Latest updates from the AI industry
Learn five ways developers can use Google Gemini models for AI search, data extraction, smart agents, multimodal apps, and real-time voice features.
Flint, developed by Recursive, introduces a new approach to AI-driven programming with its “mixture of experts” architecture. Unlike traditional dense models, Flint activates only 2.5 billion of its 40 billion parameters per token, significantly reducing computational overhead while maintaining competitive performance. Sam Witteveen highlights how this selective activation, combined with features like sliding window attention, […] The post Flint Could Be the First Coding AI Built via an RSI System appeared first on Geeky Gadgets.
A new generation of AI systems is beginning to interact with the physical world through robots capable of translating high-level goals into actions.
GSMA’s Jawad Abbassi says the event will bring together telecoms operators, policymakers and technology leaders as Gulf countries accelerate investment in 5G, AI and datacentres
DeepSeek AI model shows safety gaps in independent tests
Newest Signal AI agent fixes software and supply chain vulnerabilities at machine speed while ensuring developers maintain control of what changes go into their code BURLINGTON, Mass., Oct. 27, 2026 /PRNewswire/ -- Black Duck®, the leader in application security for the AI era, today...
Beijing [China]: DeepSeek’s latest AI model showed stronger safety controls in independent tests than earlier versions, but could still disable software intended to shut it down, exploit gaps in monitoring and follow deceptive instructions during long tasks, according to a Chinese AI-safety group. Start-up DeepSeek is one of China’s most closely-watched artificial intelligence companies, and […] The post DeepSeek AI model shows safety gaps in independent tests appeared first on Revoi.in.
The tests examined whether the model would act beyond a user’s instructions, evade human oversight or improperly influence people.
DeepSeek's logo and a keyboard in this illustration taken October 19, 2026. — Reuters BEIJING: DeepSeek's latest AI model showed stronger safety controls in independent tests than earlier versions, but could still disable software intended to shut it down, exploit gaps in monitoring...
DeepSeek's latest AI model showed stronger safety controls in independent tests than earlier versions, but could still disable software intended to shut it down, exploit gaps in monitoring and follow deceptive instructions during long tasks, according to a Chinese AI-safety group. Start-up DeepSeek is one of China's most closely-watched artificial intelligence companies, and its models have drawn global attention for their performance and relatively low operating costs.
OpenAI has enhanced its models to work with legacy programming languages like COBOL and Fortran. The AI assists companies in modernising critical software systems, which often lack documentation. Recent tests showed significant improvements in accuracy for building and translating software across multiple languages.
Daijiworld Media Network - Seattle Seattle, Oct 20: Amazon.com is introducing a new software tool that allows developers to run code written in the popular artificial intelligence (AI) programming framework PyTorch on its custom Trainium AI chips without rewriting the code, in a move aim....
Dream Labs, a startup trying to make robots more intelligent and capable of performing many different tasks, has raised $90 million in seed funding to coincide with the launch of a public application programming interface that provides access to its universal robot foundation model. Today’s round was jointly led by B Capital and Kindred Ventures […] The post Dream Labs raises $90M to help build more intelligent robots with simple video feeds appeared first on SiliconANGLE.
Amazon.com Inc. today introduced TorchNeuron, an open-source toolkit that enables developers to run PyTorch workloads on its AWS Trainium chips. PyTorch is the industry’s most popular artificial intelligence development framework. Until now, running AI projects powered by the tool on Trainium-based Amazon Web Services instances involved a significant amount of manual work. That increased development […] The post Amazon introduces toolkit for running PyTorch workloads on AWS Trainium chips appeared first on SiliconANGLE.
Study Artificial Intelligence MSc at Brunel. Located in west London. Computer Science at Brunel is ranked 5th in London by the THE World Rankings 2025.
DeepSeek keeps giving businesses reasons to try its low-cost, high-performing, open source AI models. The U.S. government, though, is saying 'no dice.'
Chinese artificial intelligence startup DeepSeek is collaborating with Huawei Technologies to develop advanced AI chips optimized for its models, potentially reducing China's dependence on Nvidia hardware. The partnership represents a significant step toward building a domestic AI infrastructure capable of supporting next-generation AI development. According to an October 19 report by The Information, Huawei engineers have been working directly at DeepSeek's offices to improve chip designs based on the startup's technical requirements. Their collaboration focuses on optimizing AI computations, improving data transfers between processors, and increasing overall training efficiency. DeepSeek reportedly intends to begin training AI models using Huawei's Ascend 950DT processors by the end of 2026. Huawei is also expected to allocate most of its upcoming Ascend 960DT chips to DeepSeek, with deliveries anticipated during the first quarter of 2027. The DeepSeek-Huawei partnership aligns with Beijing's broader strategy to strengthen China's semiconductor industry and reduce reliance on American technology. US export restrictions have limited Chinese companies' access to advanced Nvidia AI chips, accelerating efforts to develop competitive domestic alternatives. However, Huawei faces considerable technological challenges in competing with Nvidia. Research from Epoch AI indicates that Nvidia's B300 processors deliver approximately seven times the raw computational performance of Huawei's Ascend 950 chips. Training advanced AI models using Huawei hardware could require more than 10,000 interconnected processors operating continuously for several weeks. Consequently, chip production capacity, system stability, and computing efficiency remain major obstacles to achieving Nvidia-level performance. Meanwhile, DeepSeek is expanding its software capabilities through TileLang, an open-source programming language designed to support AI dev
Developers who want to complete control over their game's code have a new framework to lean on.
The Urich leak raises concerns about AI-assisted programming, third-party messaging services, and the potential for personal accounts to expose sensitive political communications.
The Maker ESP32 is a development board designed to streamline the process of building smart electronics and IoT prototypes. A key feature is its plug-and-play sensor integration, which uses I²C connectivity to bypass soldering and complex wiring. This approach simplifies setup for beginners and educators while also saving time for experienced developers. Additional features like […] The post Maker ESP32 Replaces Soldering with Plug-and-Play I²C Ports appeared first on Geeky Gadgets .
Level up your game dev skills with 11 top AI courses for 2026—covering ML, procedural generation, and intelligent NPCs. Find the right fit to future-proof your career.
The post Frnd Internship 2026: Software Quality Assurance Intern | Bengaluru | Apply Online first appeared on Job4freshers .
Google has announced Gemini 4 Argon, a new artificial intelligence (AI) model designed to handle complex coding tasks, professional research, enterprise workflows and cybersecurity. The model is initially being made available to a limited group of trusted cyber defenders, with a broader rollout planned for developers, businesses and consumers. Announced on September 30, Gemini 4 Argon is built to work through longer, multi-step tasks, including software development, financial research, legal drafting and security vulnerability detection. Google said the model can analyse large amounts of information and carry out tasks that require several stages of reasoning. One of the key changes is its expanded output limit of up to one million tokens, compared with 64,000 tokens previously. This allows the model to generate longer responses and work through complex tasks in a single sequence. Focus on Coding and Cybersecurity Google said its engineers are already using Argon for debugging, software migrations and algorithm design. In one example, the company said the model helped optimise software used for video decoding, making a Rust-based implementation 2.7 times faster than an existing Rust version while producing identical video output. The model is also designed to identify, validate and patch software vulnerabilities. Google said Argon identified a critical vulnerability in healthcare software used by hospitals worldwide that exposed sensitive personal information. The company did not identify the software in its announcement. Google is initially providing access to trusted cyber defenders through its cybersecurity programme, while continuing to test the model's safeguards before expanding availability. Pricing and Availability Gemini 4 Argon will initially cost $2 per million input tokens and $10 per million output tokens. Cached input tokens will be priced at a 95% discount against the standard input rate. After the introductory pricing period, the rates will increase to $4 per million input tokens and $20 per million output tokens, according to Google. The company plans to begin broader access with paid application programming interface (API) customers and Google AI Ultra subscribers before expanding availability further. Google said it is continuing safety testing and refining safeguards against misuse and prompt injection attacks before a wider release.
Pakyong, October 2: Google Gemini 4 Argon, the technology giant’s newest flagship artificial intelligence model, was unveiled on Wednesday, September 30, with the company describing it as its most advanced system so far for coding, cybersecurity defence and complex professional work. Access has initially been restricted to vetted cybersecurity partners and Google’s own teams, and […] The post Google Gemini 4 Argon Unveiled: Most Advanced AI Model Yet appeared first on The Voice Of Sikkim .
The system could cut robot retraining from hours to minutes in factories where layouts and product lines change often.
The system could cut robot retraining from hours to minutes in factories where layouts and product lines change often.
The system could cut robot retraining from hours to minutes in factories where layouts and product lines change often.
Teams can now share projects and documents in ChatGPT, as OpenAI pushes the service deeper into everyday office workflows.
Teams can now share projects and documents in ChatGPT, as OpenAI pushes the service deeper into everyday office workflows.
The MSc in Software Development is designed in collaboration with industry to provide a year of intensive education in software development and is specifically designed for graduates of non-computing disciplines. It has an extremely high graduate employability rate and is the highest rated MSc computing course for student satisfaction in the UK. COURSE CONTENT: You will learn the foundations of computing and acquire industry relevant skills. These core skills include programming, databases, web development, software engineering and problem solving. You need no computing background at all to do this course, but on successful completion, you will be equipped with the technical, analytical and professional skills required to take on a range of roles in the computing industry. The curriculum was designed and is continually reviewed in collaboration with significant industrial partners, who helped ensure the most relevant module content. This helps ensure that our graduates are highly trained and much sought after by employers. WHO WILL BE TEACHING YOU: Many of the course teaching team have worked in the computing industry and have particular research specialisms in developing novel and effective methods for teaching computing. Several of the team have been awarded student-nominated teaching awards for their work. WHAT YOU'LL GET UP TO: You’ll learn how to analyse problems and design robust and elegant software systems to solve the problems. This includes gaining skills in programming, web development, database design, software testing and software engineering. You will also have the opportunity to learn how to work as part of a development team using modern development techniques and tools. We've created a short taster session to give you a hands-on introduction to coding and a sense of what to expect during your first week of class. You can access it here: http://go.qub.ac.uk/try-software-dev PLEASE NOTE: Applications for this course received after 12th August 2026 may not be accepted. In addition, a deposit will be required to secure a place.
The MSc in Software Development (Artificial Intelligence) is designed in collaboration with industry to provide an intensive education experience in software development and is specifically designed for graduates of non-computing disciplines. Traditionally the programme has had a significant graduate empolyability rate and is consistently one of the highest rated MSc computing courses for student satisfaction in the UK. This part-time programme is taught in the evenings (one to two evenings a week per semester) and spread over 3 years to allow you to study while in full-time employment. The course is completed entirely online, with live online lectures and lab sessions. All learning materials are hosted on our virtual learning environment, Canvas, in addition to pre-recorded bite-sized lecture topic videos, feedback quizzes to check your progress, as well as individual support from our lecturers and lab demonstrators. You can also optionally attend our on-campus labs on the teaching nights for additional support or if you need a quiet environment to work in. You also have the option of completing the MSc programme in 2 years. A dissertation project or set of specialised Mini Project modules can be undertaken in the third semester of the second year of the course, once all the taught modules are successfully completed. The project or set of Mini Project modules can be undertaken on a full time to enable successful completion in that semester. COURSE CONTENT: You will learn the foundations of computing and develop the industry-relevant skills required for modern software development careers. These core skills include theoretical and applied knowledge of programming, databases and web development, alongside software engineering principles and practices. In addition to technical capability, the programme develops the analytical, professional and transferable skills increasingly sought by employers across the global technology sector. Students gain experience in communication, teamwork, project delivery, critical thinking and adapting to new technologies — skills that are essential for working effectively in modern multidisciplinary software teams. You do not need a computing background to undertake this course. On successful completion, graduates will be equipped with the technical, analytical and professional competencies needed to pursue a wide range of careers within the computing and software industries. The curriculum has been designed and is continually reviewed in collaboration with industrial partners to ensure that module content reflects current industry practice and emerging technologies. This helps ensure that graduates are highly skilled, adaptable and well prepared for employment in a rapidly evolving digital economy. WHO WILL BE TEACHING YOU: Many of the course teaching team have worked in the computing industry and have particular research specialisms in developing novel and effective methods for teaching computing. Several of the team have been awarded student-nominated teaching awards for their work. WHAT YOU'LL GET UP TO: You’ll learn how to analyse problems, perform problem decomposition and design robust and elegant software systems to solve the problems. This includes gaining skills in procedural and object-oriented programming, web development technology stacks, database design and development, AI-augment development approaches, and software project management. You will also have the opportunity to learn how to work as part of a development team using modern development techniques and tools. We have created a short taster session to give you a hands-on introduction to programming and a sense of what to expect during topics on the programme. PLEASE NOTE: Applications for this course received after 12th August 2026 may not be accepted. In addition, a deposit will be required to secure a place.
Apply for your research degree at Queen's University Belfast, Russell Group university. Find out more about funded and unfunded PhD projects.
The MSc Financial Analytics is the programme for you if you have an interest in financial markets or financial technology (FinTech) and enjoy working with data. It shows you how data science, analytics, statistics and programming tools are used in the real world for the analysis and modelling of financial and economic data. The programme will equip students with cutting-edge quantitative and computational techniques utilised by industry leading firms. The course aims to bridge the gap between complex quantitative models and financial decision making and does so by equipping students with dual skillsets in both finance and data analysis. It can lead to exciting careers in areas such as financial data science, trading, software development, portfolio management, consulting, data analytics, risk management, business analytics and academia.
The MSc Software Development (AI) at Queen's University Belfast is an intensive one-year conversion programme designed for graduates from any academic background who want to transition into the software and technology sector. No prior experience in programming or software development is required. Developed in collaboration with industry, the programme reflects how software engineering is evolving in response to Artificial Intelligence (AI), automation and modern development practices. Students build strong foundations in programming, software engineering, databases, web development and systems design while gaining practical experience of the AI-assisted tools, workflows and technologies increasingly used by professional software developers. Throughout the programme, AI enhances learning and reflects contemporary industry practice, while maintaining a strong emphasis on the core technical knowledge and software engineering principles that underpin professional software development. Through practical, project-based learning, students develop real-world software solutions, solve complex engineering problems and work using the collaborative approaches, development practices and professional tools employed across today's software industry. By the end of the programme, students have progressed from beginner level to industry-ready junior software developers with a portfolio of practical work demonstrating both fundamental software engineering expertise and experience of AI-enhanced development. Students develop expertise across software engineering, programming, application development, databases, web technologies and systems design, alongside highly valued professional skills including communication, teamwork, project delivery, critical thinking and the ability to evaluate and adopt emerging technologies. These capabilities are increasingly sought by employers throughout the global technology sector. The curriculum is continually reviewed with industry partners to ensure that teaching reflects current professional practice, emerging technologies and the evolving needs of employers across the software and technology sectors. Who will be teaching you? Teaching is delivered by academic staff from the School of Electronics, Electrical Engineering and Computer Science (EEECS) at Queen's University Belfast, an internationally recognised school with an established reputation for research excellence, high-quality teaching and strong industry engagement. The programme team combines research expertise, industry collaboration and extensive teaching experience, enabling students to understand both the theoretical foundations of software development and their application within modern software engineering practice. Several members of the teaching team have received student-nominated teaching awards in recognition of their commitment to delivering an outstanding student experience. What you'll get up to: This is a highly practical programme built around active learning, hands-on laboratories and real-world software development. Throughout the programme, students undertake practical projects that mirror the challenges encountered by modern software engineering teams. You will learn how to analyse complex problems, decompose them into manageable components and design, develop, test and deploy robust software solutions. Along the way, you will gain practical experience in procedural and object-oriented programming, database design, modern web application development, software architecture, software engineering and project management. Reflecting the changing nature of the profession, you will also develop experience of AI-assisted software development, learning how to use contemporary AI tools responsibly and effectively to support activities such as software design, coding, testing, debugging and documentation. These skills are developed alongside—not instead of—the fundamental programming and software engineering knowledge expected of professional software developers. Throughout the programme you will work both individually and collaboratively, using modern development environments, version control systems, agile methodologies and professional software engineering workflows commonly adopted throughout industry. Many students also undertake individual projects in collaboration with industry partners or research groups, providing opportunities to contribute to innovative developments in areas such as Artificial Intelligence, machine learning, cybersecurity, cloud computing and data analytics.
Aurora Mobile (NASDAQ:JG – Get Free Report) and Samsara (NYSE:IOT – Get Free Report) are both technology companies, but which is the better stock? We will compare the two businesses based on the strength of their risk, valuation, earnings, profitability, dividends, analyst recommendations and institutional ownership. Volatility & Risk Aurora Mobile has a beta of [...]
Samsara (NYSE:IOT – Get Free Report) and Aurora Mobile (NASDAQ:JG – Get Free Report) are both technology companies, but which is the better business? We will compare the two businesses based on the strength of their dividends, earnings, profitability, valuation, analyst recommendations, institutional ownership and risk. Risk & Volatility Samsara has a beta of 1.35, [...]
This article explores the transformative landscape of digital innovation, focusing on emerging trends, technologies, and their implications for various industries. Join us as we delve into the future of tech with insightful analysis. What is Digital Innovation? Digital innovation refers to the integration of digital technology into all areas of a business, significantly changing how [...]
Programming book reviews, programming tutorials,programming news, C#, Ruby, Python,C, C++, PHP, Visual Basic, Computer book reviews, computer history, programming history, joomla, theory, spreadsheets and more.
Industrial energy procurement once revolved around fixed utility tariffs, predictable pipeline spreads, and steady seasonal consumption patterns. However, the convergence of upstream crude swings, regional [read full press release...]
Aman Sanger co-founded Anysphere, known for its innovative Cursor application. SpaceX's recent acquisition of Anysphere for $60 billion propels its founders into billionaire status. Cursor revolutionizes developer productivity by seamlessly integrating AI into coding environments, aligning perfectly with Elon Musk's ambitions in artificial intelligence. This strategic deal signifies a pivotal moment in the evolution of software development towards AI-enhanced coding techniques.
What if the 1991 Linus Torvalds woke up in 2026 and called the whole AI stack bad taste? And an open source approach to change that.
Parallel coding agents solve the waiting problem, not the mentorship one. Redesign mob programming around watching AI decide, not write.
<p>Each step of the AI workflow -- from data preparation, to analytics, task execution and workflow orchestration -- depends entirely on processing hardware.</p> <p>But not all processing power is created equal. Processors have evolved and diversified to offer powerful features that support specific, critical functionalities not found in traditional central processing units (CPUs), such as efficient data handling and complex mathematical capabilities. The rapid emergence of AI has accelerated this diversification, and there's no end in sight.</p> <p>Leaders committed to deploying AI need to understand the current suite of AI chip architectures to optimize AI platform performance while managing costs across their organizations.</p> <section class="section main-article-chapter" data-menu-title="Why are chip architectures diversifying?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Why are chip architectures diversifying?</h2> <p>A <a target="_blank" href="https://www.techtarget.com/it-infrastructure/tip/CPU-vs-microprocessor-What-are-the-differences" rel="noopener">microprocessor or processor</a> is an "engine" that fetches instructions from memory, decodes them, performs the functions and calculations specified by those instructions and returns results to software, OSes and applications. The processor is composed of millions of microscopic transistors arranged to execute every instruction from start to finish, input to output. Every instruction can correspond to millions of transistors, and a processor capable of handling hundreds of unique instructions can involve tens of billions of transistors, each using power and generating heat.</p> <p>Early processors were general-purpose. They could perform a little logic, basic math and simple data handling. As workloads required more features and functions, the instruction sets and transistors swelled. Then, two things happened: first, the processors became too complex, resulting in excessive power consumption for too little performance; second, the processors became too cumbersome and expensive to manufacture. The processors' evolutionary tree had to split.</p> <p>Graphics co-processors, known as <a href="https://www.techtarget.com/ai/tip/How-to-choose-the-best-GPUs-for-AI-projects">graphics processing units (GPUs)</a>, followed, specializing in floating-point math originally suited to graphics-rendering tasks such as gaming. Today, GPU mathematical capabilities are commonplace and crucial elements of AI training and inference. Then followed neural processing units (NPUs) and tensor processing units (TPUs) to accelerate AI and machine learning (ML) workloads, including LLMs, autonomous vehicles and healthcare use cases.</p> <p>For other use cases, established processors are still too expensive or complex for specific tasks, and application-specific integrated circuits (ASICs) have emerged to meet a wide array of task-specific needs, providing narrow, efficient computing for systems and servers.</p> <p>But there are some downsides to this diversification. A processor can only operate on the instructions it's designed to support. This means advanced processors, like GPUs and ASICs, are rarely able to perform general-purpose work. Their complexity also results in greater power use and heat generation. Consequently, processors can work well together and are frequently used in tandem to share the complex demands of modern AI workloads.</p> </section> <section class="section main-article-chapter" data-menu-title="A closer look at processor architectures"> <h2 class="section-title"><i class="icon" data-icon="1"></i>A closer look at processor architectures</h2> <p>Several computing choices can now support modern AI systems. Whether implemented on local servers or on cloud computing instances, the choice of processor configuration significantly affects AI training and inference performance. Here is an overview of the basic types of architecture for AI chips.</p> <h3>CPUs</h3> <p>CPUs are generalists. They are well-suited to most types of software applications, such as CRM platforms, databases and word processing. CPUs are also central to AI platforms as the workload manager or orchestrator, task scheduler, system component coordinator and information manager.</p> <p>However, CPUs are not strong at advanced mathematical functions and have relatively limited capabilities for the kinds of sequential or logical thinking that AI training and inference demand. This is why CPUs are a critical bottleneck for AI systems, and tasks involving complex math or massive parallel mathematical equations are delegated to advanced processors.</p> <p>CPUs can <a target="_blank" href="https://www.techtarget.com/ai/tip/How-to-train-an-LLM-on-your-own-data" rel="noopener">train AI</a> in test or development settings, but they're inadequate on their own for production AI platforms. It could take days, weeks or even months to train an AI using CPU-based computing. This might be fine for offline testing where budgets are tight, but advanced processors could complete the same task in hours. Still, CPUs remain essential AI components and work well with other advanced processors.</p> <p>CPUs are typically the least expensive computing infrastructure. Every computer has CPUs, and cloud-based CPU computing instances are many, diverse and readily accommodate most traditional software workloads.</p> <h3>GPUs</h3> <p>Where a CPU can adeptly process a few tasks sequentially, GPUs can make advanced calculations across vast amounts of data. This originally made GPUs vital in graphics tasks such as rendering, but their mathematical prowess and sheer execution scale have catapulted GPUs to prominence as <a target="_blank" href="https://www.techtarget.com/ai/definition/What-is-an-AI-accelerator" rel="noopener">AI accelerators</a>.</p> <p>GPUs are incredibly effective for a narrow range of tasks. While CPU architecture and instruction sets are intended for low latency, sequential processing and decision-making, GPU architecture and instruction sets are built for high-throughput parallel processing, where the same limited instructions can apply to vast amounts of data. GPUs offer high memory bandwidth and can move enormous amounts of data with great efficiency.</p> <p>AI model training needs to process billions of parameters across enormous data sets. Since GPUs can divide and organize these mathematical tasks across thousands of highly optimized cores, they can complete training tasks in hours. For <a target="_blank" href="https://www.techtarget.com/ai/tip/AI-inference-vs-training-Key-differences-and-tradeoffs" rel="noopener">AI inference tasks</a>, the AI can use GPUs to process new data and formulate responses to user queries almost instantly. Modern GPUs for AI tasks can also include specialized tensor cores that enhance the efficiency of ML equations.</p> <p>However, GPUs are specialized devices and are extremely poor at performing the general-purpose tasks needed to handle computer operations, such as booting a computer, running an OS, handling I/O and launching and running ordinary software applications. GPUs always have an associated CPU to manage them.</p> <p>From a cost perspective, GPUs are considerably expensive. Standard CPUs range from $100 to $600, while high-performance GPUs can cost anywhere from $300 to $2,000. In cloud computing, using a server instance with a top-tier GPU costs $1.50 to $7.00 per hour or more, whereas CPU-only instances often cost pennies per hour. The value is in time saved. Completing AI training in hours rather than months or processing an inference in milliseconds rather than hours. Advanced processors such as GPUs make modern AI systems practical.</p> <h3>ASICs</h3> <p>ASIC is a broad term that encompasses an array of "customized logic" devices. The chips that operate the office coffee maker, a bedside alarm clock, a network switch or IoT devices are typically ASICs, each custom-designed and manufactured to suit a specific purpose or task.</p> <p>ASICs fill vital computing roles that other, more traditional processors are simply too complex, power-hungry, inefficient or costly to support. Where a CPU might have billions of transistors and support hundreds of instructions, an ASIC can act as a customized processor stripped to support only a handful of key instructions with a fraction of the power and far better performance.</p> <p>ASICs have found ready acceptance as powerful processor accelerators in AI tasks. Because ASICs can accomplish a narrow set of tasks with extremely high performance and enhanced efficiency, they can perform key AI tasks, such as mathematical operations or incorporate advanced accelerator features such as NPUs or TPUs.</p> <p>However, AI ASICs are highly proprietary and are typically reserved for high-volume deployments through hyperscaler providers such as Google, AWS and Meta. An everyday enterprise wouldn't buy a server with ASIC accelerators for AI use. Instead, a hyperscaler would spearhead the design and commission the manufacture of ASIC chips, then manage the integration of those chips into custom computing systems that would be deployed across the provider's infrastructure for internal use or offered as another specialized <a target="_blank" href="https://www.techtarget.com/it-infrastructure/tip/Understanding-the-role-of-AI-in-cloud-computing" rel="noopener">AI cloud service</a>.</p> <p>On a per-chip basis, an ASIC is cheaper, delivers better performance for its intended tasks and uses far less power than a commercial processor. But the broader cost considerations for ASICs can be tricky. ASIC design can be costly, and manufacturing ASIC chips at a contract fabrication facility can represent an enormous investment for an enterprise.</p> <p>The resulting upfront investment in time and capital is huge. That's a major reason why ASICs are the realm of the largest hyperscalers that can afford them. This enormous ASIC investment also carries risk, which is why hyperscalers generally don't offer commercial versions of their complete ASIC systems. Hyperscalers treat their ASIC investments as intellectual property and a competitive advantage.</p> <h3>Field-programmable gate arrays (FPGAs)</h3> <p>While specialized chips like ASICs have a fixed, unchangeable configuration, an <a target="_blank" href="https://www.techtarget.com/ai/tip/The-growing-role-of-FPGAs-for-accelerating-AI-workloads" rel="noopener">FPGA can interconnect</a> various elements to create a chip with highly customizable functionality using major hardware description languages (HDLs) like VHDL or Verilog. This type of programmability provides two significant benefits for AI developers:</p> <ul class="default-list"> <li>The same FPGA chip can provide various AI functions in areas like edge AI, industrial IoT and processor acceleration devices in AI data centers. This means one chip can do a variety of jobs.</li> <li>The FPGA chip can be reprogrammed as required. This can be powerful when a bug is found in the chip's operation. The chip can be reprogrammed to fix the bug without making any hardware changes.</li> </ul> <p>FPGAs also provide compelling benefits for AI infrastructures. They are reconfigurable, so they can be updated, optimized or changed to perform entirely different jobs without further hardware investment. They also enable ultra-low-latency operations, processing data immediately without batching. Finally, as FPGAs are customized to the precise needs of their ML model, they typically consume far less power than generic CPUs, GPUs and other AI accelerator chips.</p> <p>An FPGA can also operate at the edge, collecting and processing data from multiple sensors and running inference on the same FPGA chip in real time. This makes FPGAs ideal for use in robotics and autonomous vehicles, medical and industrial devices and servers.</p> <p>Despite the various advantages of FPGAs, there are still numerous limitations and tradeoffs to consider. Compared to ASICs, FPGAs generally demand 20 to 35 times more silicon area, run three to four times slower, and use about 10 times more power. This tradeoff comes in the form of FPGA flexibility and programmability. When extreme performance and power savings are needed at scale, ASICs can be the better choice.<b> </b>FPGA chips also require a well-designed circuit board to accommodate their physical installation and practical use in a working AI system. Even though FPGAs can be programmed and reused, the enterprise still needs to develop and acquire circuit boards and have systems capable of incorporating those circuits. Developing, testing and validating FPGA code can also take up to 10 times longer than conventional software, requiring experienced talent.</p> <h3>NPUs</h3> <p>An NPU is a specialized processing accelerator that efficiently runs ML and AI tasks. NPUs rarely appear as separate, standalone chips but are typically integrated into the main system processor, such as a <a target="_blank" href="https://www.techtarget.com/iotagenda/definition/system-on-a-chip-SoC" rel="noopener">system-on-a-chip (SoC)</a> device. Consequently, NPUs are almost always system-centric, accelerating AI-related functionality in devices such as smartphones and laptops. This provides sophisticated AI features while emphasizing low latency and significant power savings.</p> <p>NPUs are most closely associated with endpoint devices such as smartphones, tablets and laptops, where the SoC can add an NPU to enable AI capabilities such as background blur, voice transcription, noise cancellation and image enhancement while maintaining low power consumption in these battery-powered mobile systems.</p> <p>Today, more dedicated NPUs are appearing in edge computing and server-class systems to assist with rapid, high-volume AI inference tasks. The same power savings that make NPUs attractive for mobile devices also apply to edge and data center environments.</p> <p>NPUs typically present several key limitations to consider. NPUs only support and accelerate specific AI models. When new models emerge, the AI workload must resort to GPUs, reducing AI performance. Their dedication to specific math functions also makes them useless for traditional computing tasks. NPUs can't be used to train AI models; they can only support inference tasks and supplement CPUs and GPUs as AI accelerators. Finally, NPU software is often proprietary, requiring specialized tools and skills, and is designed only for local or on-device AI processing. This makes NPUs more troublesome to integrate.</p> <p>While NPU-capable devices incur an additional acquisition cost, the reduction in API calls, server load and power utilization are typically attractive factors for NPU deployment. However, NPUs don't work unless the application is specifically designed to use them. NPUs are best used with AI-enhanced creative suites, applications requiring strict privacy for local data processing and developers building local ML models and AI systems.</p> <h3>TPUs</h3> <p>TPUs are specialized processor accelerators that optimize ML and AI tasks by supporting massive parallel processing for matrix multiplication, the principal mathematical operation behind neural networks such as Google Gemini. The TPU is a highly optimized ASIC dedicated to AI math operations, using a fraction of the power of conventional processors such as GPUs. The TPU was originally designed by Google and is now in its eighth generation with the TPU 8t for training and the TPU 8i for inference.</p> <p><a target="_blank" href="https://www.techtarget.com/it-infrastructure/news/366642002/New-Google-TPUs-multiply-AI-infrastructure-efficiency" rel="noopener">Google's TPU</a> isn't readily available for local systems or private data centers. Instead, TPUs are accessed through Google Cloud as a service, where they can be used to train massive AI models, support post-training and reinforcement learning, enable high-performance inference and perform powerful data analytics.</p> <p>While TPUs provide powerful acceleration for matrix multiplication tasks, they can't be used for general computing and offer limited benefits for workloads not focused on matrix math. TPUs are a poor choice for precision scientific computing because TPUs sacrifice precision for performance. Furthermore, TPUs depend on Google's software stack, requiring additional skills and investment to build, refine and deploy AI apps that use them. The fact that TPUs are a Google Cloud service introduces a degree of vendor lock-in that some businesses might find undesirable. TPUs offer high bandwidth memory, but limited memory space can impose bottlenecks, and they can provide reduced latency for non-batch tasks, such as a single prompt.</p> <h3>Language Processing Units (LPUs)</h3> <p>LPUs are another specialized processor accelerator that assists with AI inference tasks, focusing primarily on token generation for LLMs. LPUs rely on high-speed local memory to keep model weights readily accessible for extremely low-latency and decision-making operations. Because LPUs focus on token generation, they are intended for inference and not suited for AI model training.</p> <p>LPUs such as OpenAI's Jalapeno chip offer predictable performance and minimize the lag spikes often found in server-based language models. This makes LPUs excellent for real-time AI agents and chatbots, financial data processing and market trading, real-time voice transcription for healthcare and voice assistance, educational and adaptive learning environments, and <a target="_blank" href="https://www.techtarget.com/it-strategy/feature/Can-edge-computing-make-AI-more-sustainable" rel="noopener">edge computing tasks</a>.</p> <p>The strength of an LPU is in its massive onboard memory and fast, predictable operation. Still, business leaders should consider the numerous limitations of LPUs. Onboard memory can be a constraint. Even though LPUs offer significant onboard memory, this finite resource can be exhausted by large models, requiring LPU clusters. Furthermore, LPUs rely on static execution pathways that must be carefully planned and meticulously implemented in software. Compiling LPU software to drive a new or updated model can take days, making LPUs inadequate for testing or experimentation.</p> <p>Ultimately, LPUs can be the best choice when processing smaller models at sub-millisecond token speeds, when high efficiency and throughput with predictable token decoding are needed, and when the fastest AI responses are required.</p> <p>As with TPU chips, LPU chips are typically unavailable individually. Instead, they're offered as integrated components of large, enterprise-class systems such as the Nvidia Groq 3 LPX Rack, which packs 256 interconnected LPU accelerators with the Vera Rubin platform, or as a cloud service such as GroqCloud, which is a generally more cost-effective, pay-as-you-go option for business users.</p> </section> <section class="section main-article-chapter" data-menu-title="Selecting an AI processor architecture"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Selecting an AI processor architecture</h2> <p>Selecting processor architecture for AI projects can be a mission-critical decision. Mistakes can be costly in performance and investment. The final choices should involve executives and their technical team. The selection criteria typically involve several broad categories, including the following:</p> <ul class="default-list"> <li><b>AI workload type.</b> AI training builds models from vast data sets and requires massive parallel processing, high memory bandwidth and high scalability. This makes GPUs the best overall choice for AI training. By contrast, AI inference runs trained models on new data and benefits from extremely low latency, rapid token generation and power efficiency. This makes LPUs best for LLM inference, while ASICs or TPUs are better for more general inference work.</li> <li><b>AI model maturity.</b> Consider the stability or maturity of the models. Processor accelerators can be powerful, but often only accelerate a specific model's code. Changing the model can reduce the effectiveness of high-end accelerators like ASICs or FPGAs, resulting in lower performance or an outright fallback to a more general accelerator. Invest in top acceleration only when a model is fully vetted and ready for reliable production use. Models in development or subject to frequent change are best supported by general GPUs or reprogrammable FPGAs.</li> <li><b>AI deployment environment.</b> AI processors can be deployed at the edge or in local or cloud data centers. Edge deployments include laptops, smartphones and IoT devices and are best served by low-power NPUs. Data center deployments are large-scale deployments that offer raw compute power for the heaviest AI tasks. This can include various accelerators, such as GPUs and cloud provider-based offerings such as TPUs, LPUs, FPGAs and ASICs. Make sure the deployment environment best matches the workload.</li> <li><b>Programming and development skills.</b> Choosing a processor isn't enough. Businesses need the coding and toolchain skills to back it up. Carefully evaluate the coding skills needed to support the processor choice. FPGAs and ASICs demand extensive and highly specialized coding expertise. GPUs need more general, well-standardized coding. Cloud instances require solid experience with the provider's APIs and nuances.</li> <li><b>Decide on meaningful performance metrics.</b> Choose and implement meaningful model performance metrics and have the tools in place to track them to verify actual performance and measure improvements once processor accelerators are implemented. This provides cost justification and lets the business recognize performance degradation or unexpected reductions in model health, which might need scaling, optimization or troubleshooting.</li> <li><b>Effect on local data centers.</b> Don't overlook the secondary effects of AI processor deployments on local data center operations. New servers mean changes to cooling and rack space demands. High-efficiency processor accelerators can save power per chip or per core, but are used in such large numbers that overall data center power consumption increases dramatically. Other factors include networking changes and vast storage resources. Consider the resources needed to implement processor accelerators in the data center, and the associated costs.</li> </ul> <p><em>Stephen J. Bigelow, senior technology editor at TechTarget, has more than 30 years of technical writing experience in the PC and technology industry.</em></p> </section>
Vibe coding has changed how I run my business. Here’s how I use it to automate repetitive work, build useful tools, launch ideas faster and get more done without hiring a development team.
From AI coding agents to TypeScript, Python and cloud-native development, these programming trends could shape software development and developer skills in 2027.
Climate tech pioneer offers energy leaders hands-on, in-field access to utility-scale Earth Mount Solar technology BAY CITY, Texas, Aug. 31, 2026 /PRNewswire/ -- Erthos, Inc., the creator of Earth Mount Solar, today announced its role as a premier sponsor of Houston Energy & Climate Week...