impact of their AI investments and maintain a competitive advantage. Putting AI to Work in Operations: "Don’t Just Make AI, Use It" For OEMs, adopting AI continues beyond integrating it into products. The real value comes when building AI into your operations. One example is Arrow’s internal AI chatbot powered by natural language processing. This tool will offer real-time information to team members across departments, streamlining workflows and enabling faster decision-making. A recent study from Stanford and MIT showed that similar AI chatbots improved productivity by around 14%. AI tools can help streamline internal operations for research or sales support. By utilizing AI within your processes, you can enhance productivity and improve your competitive position. Conclusion: Moving Beyond AI as a Buzzword It’s more apparent than ever that AI is more than a trend—it’s a transformative technology that can elevate an entire business when deployed thoughtfully. For OEMs, crafting an edge strategy to deploy AI that aligns with specific goals, maximizes operational efficiency, and supports long-term growth is essential and inspiring. Companies can turn this incredible potential into practical and meaningful gains by strategically integrating and aligning AI with evolving business needs.

The Evolution of Healthcare: A New Era Powered by Technology Healthcare as we know it today has advanced dramatically compared to the past four to five decades. Technology has brought groundbreaking innovations, improving life expectancy rates, healthcare delivery, prevention, monitoring, and overall patient care. In 2025, it’s estimated that 75 percent of medical data will occur at the edge. AI and Edge Computing: Driving Precision Medicine Forward Artificial Intelligence (AI) is at the forefront of revolutionizing the landscape of edge-based medical devices, playing an instrumental role in diagnostics and treatment strategies. AI in modern medical devices connected to the edge plays a pivotal role in the new era of data-driven, precision medicine, optimizing patient outcomes and empowering healthcare practitioners with advanced tools for more immediate, accurate diagnosis and intervention. AI algorithms embedded in medical devices analyze vast amounts of patient data swiftly and accurately, enabling early detection of diseases and conditions and improving prognosis and treatment outcomes. The predictive analytics drawn from AI can also significantly reduce unnecessary tests and procedures and guide and augment surgeons during treatment. Given the importance of these devices, the amount of generated data, the reliability necessary for critical medical environments, and the speed at which data must be available and accessible, edge architectures are vital. Data processing, analysis, and storage must occur at the collection point to enable urgent decisions. The low latency, high bandwidth, and security are imperative to real-time decision-making that is available with the support of edge computing. Advancing medical devices in the age of AI According to the World Health Organization[1], there are 2 million medical device types on the world market. In the US alone, the Food and Drug Administration (FDA) reports that there are more than 16,000 registered medical device companies, and countless modalities. The modalities range from diagnostic imaging (CT, MRI, x-ray, ultrasound) to endoscopes and laparoscopes, robotics, microscopy, DNA sequencers, cryogenic electron microscopy (cryo-EM), and more. Developing these high-performance medical devices can be challenging, and there is much to consider regarding computing power, hardware, and software. There must also be careful considerations for regulations, certifications, and standards for installation in the medical field. Hence, developers require an end-to-end, medical-grade platform with software and hardware that supports the medical AI lifecycle from training models to building and fine-tuning medical AI applications. When trying to enter the market quickly and accurately, performing preproduction tests, validation tests, and clinical trials is essential and will help ensure the product is ready for commercial production. How Arrow can support medical device AI innovation Arrow Electronics helps reduce this complexity for developers and medical device manufacturers. Our Intelligent Solutions business works with a rich ecosystem of technology providers, including NVIDIA, to offer a sophisticated combination of hardware, software, services, and AI expertise to support you in bringing innovative solutions to the market. Leveraging the latest in AI technology, we work with you to develop high-performing, next-gen solutions that address the evolving needs of the healthcare market and set you apart from the competition. Arrow’s AI design engineering expertise helps you develop state-of-the-art medical devices that help ensure healthcare professionals and patients benefit from the most advanced AI-driven technology. To discover the Top 5 edge AI applications shaping healthcare delivery and learn how Arrow helps medical device manufacturers get their products to market, download our Industry Brief.