Tags: semiconductor

Belgian research institute imec has demonstrated a prototype 3D charge‑coupled device (CCD) memory that merges the speed of DRAM with the density of NAND. By stacking cells vertically and using indium gallium zinc oxide (IGZO) instead of silicon, the hybrid architecture promises lower leakage, higher endurance and faster data transfer—key advantages for AI training and inference workloads. The first chips already show charge‑transfer speeds above 4 MHz, and imec plans to integrate the design into a CXL‑compatible device that could reshape the economics of AI hardware. Weiterlesen

The White House Office of Science and Technology Policy released a memo on Wednesday alleging that entities in China are running industrial‑scale campaigns to distill U.S. artificial‑intelligence models. The memorandum pledges to share threat intelligence with American AI firms and to explore sanctions against the perpetrators. The claim builds on accusations from OpenAI and Anthropic that Chinese labs have used millions of queries to replicate frontier models. Lawmakers responded with the Deterring American AI Model Theft Act, while the memo arrives weeks before a planned Trump‑Xi summit in Beijing. Weiterlesen

Amazon announced a fresh commitment of $5 billion to AI start‑up Anthropic, with the option to add another $20 billion if the firm hits predefined milestones. The deal builds on two prior $4 billion rounds and ties Anthropic to Amazon’s custom Trainium chips and a decade‑long $100 billion spend on AWS services. In return, Anthropic will make its Claude platform available directly through the AWS console, eliminating separate login steps for customers. Weiterlesen

Cerebras Systems, the San Diego‑based AI chip maker, filed to go public this week, targeting a mid‑May offering. The filing follows two massive funding rounds that lifted the company’s valuation to $23 billion and brought total capital raised to $2.1 billion. Recent contracts with Amazon Web Services and a reported $10 billion deal with OpenAI underscore the firm’s rapid ascent. Revenues hit $510 million in 2025, while net income reached $237.8 million, according to the prospectus. The IPO marks a new chapter for a startup that once withdrew a 2024 filing amid a federal review. Weiterlesen

Anthropic announced a new agreement with Google Cloud and Broadcom that will add roughly 3.5 gigawatts of compute capacity, primarily in the United States, to power its Claude models. The expansion builds on a 2025 deal and is slated to be operational by 2027, reflecting surging demand from enterprise customers despite recent U.S. Defense Department concerns. CFO Krishna Rao called the move the company’s “most significant compute commitment to date.” Weiterlesen

Cognichip, a startup that builds a deep‑learning model to assist engineers in designing semiconductor chips, announced a $60 million financing round led by Seligman Ventures, with Intel CEO Lip‑Bu Tan joining its board. The company claims its AI system can cut chip‑development costs by more than 75 percent and halve the design timeline. Cognichip trains its own model on proprietary and synthetic chip‑design data, allowing secure collaboration with manufacturers. The firm, founded in 2024, has raised a total of $93 million but has not yet disclosed a chip designed with its technology. Weiterlesen

The cost of AI infrastructure is increasingly driven by memory expenses, with DRAM prices jumping roughly 7x in the past year. As hyperscalers expand data centers, managing prompt caching and memory orchestration is emerging as a key competitive factor. Companies that master cache optimization can reduce token usage and inference costs, opening new avenues for profitability in AI applications. Weiterlesen

India's finance minister unveiled a budget proposal that grants foreign cloud providers a tax exemption on revenues from AI workloads run in Indian data centers and sold abroad through 2047. The plan also includes a cost‑plus safe harbour for Indian data‑center operators, expanded incentives for electronics and semiconductor manufacturing, and support for rare‑earth mineral development. Major global tech firms have already pledged billions to build AI‑focused data‑center campuses in the country, while domestic projects are also scaling up. The initiative aims to position India as a long‑term hub for AI infrastructure despite challenges such as power reliability and water scarcity. Weiterlesen

ASML, the Dutch photolithography specialist, posted record new orders worth 13 billion euros, more than double the previous quarter. The surge reflects strong demand from artificial‑intelligence data‑center builders, underscoring the continued momentum of the AI infrastructure boom. CEO Christophe Fouquet highlighted that customers are increasingly confident about medium‑term AI‑related demand, suggesting that the industry expects sustained growth in chip production for AI workloads. Weiterlesen

OpenAI has signed significant agreements with AMD and Broadcom to supply AI chips and accelerators for its data‑center deployments. The AMD deal focuses on large‑scale GPU production, while Broadcom will provide custom AI accelerators and Ethernet systems. Both partnerships underline OpenAI's strategy to diversify its silicon supply and reduce reliance on a single vendor. However, the majority of the chips are still expected to be fabricated by Taiwan Semiconductor Manufacturing Company (TSMC), whose capacity constraints and geographic concentration remain a critical bottleneck for the broader AI and semiconductor industries. Weiterlesen

OpenAI announced a partnership with Broadcom to design and deploy custom AI accelerators. The collaboration will see OpenAI creating specialized hardware while Broadcom handles manufacturing and integration into server racks that include its Ethernet, PCIe and optical connectivity products. Both companies say the effort will improve performance and efficiency for large‑scale AI models such as ChatGPT and the new Sora 2 video generator. Executives highlighted the partnership as a step toward broader AI infrastructure that benefits businesses and consumers, reinforcing a trend of major tech firms joining forces on AI hardware. Weiterlesen

AMD and OpenAI have entered a multi‑year agreement that lets OpenAI buy up to 160 million AMD shares at a nominal price while securing a substantial supply of AMD AI chips. The deal adds roughly 6 gigawatts of AMD‑based compute capacity to OpenAI’s growing AI infrastructure, complementing its existing relationships with Nvidia and its own silicon efforts. Industry sources say the partnership does not alter OpenAI’s broader compute strategy, including its work with Microsoft and Broadcom. Weiterlesen

OpenAI announced a strategic partnership with AMD to procure AMD's Instinct GPUs for large‑scale AI infrastructure. The agreement envisions deploying up to six gigawatts of compute power, with the first gigawatt slated for the second half of 2026. In addition to the hardware purchase, OpenAI will acquire 160 million AMD shares at a nominal price, potentially giving it a ten‑percent stake in the chipmaker. The deal positions AMD as a core compute partner and underscores OpenAI’s push to diversify its hardware sources beyond its existing relationships. Weiterlesen

Microsoft announced progress on a microfluidic cooling technique that channels liquid coolant directly through tiny channels etched onto silicon chips. Lab tests show the method can remove heat up to three times more effectively than traditional cold‑plate systems and cut peak chip temperatures by roughly 65 percent. By delivering cooling straight to the silicon, the approach could enable more powerful processors, tighter server packing, and lower energy use in data centers. Microsoft says the technology may also support future 3‑D chip architectures, though broader deployment will require manufacturing changes and further testing. Weiterlesen

Morningstar Equity Research forecasts Nvidia's AI‑related sales could approach $400 billion by 2028, driven largely by its accelerator products. The analysis highlights that hyperscaler cloud providers are expected to push annual capital expenditures beyond $450 billion by 2027, fueling demand for AI hardware. While growth remains robust, the pace is projected to decelerate after 2024, raising questions about sustainability. Competitive pressures from Broadcom, AMD and other semiconductor firms, together with rising energy needs, regulatory scrutiny and geopolitical factors, could temper Nvidia's dominance in the long run. Weiterlesen

Huawei announced a new AI infrastructure called SuperPoD Interconnect at its Huawei Connect conference, capable of linking up to 15,000 graphics cards, including its Ascend AI chips. The technology is positioned as a competitor to Nvidia's NVLink and aims to boost compute power for AI workloads. The launch follows a Chinese ban that prevents domestic tech firms from purchasing Nvidia hardware such as RTX Pro 600D servers. Weiterlesen

China’s Cyberspace Administration has prohibited domestic technology companies, including ByteDance and Alibaba, from buying Nvidia AI chips such as the RTX Pro 6000D server. The move follows earlier encouragement for firms to use locally‑made alternatives. Nvidia’s CEO Jensen Huang expressed disappointment but affirmed the company will respect the market’s wishes. The ban adds to a broader backdrop of U.S. licensing requirements that have limited Nvidia’s sales in China, with the firm forecasting significant revenue loss and noting that recent U.S. license approvals have yet to translate into actual shipments. Weiterlesen

China has halted the sale of Nvidia's AI chips after regulators summoned major domestic firms, including Huawei, Cambricon, Alibaba and Baidu, to compare their own processors with Nvidia's limited‑supply products. Officials concluded that Chinese AI chips now match or exceed the performance of the Nvidia chips allowed under U.S. export controls, prompting a consensus that domestic supply can meet demand without further Nvidia imports. Nvidia recently introduced the RTX Pro 6000D in Beijing, while regulators warned companies against purchasing Nvidia’s H20 chip, signalling a tightening of restrictions on foreign AI hardware. Weiterlesen