Category: AI Chips

  • Why Tech Giants Are Developing Custom AI Chips to Cut Costs and Boost Performance

    Why Tech Giants Are Developing Custom AI Chips to Cut Costs and Boost Performance

    Nvidia’s AI GPUs have become a cornerstone of the AI infrastructure boom, but their high cost is driving major technology companies to develop custom chips. Purpose-built AI silicon can be tailored for specific workloads, helping companies control expenses and improve performance at scale.

    Google, Amazon, Microsoft, and Meta are all creating their own AI accelerators to reduce reliance on merchant GPUs. Google uses TPUs, Amazon has Trainium, Microsoft develops Maia, and Meta has MTIA. These chips are designed around each company’s unique AI workloads and data center requirements.

    The primary benefit is lower inference cost. Custom ASICs can be optimized for specific AI tasks rather than supporting every possible workload. Amazon reports that its Trainium chips can enhance price-performance, while its latest Trainium3 systems target reduced operating costs for large AI workloads.

    Building custom AI chips also diversifies the supply chain. Broadcom and Marvell help turn hyperscaler chip designs into production-ready ASICs, while Arm provides CPU architecture and licensing. This process connects chip design with advanced manufacturing and packaging.

    Memory and networking are equally critical. AI accelerators require high-bandwidth memory and fast inter-chip connections. Companies such as Micron, Astera Labs, Credo, and Arista operate across memory, connectivity, Ethernet, and data-center networking needed for large AI systems.

    Manufacturing and advanced packaging remain essential parts of the AI chip ecosystem. TSMC is a major foundry for advanced AI chips, while CoWoS packaging supports high-performance accelerator systems. Samsung also provides semiconductor manufacturing capacity. The supply chain extends from wafers to HBM and packaging.

    The AI chip race does not end with the processor. Higher-density AI racks require advanced networking, optical connections, power systems, and cooling. Companies across these areas support the growing infrastructure. Custom AI silicon is therefore becoming a full ecosystem spanning chips, memory, factories, networking, power, and cooling.

  • Snapdragon X2 AI PCs: Five Capabilities Set to Reshape Everyday Computing

    Snapdragon X2 AI PCs: Five Capabilities Set to Reshape Everyday Computing

    Qualcomm’s Snapdragon X2 generation is pushing AI PCs beyond marketing language. The platform pairs dedicated neural processing with efficient CPUs and GPUs, aiming to make AI-assisted work feel more native on laptops. Here are five areas where the new Snapdragon AI PCs are expected to make a difference.

    1. More capable on-device AI

    The centerpiece is the NPU. Snapdragon X2 Elite processors are listed with up to 85 TOPS of NPU performance in some configurations, while other listings cite 80 TOPS. The NPU works alongside the CPU and GPU, so supported apps can run AI workloads locally instead of sending every request to the cloud. That opens the door to multimodal tasks involving text, images, video, and audio, with faster responses and smoother AI tools when software is optimized.

    2. AI agents that can handle multi-step tasks

    Snapdragon X2 platforms are built to support agentic AI experiences. Rather than only answering a question, supported applications can break a user’s goal into steps and use available tools to carry out parts of the workflow. Qualcomm has demonstrated native agentic apps for document search, content creation, web research, and workflow automation. The caveat: real-world usefulness will depend on developer support and app integration.

    3. Battery life and efficient processing

    Performance matters only if a laptop can sustain it. Snapdragon PCs continue to emphasize power efficiency, and Qualcomm says Snapdragon X Series devices can reach up to 22 hours of battery life. Actual mileage will vary by hardware, software, display, and workload. The architecture spreads tasks across processing units, so AI work can use the NPU instead of overloading the CPU. That can help laptops stay responsive while cutting unnecessary power draw.

    4. Smarter cameras, audio, and connectivity

    AI features are also moving into everyday communication. Snapdragon platforms support AI-enhanced camera and audio processing, which can improve video calls, background effects, voice handling, and related tools. Snapdragon X2 systems include Wi-Fi 7 support and advanced display capabilities, with some configurations able to drive up to three 5K displays at 60Hz. That makes these PCs relevant for mobile productivity and workstation-style setups.

    5. Local processing with more privacy

    On-device AI can offer more than speed. When suitable tasks are processed locally, less sensitive data has to leave the device. Qualcomm also highlights chip-to-cloud security across Snapdragon X2 platforms. This does not remove the cloud from the picture. Demanding workloads can still use cloud services, while the PC follows a hybrid model: local processing for appropriate tasks and cloud capacity when needed.

    What the next Snapdragon AI PCs could mean for users

    The next generation of Snapdragon laptops is becoming more AI-focused. Dedicated NPUs are getting faster, and software is catching up. Agentic AI may be one of the most noticeable shifts, moving AI tools from simple Q&A toward workflow assistance. Battery efficiency remains a major advantage for mobile users, and local processing could make some AI experiences faster and more private. Still, hardware specifications alone do not guarantee a better AI experience. The final result will depend on the device, the applications, and how well developers use the platform.

    FAQ

    What are Snapdragon AI PCs?

    They are computers powered by Qualcomm Snapdragon platforms with dedicated AI processing. Their NPUs can handle supported AI workloads alongside the CPU and GPU, enabling local AI features and AI-assisted applications.

    What is the Snapdragon X2 Elite NPU?

    It is a dedicated neural processing unit for AI workloads. The article cites up to 80 TOPS of NPU performance, with some configurations listed at up to 85 TOPS. Actual capabilities depend on the specific device and software.

    What can on-device AI do on Snapdragon PCs?

    On-device AI can process supported workloads directly on the computer, including tasks involving text, images, audio, and other content. This can reduce dependence on cloud processing for suitable workloads.

    What is agentic AI on a Snapdragon PC?

    Agentic AI goes beyond simple question-and-answer interactions. Supported applications can break a user’s goal into multiple steps and use applications and tools to complete parts of that workflow.

    Do Snapdragon AI PCs have better battery life?

    Snapdragon PCs emphasize power efficiency alongside performance. Qualcomm says Snapdragon X Series devices can deliver up to 22 hours of battery life. Actual battery performance varies by device, software, workload, display, and usage conditions.

  • AI Reshapes Chip Industry: Top Semiconductor Companies by Market Cap

    AI Reshapes Chip Industry: Top Semiconductor Companies by Market Cap

    AI’s Impact on Semiconductor Valuations

    Artificial intelligence is fundamentally transforming how investors evaluate semiconductor companies. The industry now spans GPUs, memory, foundries, networking chips, manufacturing equipment, and advanced processors. Recent market-cap data underscores this shift, with several chipmakers reaching historic valuations as demand for AI infrastructure expands across data centers and increasingly complex computing workloads worldwide.

    Nvidia Leads the Pack

    Nvidia dominates the semiconductor rankings with a market capitalization of approximately $5.33 trillion. Its position reflects surging demand for accelerated computing, AI processors, and data center infrastructure. The company’s valuation is substantially larger than other listed semiconductor businesses, highlighting the financial importance of AI computing platforms and supporting technologies.

    TSMC Powers Advanced Manufacturing

    TSMC ranks second with a market cap of about $2.24 trillion. As the leading contract chip manufacturer for major technology companies, TSMC’s strategic importance has grown with the need for advanced processors — especially energy-efficient chips that can handle demanding AI workloads.

    Broadcom’s Networking Edge

    Broadcom appears among the largest with an estimated market capitalization of $1.72 trillion. Its portfolio includes networking and connectivity technologies that support large-scale computing infrastructure. As AI data centers expand, demand is rising not only for processors but also for high-speed networking and specialized components that connect complex systems.

    Memory Plays a Big Role

    Samsung and Micron hold market caps of approximately $1.20 trillion and $1.12 trillion respectively, while SK Hynix follows at around $907 billion. Their positions reflect the growing importance of memory technologies, particularly high-bandwidth memory (HBM), which is essential for demanding AI workloads and large-scale accelerated computing systems.

    Other Key Players

    AMD is shown at approximately $894 billion, ASML reaches $633 billion, CXMT stands at $565 billion, and Intel follows at about $547 billion. This ranking demonstrates that AI-related value extends beyond GPU manufacturers, encompassing processors, manufacturing equipment, and memory — all critical to producing advanced chips.

    AI Lifts the Entire Stack

    The semiconductor market is becoming increasingly interconnected as AI accelerates demand across multiple layers. Processor designers, memory manufacturers, foundries, networking companies, and equipment suppliers all support expanding AI infrastructure. The market-cap figures illustrate how investors are assigning significant value across this ecosystem, reflecting expectations for continued computing demand and technological investment.

  • Qualcomm’s Snapdragon X2 Powers New Microsoft Surface PCs and Expands to Googlebook and Linux

    Qualcomm’s Snapdragon X2 Powers New Microsoft Surface PCs and Expands to Googlebook and Linux

    Qualcomm is expanding its Snapdragon X platform beyond the initial Windows-on-Arm push, bringing the technology into new Microsoft Surface devices, Google’s emerging Googlebook category, and a wider range of PC form factors.

    Announcements at Snapdragon Summit 2026 reveal Qualcomm’s strategy to establish its Oryon-based platform across different operating systems and device segments. Microsoft is introducing new Surface hardware with Snapdragon X2 Plus, while Google has partnered with Qualcomm for Googlebook laptops powered by Snapdragon X Elite.

    New Surface Devices with Snapdragon X2 Plus

    Microsoft unveiled a 12-inch Surface Pro and 13-inch Surface Laptop, both powered by the six-core Snapdragon X2 Plus processor. The devices go on sale in select markets from October 13. According to Microsoft, the new machines deliver up to 17% faster Office performance, up to 18% better battery efficiency, more than 60% faster graphics, and up to 95% faster local AI inference than the previous generation. The Snapdragon X2 Plus features an 80-TOPS NPU designed to handle on-device AI workloads.

    Googlebook Takes Snapdragon Beyond Windows

    A significant development is Google’s adoption of Snapdragon X for its new Googlebook category. Unlike conventional Chromebooks, Googlebook laptops are positioned around Gemini-powered features, native Android applications, and closer integration with Android phones and tablets. The first Googlebook models from Dell and HP use Snapdragon X Elite rather than the newer X2 generation. The platform supports thousands of native Android applications and games, expanding Qualcomm’s reach beyond Microsoft’s ecosystem.

    X2 Elite Pushes Performance Boundaries

    At the high end, Snapdragon X2 Elite scales to 18 Oryon CPU cores, with selected versions reaching speeds up to 5GHz. Qualcomm lists up to 85 TOPS of NPU performance and as much as 228GB/s of memory bandwidth. The company is also extending Snapdragon X2 support to Linux, adding another operating system to its PC strategy alongside Windows and Googlebook.

    Qualcomm Faces Wider Competition

    The expansion comes as Qualcomm faces competition from Intel, AMD, and Apple across premium and AI-focused PCs. Snapdragon X must compete on battery life, software compatibility, graphics performance, AI workloads, and developer support. With Surface, Googlebook, and Linux now joining Windows in the Snapdragon ecosystem, Qualcomm’s PC strategy is becoming broader. The next test will be whether manufacturers and users adopt the platform across these different categories.

  • AMD Surpasses $1 Trillion Market Capitalization as AI Chip Demand Fuels Growth

    AMD Surpasses $1 Trillion Market Capitalization as AI Chip Demand Fuels Growth

    AMD Reaches $1 Trillion Market Cap

    Advanced Micro Devices (AMD) crossed the USD 1 trillion market capitalization mark for the first time, as CEO Lisa Su’s push into artificial intelligence continues to drive investor interest. AMD shares touched a record price of USD 615.52 on September 21 after rising nearly 10% in a single session. This makes the company the fourth American chipmaker to enter the trillion-dollar club after NVIDIA, Broadcom and Micron. AMD’s stock has gained 185% this year and nearly 300% over the past 12 months.

    “AI is driving a significant expansion in demand for compute across all of our markets,” Su said in AMD’s second-quarter results statement.

    Data Center Business Drives Growth

    AMD reported a 50% increase in second-quarter revenue to a record USD 11.5 billion. Data center sales rose 107% to USD 6.7 billion, accounting for 58% of total revenue. The company has guided for about USD 13 billion in third-quarter revenue and expects data center revenue to more than double in 2027. AMD also expects server revenue to grow more than 80% in the second half of 2026.

    “Demand for both accelerators and CPUs is growing well above our prior expectations,” Su told analysts on the earnings call.

    Major Deals with Anthropic, OpenAI and Oracle

    Major deals with technology companies have supported AMD’s AI strategy. Anthropic has agreed to deploy up to 2GW of AMD Instinct MI450 GPUs in Helios, with the first gigawatt expected in the first half of 2027. AMD has committed up to USD 5 billion in equity investment in Anthropic. OpenAI signed a 6GW supply agreement with AMD last October, while Oracle is building a publicly available cluster featuring 50,000 MI450 GPUs.

    AMD Still Trails NVIDIA in AI Accelerators

    NVIDIA remains far ahead in the AI accelerator market despite AMD’s rapid growth. It is currently valued at about USD 5.4 trillion and holds an estimated 81% market share, compared with AMD’s 13%. AMD’s third-quarter results will be its next major update. The company has guided for nearly USD 13 billion in revenue and a 56% non-GAAP gross margin, as investors assess whether its AI expansion can support its trillion-dollar valuation.

  • Samsung zHBM: Stacking Memory for Faster AI Workloads

    Samsung zHBM: Stacking Memory for Faster AI Workloads

    AI accelerators are becoming increasingly powerful, but moving data between memory and processors can create bottlenecks. Samsung’s zHBM concept takes a different approach by vertically stacking high-bandwidth memory directly above an AI accelerator to shorten the data path.

    Memory Moves Closer

    Traditional HBM sits alongside the processor within advanced packages. Samsung’s zHBM concept places memory vertically above the AI accelerator. Shortening the distance between the two components is designed to improve data movement and support faster processing for demanding AI workloads.

    Higher Bandwidth Is the Goal

    Large AI models constantly move huge amounts of information between compute and memory. Samsung says its next-generation interface system incorporating zHBM is expected to deliver approximately eight times the performance of HBM5, targeting substantially faster AI data processing.

    Density Could Jump

    Samsung says zHBM could achieve more than 10 times the memory density of HBM5 using next-generation wafer-bonding technology. Higher density could allow AI systems to access more memory within compact architectures, potentially supporting increasingly large models and data-intensive workloads.

    Power Efficiency Matters

    Faster AI processing is not useful if energy consumption rises uncontrollably. Samsung says zHBM could improve energy efficiency by three times while reducing thermal resistance by more than half. Those improvements could help address power and heat challenges in high-performance AI systems.

    Custom AI Designs Are Possible

    Samsung says zHBM can support customer-specific designs by allowing customised intellectual property to be integrated into the interlayer between memory and an AI accelerator. This could give customers more flexibility to tailor memory capacity and accelerator performance for specific workloads.

    A Future AI Memory Architecture

    zHBM remains a next-generation concept rather than a mainstream memory product today. However, its vertically integrated design shows where AI hardware could be heading: closer memory, higher bandwidth, greater density and improved efficiency as models continue demanding more computing resources.

  • Alibaba Unveils Zhenwu V900 AI Chip to Power Domestic AI Ecosystem and Massive Data Center Expansion

    Alibaba Unveils Zhenwu V900 AI Chip to Power Domestic AI Ecosystem and Massive Data Center Expansion

    Alibaba has taken a significant step in advancing China’s domestic AI capabilities with the unveiling of its latest artificial intelligence chip, the Zhenwu V900. Developed by Alibaba’s semiconductor unit T-Head, the new accelerator is designed to handle both AI training and inference tasks, marking a key milestone in the company’s push to build a full-stack AI ecosystem.

    According to Alibaba, the V900 delivers three times the performance of its predecessor, the Zhenwu M890 chip, which was launched in May. While the company did not provide independent benchmark results to support this claim, the V900 is scheduled to enter mass production and commercial release in the first quarter of 2027.

    Zhenwu V900 Specifications

    The V900 comes equipped with 216GB of memory and offers 1,200GB/s of chip-to-chip bandwidth. It also supports FP8 and FP4 precision formats, which are becoming increasingly critical for efficient AI inference and large-scale model workloads.

    Alibaba stated, “The chip can be connected into clusters containing as many as 500,000 accelerators. This architecture is designed to handle demanding AI workloads that require thousands of processors to work together.”

    The company is also integrating the V900 with its broader hardware stack, including networking and storage components, allowing Alibaba to control more of the infrastructure required to train and deploy large AI models. This strategy directly links Alibaba’s chip development with its AI model roadmap, as more powerful accelerators provide the computing capacity needed for increasingly large models.

    Alibaba Expands AI Infrastructure

    Alibaba is also planning a major expansion of its cloud infrastructure. CEO Eddie Wu stated, “Alibaba Cloud aims to increase its global data-center capacity to more than 20 gigawatts by 2032.”

    The V900 highlights Alibaba’s effort to develop chips, AI models, and cloud infrastructure as one integrated ecosystem. This initiative comes as Chinese technology companies accelerate domestic AI chip development amid restrictions on access to advanced foreign processors.

    The tech giant further explained that demand for AI computing is growing rapidly while supply constraints limit the pace of expansion. Alibaba also plans to bring AI supernodes into commercial operation this quarter.

  • AMD Reaches $1 Trillion Market Cap on Surging AI Chip Demand

    AMD Reaches $1 Trillion Market Cap on Surging AI Chip Demand

    AMD’s market value surpassed $1 trillion for the first time, driven by surging demand for AI chips and record data center sales.

    The chipmaker’s share price jumped approximately 10% on September 21, reaching a record high of $615.52. The stock gained about 25% in five trading sessions as investors focused on AMD’s AI chip and data center growth.

    Strong Financial Performance

    AMD reported second-quarter revenue of $11.54 billion, up 50% year-over-year. Data center sales reached $6.7 billion, more than double the figure from the same period last year. CEO Lisa Su stated that the company expects to double its data center sales again in 2027.

    AI Chips Drive Growth

    The growth is not just from traditional PC and server processors but increasingly from AI chips. Companies are heavily investing in AI computing, requiring more chips to run these systems. AMD is capitalizing on this demand to expand its data center business and compete for a larger share of the AI hardware market.

    AMD is shifting its strategy from selling individual chips to offering complete systems that include processors, AI accelerators, networking equipment, and other hardware. This approach provides customers with a comprehensive solution for building and running AI infrastructure.

    Market Position

    The $1 trillion milestone underscores how AI has transformed AMD’s business. However, the company still trails behind NVIDIA, which has a market valuation of over $5 trillion. AMD will need to refine its strategies to become a stronger competitor in the AI chip market.

  • AI Rally Propels Nasdaq to Record Close, AMD Joins Trillion-Dollar Chip Club

    AI Rally Propels Nasdaq to Record Close, AMD Joins Trillion-Dollar Chip Club

    Wall Street opened the week with a strong tech rally. The Nasdaq Composite jumped 2.26% to 27,122.09 points, setting a new record close and marking its first record high since June 2. The S&P 500 rose 1.49% to 7,764.70 points, while the Dow gained 0.71%.

    Notably, chip stocks led the rise. AMD jumped nearly 10% and crossed a USD 1 trillion market value for the first time. Intel gained 12.2%, Arm Holdings rose 10% and Meta climbed 11.4%.

    The gains were not limited to stocks. Bitcoin rose more than 6% to a fresh seven-month high. Falling oil prices and lower US Treasury yields also helped markets. Brent crude settled at USD 100.34 a barrel, while the 10-year Treasury yield fell below 5%. Investors are still watching US interest rates, Middle East developments and the upcoming US-China meeting.

    AMD was one of the biggest names in Monday’s rally. Its shares have gained about 25% in five trading sessions. The move shows that investor interest in AI stocks is picking up again after concerns about high AI spending had cooled the market.

    AMD now joins NVIDIA, Broadcom and Micron as the fourth US chipmaker to reach a USD 1 trillion valuation. NVIDIA remains far ahead at more than USD 5 trillion. The growing list shows how much value investors are placing on chips as AI use expands across data centers and other services.

    AMD’s milestone is another sign that AI is reshaping the chip industry. For investors, the focus is no longer only on software or AI models. Companies building the chips and systems behind them are also drawing major market attention.

  • Alibaba Stock Rallies on New AI Chip and Flagship Model as China’s AI Ambitions Intensify

    Alibaba Stock Rallies on New AI Chip and Flagship Model as China’s AI Ambitions Intensify

    Alibaba shares rose in premarket trading on Tuesday after the Chinese technology giant unveiled its latest artificial intelligence chip and its most powerful flagship AI model to date. The announcements mark another step in China’s efforts to close the gap with US tech leaders, though significant challenges remain.

    Alibaba Unveils Its Most Powerful AI Chip

    The company introduced the Zhenwu V900, which it described as its most powerful AI chip to date. Alibaba said the new chip delivers three times the performance of its predecessor, which was released just four months ago. Alongside the chip, Alibaba also launched Qen3.8-Max, its most powerful flagship AI model so far.

    Data-Center Expansion Plans

    Alibaba CEO Eddie Wu said the company expects significant growth in annual AI chip shipment volumes. He also pointed to shortages across the global AI data-center supply chain as a constraint on the pace at which Alibaba can expand its computing infrastructure. Wu said Alibaba would work with partners to address these challenges.

    Alibaba Cloud is targeting more than 20 gigawatts of global data-center capacity by 2032. The company sees expanding computing infrastructure as a core part of its broader AI strategy as demand for AI services continues to grow.

    Alibaba Still Trails US AI Leaders

    Despite the latest announcements, Alibaba’s AI infrastructure still lags behind leading US companies. The Zhenwu V900 is considered competitive with NVIDIA’s Hopper hardware, which is now two generations old. The data-center gap is also significant: one major US player has said it intends to have enough power and cooling to support 20 gigawatts of computing capacity by the end of next year.

    Parv Sharma, senior analyst at Counterpoint Research, said China’s AI infrastructure is more supply-constrained because AI has become a strategic priority. He noted that Alibaba’s 20-gigawatt target by 2032 should be viewed against the scale already operated by US hyperscalers.

    US-China Talks Could Shape AI Competition

    Investors are also watching talks between US President Donald Trump and Chinese President Xi Jinping, particularly for any developments affecting AI technology and access to advanced chips. According to Barron’s, China’s ability to close the gap with the US could remain constrained unless access to cutting-edge chips improves.

    Before Tuesday’s trading session, Alibaba’s American depositary receipts had fallen 21% so far this year.