Tag: AI chips

  • TSMC to Manufacture OpenAI’s AI Chips, Not Samsung Foundry

    TSMC to Manufacture OpenAI’s AI Chips, Not Samsung Foundry

    Key Takeaways

    1. AI-powered tools are attracting significant investor interest, but developing them can be costly.
    2. OpenAI is seeking to reduce operating expenses by creating its own hardware instead of relying on Nvidia.
    3. Samsung Foundry was considered for producing AI hardware, but OpenAI has chosen TSMC’s 3nm process instead.
    4. OpenAI plans to mass-produce its AI chips by 2026, with designs nearing completion.
    5. OpenAI is investing $500 million in its proprietary AI chip, aiming for long-term savings in operational costs.

    Development of AI-powered tools appears to be a profitable venture. An increasing number of investors are keen on backing these kinds of projects. Nonetheless, creating AI products can be pretty costly if one aims to compete with top players in the market. OpenAI, which is the parent organization of ChatGPT, is fully aware of these financial challenges, prompting them to seek ways to reduce operating expenses.

    OpenAI’s Strategy for Cost Reduction

    As per various reports, one of the strategies OpenAI intends to adopt for long-term savings involves crafting its own hardware to manage AI services. At present, the company relies on Nvidia, which holds a dominant position as the leading supplier of AI hardware globally. Nevertheless, Nvidia’s stronghold allows it to dictate prices that some companies, including OpenAI, find excessive.

    Samsung Foundry vs. TSMC

    Samsung Foundry has surfaced as a key potential producer for OpenAI’s AI hardware after a dialogue took place between Samsung Electronics Chairman Jay Y. Lee and OpenAI CEO Sam Altman during the week of February 3 to February 9. Some insiders hinted that the production of OpenAI’s AI chips using Samsung’s 3nm process was among the matters discussed. However, a recent report from Reuters suggests that OpenAI has opted for TSMC’s 3nm process instead.

    Future Plans and Investments

    The AI-centric firm won’t be the first significant player to part ways with Samsung after encountering issues and dissatisfaction with its wafer performance. Other companies like Qualcomm and Nvidia have also transitioned to TSMC for their needs. OpenAI is said to aim for the mass production of its AI chips by 2026. In the near future, TSMC might receive designs from OpenAI to kick off production tests, with reports indicating that the hardware design is nearing completion.

    OpenAI plans to invest approximately $500 million in the creation of its proprietary AI chip. Though this initial expense seems steep, the long-term savings in operational costs could be substantial. Apple previously took a similar path by moving away from Intel in favor of its own ARM chips for the Mac lineup.

    Source:
    Link

    Tags: , ,
  • How Smartphone Chips Fuel the AI Revolution on Devices

    How Smartphone Chips Fuel the AI Revolution on Devices

    The rapid shift towards on-device AI is changing the landscape of smartphones, transforming them into robust centers for artificial intelligence applications. Thanks to improvements in chip technology, companies are fine-tuning processors so they can execute AI functions directly on the device. This shift allows for quicker, safer, and more effective user experiences.

    The Role of AI-Focused Chips

    The surge in on-device AI is largely attributed to the adoption of specialized AI chips in smartphones. Top chip manufacturers like Qualcomm, MediaTek, Samsung, and Apple are incorporating Neural Processing Units (NPUs) into their System-on-Chips (SoCs). NPUs are specifically designed to speed up machine learning processes while using less energy compared to conventional CPUs or GPUs.

    Advanced Processing Power

    For example, Qualcomm’s Snapdragon 8 Elite and MediaTek’s Dimensity 9400 showcase advanced NPUs that can execute trillions of operations each second (TOPS). These powerful processors support real-time AI capabilities, including generative AI for text, images, and videos, all while ensuring high energy efficiency.

    Energy Efficiency Matters

    By leveraging low-precision arithmetic along with various other strategies, these chips dramatically lower power usage, which is a vital aspect for mobile devices, especially where battery longevity is essential.

    Tags: , ,
  • Nvidia CEO Visits Beijing Employees Instead of Trump’s Inauguration

    Nvidia CEO Visits Beijing Employees Instead of Trump’s Inauguration

    Nvidia’s CEO, Jensen Huang, recently made a trip to China to celebrate the Chinese New Year with his employees and to restate the firm’s dedication to its team and tech partnerships. This visit takes place against the backdrop of ongoing geopolitical tensions between the US and China, as well as export limits on advanced AI chips. Huang’s goal was to boost the spirits of the staff and underscore Nvidia’s commitment to the tech landscape in China. This visit underscores how crucial it is to uphold global teamwork to promote technological advancements, even when geopolitical issues are on the rise.

    Celebrating Employee Loyalty

    During his stay, Huang engaged in several events at Nvidia’s offices in Shenzhen and Beijing, including a Spring Festival party in the capital. He praised the team and highlighted Nvidia China’s impressively low turnover rate of 0.9%, which is significantly lower than the global average of 2%. Over the years, Nvidia’s workforce in China has expanded more than 50%, now numbering nearly 4,000 employees. Huang appreciated the dedication of the employees and their substantial role in Nvidia’s global achievements.

    China’s Contribution to Nvidia

    China plays an essential role for Nvidia, bringing in $5.4 billion in revenue for Q3 FY2024, marking a 34% increase from the previous year. The region represents 17% of Nvidia’s total income, making it one of the company’s major markets after the US and Singapore. Nvidia has partnered with over 3,000 Chinese startups and supports 1.5 million local developers utilizing CUDA, its unique AI programming platform.

    Navigating Geopolitical Tensions

    Huang’s trip coincided with stricter US export controls on AI chips directed at China. Nvidia has opposed these limitations, cautioning that they could damage innovation and the technological leadership of the US. In retaliation, China initiated an antitrust probe into Nvidia. Despite these obstacles, Huang reaffirmed Nvidia’s commitment to the Chinese market, calling the company’s 25-year journey in China a privilege.

    Huang’s visit also sparked rumors regarding the possible establishment of Nvidia’s Asia-Pacific headquarters, with Taiwan being a potential site. As China continues to aim for technological independence, Huang stressed Nvidia’s role in modernizing the industry and confirmed the company’s commitment to fostering innovation and collaboration.

    Tags: , ,
  • Nvidia Slams US Last-Minute Restrictions on AI Chip Exports

    Nvidia Slams US Last-Minute Restrictions on AI Chip Exports

    Nvidia is strongly against the forthcoming chip export limitations from the White House, characterizing them as a hurried initiative that might exceed its intended objectives. The anticipated regulations—expected to be announced imminently—aim to establish a three-tiered system to manage AI chip exports, tailored both to specific countries and companies.

    Concerns Over Policy Impact

    Ned Finkle, who leads Nvidia’s government relations, expressed his worries regarding the extensive implications of this policy. He stated that the proposed “extreme country cap policy” would negatively affect everyday computing worldwide without genuinely enhancing national security. These new restrictions would particularly influence AI accelerators, an area where Nvidia holds a dominant position.

    Proposed Access Levels

    According to the suggested rules, American semiconductors would be allocated based on different access categories. Some of the U.S.’s closest allies would receive unrestricted import permissions, while many other countries would face new limitations on overall computing capacity. These restrictions would not only target specialized AI chips but also general-purpose GPUs utilized in various devices, including gaming PCs and data centers.

    Timing and Economic Concerns

    The timing of this announcement is significant, occurring less than two weeks prior to the presidential changeover. “This last-minute policy from the Biden administration could become a legacy that will draw criticism from both the U.S. industry and the global community,” Finkle warned, cautioning that it may damage American economic interests.

    Meanwhile, Nvidia’s CEO, Jensen Huang, has expressed his willingness to work with the new administration coming in. He showed interest in meeting with Trump and even offered his assistance. Speaking at CES in Las Vegas, Huang seemed optimistic about the possibility of reduced regulations under Trump, stating, “As an industry, we want to move fast.”

    Market Implications for Nvidia

    These new regulations could significantly impact Nvidia’s position in the market, especially considering the company’s remarkable growth: its stock price nearly tripled last year, following a 239 percent increase in 2023, largely fueled by a rise in AI investments.

    Source:
    Link

    Tags: , ,
  • Run:ai Joins Nvidia, Announces Open-Source Software Plans

    Run:ai Joins Nvidia, Announces Open-Source Software Plans

    AI chip giant Nvidia has finalized the purchase of Israeli AI start-up Run:ai, after getting the green light for the merger from the European Commission.

    Nvidia had initially shared its intentions to buy the start-up back in April, but the deal faced a review from the European Commission, which approved it just this month. The orchestration software firm stated that it "will keep assisting customers in maximizing their AI Infrastructure."

    Open-source Ambitions

    Run:ai has expressed its intention to make its software stack open-source "to support the community in building better AI, quicker." At present, Run:ai’s workflows are compatible only with Nvidia GPUs, but the firm is optimistic that this will evolve once the software is made available "to the whole AI ecosystem."

    Market Share and Scrutiny

    As per Nasdaq, Nvidia commands an 80% stake in the rapidly growing data center market for AI chips. However, the company is also facing intense scrutiny from lawmakers regarding potential monopoly issues.

    Earlier this month, China’s Administration for Market Regulation revealed it is conducting an investigation, and the US Commerce Department is also looking into how Nvidia’s chips made their way to China despite existing restrictions. Furthermore, the US Department of Justice is delving into the company following complaints from competitors.

    Reuters | Nvidia | Run:ai | Nasdaq

    Source: Link

    Tags: , ,
  • AMD Job Cuts: Company Focuses on AI Development and Strategy

    AMD Job Cuts: Company Focuses on AI Development and Strategy

    AMD has made the decision to cut its workforce by 1,000 employees globally. This semiconductor firm, often viewed as a key competitor to Nvidia, shared an earnings report for Q3 in September that showed mixed results.

    Financial Performance Insights

    According to a report from Reuters, AMD saw its revenue in the data center sector, which includes AI chips, increase more than double in the last quarter. Additionally, the personal computer segment experienced a growth of 29%. However, the gaming division faced a significant drop, with sales falling by 69%.

    Future Projections

    The London Stock Exchange Group (LSEG) has predicted that AMD’s data center segment will grow by 98% in 2024, which greatly surpasses the expected 13%. Even with this growth, AMD still trails behind its rivals Nvidia and Intel in the AI chip market. In the first quarter of 2024, Nvidia maintained a 65% market share, while Intel held 22%, and AMD was at 11%.

    An AMD representative commented to Reuters, "In order to align our resources with our biggest growth opportunities, we are implementing several targeted measures." They also mentioned that the company is "dedicated to treating affected employees with dignity and assisting them during this transition."

    Conclusion

    In summary, while AMD is showing signs of growth in certain areas, it faces significant challenges in the competitive landscape of AI chips and has had to make tough decisions regarding its workforce. The company’s commitment to supporting its employees through these changes is notable.

    Source: Link,Link

    Tags: , ,
  • India’s First AI Chips Launched by a Ride-Hailing Company

    India’s First AI Chips Launched by a Ride-Hailing Company

    Ola Electric, a company renowned in India for its electric scooters and ride-hailing services, has taken the industry by surprise with its announcement to venture into AI hardware. The company revealed its strategy to create a series of AI chips, marking a significant step towards India’s goals in the expanding field of artificial intelligence.

    AI Chip Models and Features

    The initial lineup of chips, set to be released in 2026, consists of three models: Bodhi 1, Ojas, and Sarv 1. Bodhi 1, recognized as the first AI chip to be designed and manufactured in India, focuses on AI inferencing. It is poised to be a formidable option for large language models (LLMs) and applications that need high-performance vision processing. Ola asserts that Bodhi 1 stands out in power efficiency, an essential attribute as AI systems increasingly demand more power.

    Ojas, the second chip, is tailored to meet the rising demand for edge AI solutions. It is versatile enough for a range of applications in the automotive, mobile, and IoT sectors. Ola intends to incorporate Ojas into its future electric vehicles, potentially enhancing capabilities like charging optimization and advanced driver assistance systems (ADAS).

    Sarv 1 and Market Challenges

    The third chip, Sarv 1, is a general-purpose server CPU utilizing the Arm instruction set, crafted to address the growing AI computational requirements of the data center industry.

    During Ola’s presentation, the company showcased impressive performance and power efficiency metrics for their prototype chips, making comparisons with Nvidia GPUs. However, some important details were omitted, such as the specific Nvidia GPU model used as a benchmark and the location where these chips will be manufactured.

    India’s Position in the Global AI Market

    This development highlights India’s aspiration to join the global AI competition, which is currently dominated by the US and China. With a large pool of tech talent, India is well-positioned to advance in AI. Furthermore, the ongoing restrictions on the sale of advanced technology to China by companies like Nvidia and ASML may make India an appealing alternative market for these firms.

    Nonetheless, Ola is confronted with substantial obstacles to make its bold AI chip initiative a reality. The AI hardware sector is currently ruled by well-established companies, and Ola must prove that its chips can hold their own in terms of performance, power efficiency, and cost. Additionally, the company needs to build a robust manufacturing infrastructure to produce these sophisticated silicon components.

    Tags: , ,
  • TSMC’s Advanced Packaging Capacity Fully Booked by Nvidia and AMD

    TSMC’s Advanced Packaging Capacity Fully Booked by Nvidia and AMD

    TSMC, the prominent semiconductor manufacturer globally, has declared that its advanced packaging capacity has been completely reserved for the next two years. This announcement coincides with Nvidia, AMD, and Guanghuida securing TSMC's cutting-edge packaging technologies for their high-performance computing (HPC) endeavors.

    Growing Demand for AI Processors

    The emphasis on high-performance computing is driven by its crucial role in supporting artificial intelligence (AI) tasks. TSMC foresees a substantial revenue increase from AI processors, with estimates suggesting a doubling of revenue just this year. Projections indicate that over the next five years, the compound annual growth rate for AI chips will reach 50%, with AI processors anticipated to contribute more than 20% of TSMC's revenue by 2028.

    Key Players Embrace TSMC's Technologies

    Nvidia and AMD have both secured TSMC's Chip-on-Wafer-on-Substrate (CoWoS) and System-on-Integrated-Chip (SoIC) advanced packaging capacities for their respective products. Nvidia's flagship H100 chip, produced on TSMC's 4nm process, utilizes CoWoS packaging. In contrast, AMD's MI300 series, fabricated using TSMC's 5nm and 6nm processes, employs SoIC for CPU and GPU integration before incorporating CoWoS with High Bandwidth Memory (HBM).

    Guanghuida, an emerging player in the AI chip market, has also reserved TSMC's packaging capacity. Their H100 chips, powered by TSMC's 4nm process and CoWoS packaging, feature SK Hynix's HBM for improved performance. Furthermore, Guanghuida's latest Blackwell architecture AI chip, based on TSMC's advanced 4nm process, showcases upgraded HBM3e memory, doubling the computing power compared to earlier versions.

    Meeting the Escalating Demand

    The increasing demand for AI chips is being fueled by major global cloud service providers like Amazon AWS, Microsoft, Google, and Meta, all striving for dominance in the AI server sector. To tackle shortages from leading manufacturers such as Nvidia, AMD, and Guanghuida, these cloud giants are turning to TSMC to fulfill their orders, contributing to the chipmaker's positive revenue forecasts.

    To address this rising demand, TSMC is enhancing its production capacity for advanced packaging. By the year's end, CoWoS monthly production is expected to triple, reaching 45,000 to 50,000 wafers, while SoIC capacity is set to double, hitting 5,000 to 6,000 wafers. By 2025, SoIC monthly production is projected to double once more, reaching 10,000 wafers.

    The full booking of TSMC's advanced packaging capacity signifies the rapid pace of innovation in AI-driven computing, with key industry players strategically positioning themselves to leverage this burgeoning market.

    Tags: , ,
  • Meta Reveals Next-Gen AI Chip “Artemis”

    Meta Reveals Next-Gen AI Chip “Artemis”

    Meta Platforms recently unveiled details about its forthcoming artificial intelligence chip dubbed Artemis, designed to meet growing processing power demands for running AI features across their social media platforms such as Facebook, Instagram and WhatsApp.

    Addressing Dependency On External Suppliers

    Artemis represents Meta’s move away from external suppliers like Nvidia for AI chip production and toward in-house chip development, in an effort to take back control over their hardware infrastructure while cutting energy consumption costs.

    Optimize Architecture For Improved Functionality

    Meta has designed Artemis with its dual objectives in mind. First, Artemis’ architecture was carefully engineered to balance computing power, memory bandwidth and capacity efficiently – thus making Artemis ideal for improving ranking and recommendation systems that form integral parts of social media platforms’ operations.

    Future Growth And Collaboration Strategies

    Meta has not shied away from its partnership with Nvidia since Artemis debuted, with CEO Mark Zuckerberg affirming their intentions of purchasing significant quantities of Nvidia H100 flagship chips throughout 2019. Meta plans on amassing over 600k AI chips by 2024 from various providers other than just Nvidia alone.

    Artemis was developed using Taiwan Semiconductor Manufacturing Co’s (TSMC) cutting-edge 5nm process and boasts an incredible threefold increase in performance when compared with Meta’s initial AI processor. Artemis is now deployed within Meta data centers worldwide supporting an array of AI applications while Meta plans on expanding Artemis to accommodate generative AI workloads in near future.

    Tags: , ,
  • Microsoft & OpenAI plan $100B supercomputer Stargate AI

    Microsoft & OpenAI plan $100B supercomputer Stargate AI

    According to a recent report from The Information, Microsoft and OpenAI are in discussions for a collaborative data center project that could amount to an estimated $100 billion. This initiative aims to introduce a groundbreaking artificial intelligence supercomputer known as "Stargate" by the year 2028.

    Microsoft and OpenAI's Joint Data Center Project

    This ambitious undertaking reflects the escalating demand for AI data centers capable of managing increasingly intricate tasks, largely driven by the emergence of generative AI technology.

    Microsoft is anticipated to take on the primary financial responsibility for this venture, with projections suggesting that the project's budget will be a hundredfold greater than current data center operations.

    Key Phases of the Stargate Supercomputer Initiative

    The envisioned supercomputer, situated in the United States, will serve as the focal point of a comprehensive six-year initiative, segmented into five distinct phases. The deployment of Stargate is envisioned as the final phase, following Microsoft's ongoing development of a smaller supercomputer for OpenAI, set for unveiling in 2026.

    The acquisition of specialized AI chips is slated to be a substantial expense during the latter stages of the project, with estimates indicating a price range of $30,000 to $40,000 per chip, as highlighted by Nvidia CEO Jensen Huang.

    Financial Projections and Technological Advancements

    Microsoft's commitment to enhancing AI capabilities is exemplified through its creation of bespoke computing chips. The new data center project is designed to integrate chips from diverse suppliers, as detailed in the report.

    The estimated expenditures for the project could potentially exceed $115 billion, surpassing Microsoft's previous year's capital investment in infrastructure by a significant margin.

    Recently, OpenAI CEO Sam Altman unveiled details about the forthcoming GPT-5 model during the World Government Summit in Dubai, revealing its anticipated advancements surpassing prior iterations. GPT-5 is poised to represent a substantial leap in AI capabilities, with early demonstrations showcasing its proficiency in deciphering ancient languages. It is expected to feature functionalities such as image generation through DALL-E and video generation via Sora, with a tentative rollout projected for 2024.

    Tags: , ,