News Heat Trend
Project Overview
Node AI enabling users to participate, contribute, and benefit from the evolving landscape of artificial intelligence. By leveraging blockchain technology, Node AI ensures a transparent and secure ecosystem where participants can engage in various capacities. Such as Revenue Sharing, AI Nodes Rental and GPU Power Lending and so on.
HIVE Digital Signs $350M Five-Year AI Cloud Services Contract, Deploying 2,016 NVIDIA Blackwell Ultra GPUs
Odaily News - Digital infrastructure company HIVE Digital Technologies' high-performance computing division, BUZZ High Performance Computing, has signed a five-year AI cloud services contract worth approximately $350 million with an undisclosed investment-grade enterprise client. The contract is expected to generate approximately $70 million in additional annual revenue, bringing BUZZ HPC's annualized revenue to approximately $180 million. BUZZ HPC will deploy 2,016 NVIDIA Blackwell Ultra GPUs, utilizing the GB300 NVL72 system, NVIDIA Quantum-X800 InfiniBand networking, and VAST Data storage. The cluster is expected to become operational later this year at Bell's AI Fabric facility in Merritt, British Columbia, Canada, which runs on renewable hydroelectric power and closed-loop liquid cooling technology. HIVE estimates the project's capital expenditure at approximately $185 million, which will be funded through previously announced financing and new equipment debt. The company expects daily revenue of approximately $500,000 from its HPC and AI business once the cluster is fully operational, and plans to achieve $200 million in annualized GPU cloud services revenue by year-end. The company holds approximately 400 megawatts of capacity in Canada, which can support over 120,000 GPUs over the next two years. (Bitcoin.com News)
Serenity: The AI Supply Chain Boom Is Far From Over, With Storage, Packaging, and Power Sectors Poised for Long-Term Demand Surge
Odaily News "White-Haired Stock God" Serenity shared insights on the AI industry chain on the X platform, noting that AI infrastructure demand is driving multiple sectors—including storage, advanced packaging, computing power financing, optical communications, power supply, and electronic components—into a long-term expansion cycle. The AI supply chain remains in a phase of rapid growth.In the storage sector, Serenity cited UBS forecasts indicating that traditional DRAM manufacturers (such as Micron) could see gross margins reach an unprecedented 95% by 2027, potentially even surpassing the gross margin levels of HBM products. Additionally, SanDisk's long-term agreements already cover approximately two-thirds of its 2028 production capacity, with minimum contracted revenue reaching $93 billion. Given its current market cap of around $239 billion, this suggests its future revenue targets could persist for years, making it difficult to simply classify the company as a traditional cyclical stock.On the cloud computing infrastructure front, CoreWeave has signed agreements to use Nvidia A100 GPUs through 2029. This is a positive development for emerging cloud computing companies such as Nebius and Iren, and it also weakens some investors' bearish thesis centered on the rapid depreciation of older GPUs.AI model companies are also continuing to grow at a pace that exceeds expectations. Frontier AI labs are still maintaining extremely rapid growth rates, and a slowdown in growth would actually be a cause for concern. The market projects that Anthropic's 2028 revenue could reach $190 billion to $200 billion.However, advanced packaging and semiconductor infrastructure remain core bottlenecks. The head of advanced packaging at TSMC has stated that in the coming years, the industry may face not only memory shortages but also tight supply of ABF substrates.Serenity concluded that the AI infrastructure supply chain is continuously expanding. From GPUs, storage, and advanced packaging to power, optical communications, and electronic components, every segment is showing a long-term demand growth trend. The AI supply chain is still in a high-speed development stage.
CoreWeave and Nebius Earnings Reveal AI Cloud Computing Trends: Supply Shortage Persists, CSPs Move Toward "AI Infrastructure Operating Systems"
Odaily News - Analyst qinbafrank posted on X platform, stating that the latest earnings reports from CoreWeave (CRWV) and Nebius show the AI cloud computing (CSP) industry is entering a phase of rapid expansion. The competitive focus is shifting from simply providing GPU leasing to building AI infrastructure platforms that encompass computing power, software, data, and operational capabilities.Currently, AI computing demand still significantly exceeds short-term deliverable supply. Meanwhile, pricing power for AI computing is strengthening, but price increases are mainly concentrated on high-value resources. CoreWeave stated that prices for various GPU computing SKUs rose by approximately 25% on average in July; Nebius disclosed that prices for previous-generation GPUs increased by over 30% compared to Q1, with new contracts signed in Q2 averaging over $20 million in annualized revenue per MW, some projects reaching $20 million to $25 million, and short-term emergency capacity prices even reaching $40 million to $50 million per MW.However, price increases are mainly occurring in short-term capacity, next-generation GPUs, large-scale clusters, and production-grade AI inference scenarios. Traditional low-priority, long-term locked-in bare computing power has not seen concurrent price increases. From a profitability model perspective, project-level returns on AI computing are becoming clearer, but overall corporate return on invested capital (ROIC) still needs time to be validated. Nebius has for the first time disclosed relatively clear project payback periods, while CoreWeave is reducing GPU investment pressure through long-term contracts and asset-level financing. However, both companies remain in a high-capital-expenditure phase, with depreciation and financing costs continuing to compress profit margins.Nevertheless, an increasing number of individual projects are achieving closed-loop economic models, indicating that the AI infrastructure business model is gradually maturing. Additionally, both CoreWeave and Nebius are upgrading toward becoming "AI infrastructure operating systems." Future CSP competition will no longer be just about renting out GPU hours but will cover complete service systems including AI training, inference, storage, networking, model deployment, monitoring, security governance, and Agent runtime environments.In terms of capital models, the two companies are also taking different paths: Nebius leans more toward an asset-light model, building AI data centers through capital partners while providing AI infrastructure operations and software capabilities itself; CoreWeave, on the other hand, is promoting a hybrid cloud model through its Omni strategy, deploying complete AI cloud platforms to customers' own data centers and GPU resources, placing greater emphasis on enterprise-level and sovereign AI delivery.Overall, the AI cloud computing industry is evolving from "GPU rental providers" to "AI infrastructure platforms." Short-
AI architecture company Pathway completes $30 million seed funding round, with participation from Id4 Ventures and others
Odaily News: AI architecture development company Pathway has announced the completion of a $30 million seed funding round, with participation from Id4 Ventures, TQ Ventures, Red Bridge Ventures, Kadmos Capital, and WS Investment, the investment arm of Wilson Sonsini, among others. Databricks' Chief AI Scientist Jonathan Frankle has joined as an angel investor.Pathway is developing a Bio-inspired Dynamic Hierarchical architecture (BDH), a "post-Transformer" architecture designed to overcome the limitations of current Transformer models, which continue to rely on ever-increasing data, GPUs, energy, and capital expansion. Unlike traditional large models that require periodic retraining, BDH enables continuous learning and ongoing adaptation with significantly less data.The company also announced the appointment of Adam Kurzrok, former product lead for Google DeepMind's Gemini, as Chief Product Officer. He will oversee the product direction of BDH models, covering areas such as model packaging, evaluation systems, and commercial deployment. (Finsmes)
GPU data company Silicon Data completes $30.5 million Series A financing, led by Valor Atreides AI Fund
GPU market data infrastructure company Silicon Data announced the completion of the first closing of its $30.5 million Series A financing round, led by Valor Atreides AI Fund, with participation from CME Ventures, DRW, Samsung Next, VanEck, Jump Trading, Wintermute, and others. The funds will be used for GPU benchmark pricing, performance measurement, institutional and alternative data, as well as derivatives, insurance, and credit market risk infrastructure.
Bank of America: NVIDIA $500 Billion Third-Party Financing Diversifies Risk, No Need to Tap Balance Sheet
According to TechFlow Research, a Bank of America research report on August 10 pointed out that NVIDIA signed a memorandum of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize over $500 billion in third-party capital through an independent platform. Previously, NVIDIA invested approximately $70 billion in equity in ecosystem partners such as OpenAI and Anthropic, accounting for only 15% of the expected free cash flow of $470 billion from 2026 to 2027, without affecting the commitment to return 50% of free cash flow to shareholders. Bank of America believes the financing structure shifts the capital burden from NVIDIA to the consortium. GPU computing power can be transferred across operators, and CUDA extends the service life, with the asset quality itself resisting depreciation. The $500 billion fund pool allows non-investment grade buyers to acquire GPUs at preferential rates, transforming AI computing power acquisition from capital-intensive purchases to financial leasing, supporting the $1.7 trillion AI system TAM by 2030. Bank of America maintains a Buy rating with a target price of $350, corresponding to 26 times the expected earnings per share in 2027. The upcoming earnings conference call is the next important catalyst.
Silicon Data Secures $30.5 Million Series A Funding, Led by Valor Atreides AI Fund
Odaily News: GPU market data infrastructure company Silicon Data has announced the first closing of its $30.5 million Series A funding round, led by Valor Atreides AI Fund, with participation from CME Ventures, DRW, Samsung Next, VanEck, Jump Trading, Wintermute, and others. The funds will be used for GPU benchmark pricing, performance measurement, institutional and alternative data, as well as risk infrastructure for derivatives, insurance, and credit markets.Silicon Data currently collects data from approximately 100 GPU rental platforms across more than 40 countries worldwide, processing over 150,000 verified price records daily. CME Group plans to adopt Silicon Data's benchmarks as the reference price for its proposed cash-settled GPU futures contracts, pending regulatory approval.
South Korean Public Institution Employee Fired for Unauthorized Mining, Court Dismisses Wrongful Dismissal Lawsuit
According to Yonhap News Agency, a former department head surnamed A at the Korea Food Research Institute filed an administrative lawsuit after being fired for privately mining cryptocurrency in the institution's warehouse, but was ruled to have lost the case by the Seoul Administrative Court on the 3rd of last month. According to the investigation, between February and September 2023, A unauthorizedly installed 2 GPU servers in the idle Promotion Hall warehouse of the institute and used institutional budget to complete air conditioning, network, and electrical renovation projects, cumulatively illegally mining approximately 71 million altcoins. After the incident was discovered, A also forged approval documents attempting to retrieve the GPU servers to destroy evidence. In addition, between August 2023 and May 2024, A used an unauthorized VPN to clock in in violation of regulations a total of 117 times, indirectly causing important scientific research data of the institute to be illegally leaked. The Audit Committee of the National Research Council for Science & Technology launched a special audit on A in 2024, subsequently reported to the police, and requested the institute to impose a dismissal penalty. After A's internal appeals and relief applications to the Local Labor Relations Commission and the Central Labor Relations Commission were all rejected, A resorted to the administrative court, but still ended in defeat. The court ruled that the dismissal "did not involve any circumstances clearly violating social common sense or abuse of discretion." In the criminal case, A was charged with crimes such as theft, violation of the Information and Communications Network Act, and forgery of private documents, sentenced to 1 year in prison in the first instance, and the second instance upheld the original verdict in April this year, formally finalizing the sentence.
"Unease and anticipation are just two sides of the same coin": SK Hynix stock pulls back but remains favored by institutions, market focuses on HBM4 technology moat
Odaily News: SK Hynix's stock price has recently seen a pullback. On August 3, the stock fell 8.79%. Although it rebounded slightly by 0.64% and 5.77% on August 4 and 5 respectively, it plunged another 10.37% on August 6, closing at 1.495 million KRW; on August 7, it dropped a further 4.88%, closing at 1.422 million KRW.Against the backdrop of heightened market volatility, SK Group released an advertisement quoting founder Choi Jong-gun’s famous saying: "Despair and hope are two sides of the same coin; you can turn despair into hope as easily as flipping your palm," and adapted it to: "Unease and anticipation in the AI era are also two sides of the same coin; you can turn unease into anticipation," thereby conveying confidence in the long-term development of the AI industry.Securities institutions believe that short-term stock price fluctuations have not changed SK Hynix's fundamentals, and the market should focus on its HBM4 technology leadership and the earnings stability brought by long-term supply agreements (LTAs). Specifically:1. Hyundai Motor Securities expects SK Hynix's DRAM and NAND bit growth to reach 9.7% and 1.5% respectively in the third quarter. As HBM4 sales contributions expand, even with a higher proportion of LTAs, DRAM average selling prices (ASP) are still expected to rise 19.9% quarter-over-quarter. Companies such as OpenAI and Anthropic are advancing plans to build their own hyperscale AI data centers and intend to raise funds for related construction through IPOs. Even if some large tech companies adjust capital expenditures (Capex) in the future, this could be offset by demand from other AI infrastructure. Additionally, regarding competitive concerns over China's CXMT, given the U.S. continued tightening of semiconductor equipment export restrictions, as well as Micron's expansion of domestic U.S. investment, the likelihood of major companies like Apple adopting Chinese memory chips is relatively low.2. SK Securities is also bullish on SK Hynix's competitive advantages, believing that with its leading position in HBM, partnerships with major North American GPU companies, and AI-driven LTA demand, SK Hynix's market position remains solid. Currently, the HBM supply-demand fulfillment rate is below 70%, and the core value of LTAs lies in ensuring profit sustainability and earnings stability through a "mutual binding structure" between customers and suppliers. With value-reassessment initiatives such as an ADR listing progressing, along with dividend income from the sale of SPC assets related to Kioxia, the company's goal of achieving net cash of 100 trillion KRW may be reached earlier than expected. As shareholder return policies gradually become clearer, this will help the market re-evaluate the value of the LTA model and drive a further re-rating of SK Hynix. (Daum)
Western Digital: AI Storage Competition Is Not "Flash vs. HDD", Economic Scalability Is the Key
Western Digital Chief Product Officer Ahmed Shihab published a long article pointing out that as AI infrastructure expands rapidly, the core competition in the storage field should not be simply reduced to a contest between Flash and Hard Disk Drives (HDD), but lies in whether an AI storage architecture with long-term economic scalability can be built. The AI industry is currently facing a key question: whether the storage architecture chosen this year can support future data scale growth to the PB level or even the EB level. Many AI infrastructure designs do not fail due to insufficient performance, but fall into cost dilemmas after data scale expands. Ahmed Shihab added that Flash and HDD are not in a competitive relationship, but are complementary technologies for different workloads. High-performance scenarios, such as model weights, GPU spillover, KV cache, etc., require low-latency Flash support; while long-term storage needs such as training datasets, logs, checkpoints, compliance records, and large-scale historical data are more suitable for adopting HDDs with cost advantages. Storage architecture in the AI era will be more layered, rather than relying on a single storage medium. "Flash handles performance at critical moments, HDD handles data lifecycle. The direction of future AI storage development is not 'Flash replacing HDD', but precise layering based on different data lifecycles and business requirements." "True infrastructure is not about pursuing dazzling performance, but a reliable foundation capable of supporting long-term AI growth." US stock market trends show, Western Digital
AMD launches AI programming platform Instinct Coder, which can reduce enterprise AI coding costs by 70%
Odaily News AMD, the semiconductor giant, announced the launch of its enterprise-grade AI programming platform, AMD Instinct Coder. The platform combines AMD chips, Supermicro servers, and Spectro Cloud software, aiming to help enterprises deploy AI coding assistants locally, reduce the cost of cloud-based AI models, and protect code and data security.AMD stated that Instinct Coder is an "out-of-the-box" end-to-end AI development platform, integrating AMD EPYC processors, AMD Instinct GPUs, Supermicro AI servers, Spectro Cloud PaletteAI Inference Launchpad software, and the AMD-optimized GLM-5.2 model. It can be used for software development scenarios such as code generation, application modernization, automated testing, and code review.AMD said that compared to relying on cutting-edge cloud-based AI models, Instinct Coder can help enterprises reduce total cost of ownership (TCO) by up to 70%, with the fastest payback period shortened to 6 months.AMD noted that more and more enterprises are looking to leverage AI to improve development efficiency, but face two major challenges: on one hand, the cost of invoking top-tier cloud models continues to rise; on the other hand, entrusting enterprise source code, intellectual property, and sensitive data to third-party services poses security and compliance risks.Through a local deployment model, Instinct Coder allows enterprises to maintain control over their data and code while providing more predictable infrastructure costs. The platform supports development tools such as Claude Code, OpenAI Codex, Visual Studio Code, and Cursor, with each node supporting up to 50 users (30 concurrent users).Additionally, the PaletteAI Inference Launchpad provided by Spectro Cloud enables AI workload management, model routing, request auditing, and cost monitoring, and supports invoking external models such as Anthropic, OpenAI, Google, or xAI when needed.AMD stated that Instinct Coder aims to help enterprises break free from the high costs of cloud-based AI services, accelerate AI-driven software development processes while ensuring data security and autonomous control.
DRAM supply shortage expected to persist through 2027; NVIDIA evaluates reducing HBM configuration on Rubin Ultra
Odaily News, Citrini analyst jukan posted on X platform, according to TrendForce's latest memory industry research, the DRAM supply shortage is expected to persist through 2027, and the HBM4e certification timeline remains uncertain. NVIDIA has been reassessing the HBM configuration for Rubin Ultra since the third quarter of 2026.The original plan called for a 12-layer stacked HBM4e configuration, but NVIDIA is currently evaluating multiple designs in parallel, including HBM4e 8-layer stacking, HBM4 12-layer stacking, and HBM4 8-layer stacking, with the final specifications yet to be confirmed. In addition to NVIDIA, some CSPs are reportedly also considering reducing HBM capacity for their next-generation custom ASICs.From 2025 to the first half of 2026, NVIDIA used 12-layer stacked HBM4e as the baseline design for Rubin Ultra. However, since the beginning of the third quarter of 2026, NVIDIA has begun reviewing lower-spec alternatives. TrendForce attributes this change to two major supply-side constraints: first, the overall DRAM shortage expected in 2027 will limit the wafer capacity that memory manufacturers can allocate to HBM production; second, uncertainty remains regarding the certification timeline for 12-layer stacked HBM4e and the pace of yield improvement in mass production.TrendForce stated that NVIDIA's primary focus for the Rubin Ultra generation is improving I/O speed, with expanding GPU shipments as a secondary priority. If NVIDIA ultimately decides to downgrade the HBM specifications, it is expected to do so by reducing the number of DRAM stacking layers. Whether HBM4e can complete certification and enter mass production as planned will determine whether Rubin Ultra's I/O speed can be improved from the previous generation Rubin's 8 to 11.7Gbps to 14 to 16Gbps, or be maintained at 11 to 12Gbps through an optimized HBM4 design. Within the same product generation, the number of DRAM stacking layers determines the trade-off between HBM capacity per GPU and the number of GPUs that can be shipped.TrendForce also believes that the final configuration will depend on wafer allocation decisions made by memory manufacturers. On the supply-demand front, HBM bit shipments in 2027 are expected to grow 50% to 60% year-over-year, but are still projected to fall short of demand growth. With supply constraints persisting, HBM suppliers are expected to maintain pricing power throughout 2027. The industry has broadly anticipated significant HBM price increases, and AI chip makers will face the dual pressure of limited HBM supply and rising procurement costs, further strengthening the incentive to adopt lower HBM capacity configurations.
SK Hynix's Optical Interconnect Memory Roadmap: Connecting Larger Shared Memory Pools via Localized HBM
Odaily News, Citrini analyst jukan stated on the X platform that at first glance, SK Hynix's CPO technology roadmap appears to have one axis pointing toward HBM and another toward optical communication. However, the underlying logic is that AI competition is shifting from individual chip performance to data transfer efficiency across the entire system.Over the past few years, HBM has addressed the problem of GPUs being unable to receive data fast enough. By vertically stacking multiple layers of DRAM and placing them next to the GPU, HBM delivers extremely high bandwidth. Yet, as more HBM stacks are placed around each GPU, packaging area, interposer edge space, power supply, and thermal dissipation capabilities are all approaching their limits. Meanwhile, AI clusters have expanded to thousands or even tens of thousands of GPUs. No matter how fast a single GPU computes, the entire system will still be constrained by the "bandwidth wall" if data cannot be efficiently transferred between GPUs and racks.SK Hynix's vision is to extend optical interconnect to memory. Low-latency, high-bandwidth local HBM will remain next to the GPU, while optical fibers connect it to a larger shared memory pool. This approach avoids limiting all memory capacity within a single GPU package and enables horizontal scaling at the rack level.From an investment perspective, this does not mean HBM will be replaced in the near term. What is more likely to emerge is a new memory hierarchy: frequently accessed data remains stored in local HBM, while larger-scale, less frequently accessed data is stored in optically interconnected memory pools, HBF, or SSDs.SK Hynix is repositioning its business, shifting from selling standardized memory chips to co-designing HBM, controllers, advanced packaging, and system-level memory architectures with customers. If this roadmap comes to fruition, SK Hynix could strengthen customer stickiness, increase product added value, and enhance its ability to secure long-term contracts. At the same time, the company may also more proactively address the potential impact of future memory disaggregation on traditional HBM business models.At the supply chain level, areas that may benefit in the long term include silicon photonics chips, optical engines, lasers, fiber coupling technology, and advanced 2.5D and 3D packaging.However, this concept is still in its very early stages and essentially remains just a roadmap. jukan noted that a person involved in TSMC's packaging business whom he interviewed today was completely unaware of this plan.
Citrini analyst questions SanDisk's HBF demonstration: Possible deliberate underestimation of HBM performance
Odaily News, Citrini analyst Zephyr posted on X platform questioning that SanDisk's HBF vs. HBM performance comparison at its Investor Day may contain parameter configuration bias, arguing that the company deliberately underestimated HBM performance in its demonstration.Zephyr pointed out that SanDisk set total bandwidth for both HBM and HBF at 12.8TB/s, while running the Qwen3-480B-A35B model with bfloat16. Current model inference more commonly uses FP4 or FP8, requiring approximately 240GB to 480GB of capacity.Zephyr also noted that SanDisk fixed HBM capacity at 192GB per GPU, but 16-layer HBM4E in an 8-stack configuration can reach 512GB capacity with approximately 32TB/s bandwidth—roughly three times the parameters used in SanDisk's demonstration. Zephyr believes these parameter choices result in the company's claim that "HBF requires only 1 to 4 GPUs, while HBM requires 8 GPUs" failing to adequately represent actual performance under higher-spec HBM configurations.
CoreWeave and Nebius Earnings Reveal AI Cloud Computing Trends: Supply Shortage Persists, CSPs Move Toward "AI Infrastructure Operating Systems"
Odaily News - Analyst qinbafrank posted on X platform, stating that the latest earnings reports from CoreWeave (CRWV) and Nebius show the AI cloud computing (CSP) industry is entering a phase of rapid expansion. The competitive focus is shifting from simply providing GPU leasing to building AI infrastructure platforms that encompass computing power, software, data, and operational capabilities.Currently, AI computing demand still significantly exceeds short-term deliverable supply. Meanwhile, pricing power for AI computing is strengthening, but price increases are mainly concentrated on high-value resources. CoreWeave stated that prices for various GPU computing SKUs rose by approximately 25% on average in July; Nebius disclosed that prices for previous-generation GPUs increased by over 30% compared to Q1, with new contracts signed in Q2 averaging over $20 million in annualized revenue per MW, some projects reaching $20 million to $25 million, and short-term emergency capacity prices even reaching $40 million to $50 million per MW.However, price increases are mainly occurring in short-term capacity, next-generation GPUs, large-scale clusters, and production-grade AI inference scenarios. Traditional low-priority, long-term locked-in bare computing power has not seen concurrent price increases. From a profitability model perspective, project-level returns on AI computing are becoming clearer, but overall corporate return on invested capital (ROIC) still needs time to be validated. Nebius has for the first time disclosed relatively clear project payback periods, while CoreWeave is reducing GPU investment pressure through long-term contracts and asset-level financing. However, both companies remain in a high-capital-expenditure phase, with depreciation and financing costs continuing to compress profit margins.Nevertheless, an increasing number of individual projects are achieving closed-loop economic models, indicating that the AI infrastructure business model is gradually maturing. Additionally, both CoreWeave and Nebius are upgrading toward becoming "AI infrastructure operating systems." Future CSP competition will no longer be just about renting out GPU hours but will cover complete service systems including AI training, inference, storage, networking, model deployment, monitoring, security governance, and Agent runtime environments.In terms of capital models, the two companies are also taking different paths: Nebius leans more toward an asset-light model, building AI data centers through capital partners while providing AI infrastructure operations and software capabilities itself; CoreWeave, on the other hand, is promoting a hybrid cloud model through its Omni strategy, deploying complete AI cloud platforms to customers' own data centers and GPU resources, placing greater emphasis on enterprise-level and sovereign AI delivery.Overall, the AI cloud computing industry is evolving from "GPU rental providers" to "AI infrastructure platforms." Short-
DRAM supply shortage expected to persist through 2027; NVIDIA evaluates reducing HBM configuration on Rubin Ultra
Odaily News, Citrini analyst jukan posted on X platform, according to TrendForce's latest memory industry research, the DRAM supply shortage is expected to persist through 2027, and the HBM4e certification timeline remains uncertain. NVIDIA has been reassessing the HBM configuration for Rubin Ultra since the third quarter of 2026.The original plan called for a 12-layer stacked HBM4e configuration, but NVIDIA is currently evaluating multiple designs in parallel, including HBM4e 8-layer stacking, HBM4 12-layer stacking, and HBM4 8-layer stacking, with the final specifications yet to be confirmed. In addition to NVIDIA, some CSPs are reportedly also considering reducing HBM capacity for their next-generation custom ASICs.From 2025 to the first half of 2026, NVIDIA used 12-layer stacked HBM4e as the baseline design for Rubin Ultra. However, since the beginning of the third quarter of 2026, NVIDIA has begun reviewing lower-spec alternatives. TrendForce attributes this change to two major supply-side constraints: first, the overall DRAM shortage expected in 2027 will limit the wafer capacity that memory manufacturers can allocate to HBM production; second, uncertainty remains regarding the certification timeline for 12-layer stacked HBM4e and the pace of yield improvement in mass production.TrendForce stated that NVIDIA's primary focus for the Rubin Ultra generation is improving I/O speed, with expanding GPU shipments as a secondary priority. If NVIDIA ultimately decides to downgrade the HBM specifications, it is expected to do so by reducing the number of DRAM stacking layers. Whether HBM4e can complete certification and enter mass production as planned will determine whether Rubin Ultra's I/O speed can be improved from the previous generation Rubin's 8 to 11.7Gbps to 14 to 16Gbps, or be maintained at 11 to 12Gbps through an optimized HBM4 design. Within the same product generation, the number of DRAM stacking layers determines the trade-off between HBM capacity per GPU and the number of GPUs that can be shipped.TrendForce also believes that the final configuration will depend on wafer allocation decisions made by memory manufacturers. On the supply-demand front, HBM bit shipments in 2027 are expected to grow 50% to 60% year-over-year, but are still projected to fall short of demand growth. With supply constraints persisting, HBM suppliers are expected to maintain pricing power throughout 2027. The industry has broadly anticipated significant HBM price increases, and AI chip makers will face the dual pressure of limited HBM supply and rising procurement costs, further strengthening the incentive to adopt lower HBM capacity configurations.
Analyst: Kimi K3 Ultra-large Model to Be Open-Sourced Soon, AI Infrastructure Platforms May Be the Biggest Beneficiaries
analyst KawzInvests stated that Moonshot AI's upcoming Kimi K3 could become a significant event in the open-source AI space, and the infrastructure demand behind it may drive growth for AI cloud service platforms. Kimi K3 has approximately 2.8 trillion parameters, making it one of the ultra-large-scale open-source models. According to Moonshot's official evaluation, the model's performance is only slightly behind frontier models like Claude Fable 5 and GPT 5.6 Sol, and the full model weights are expected to be released on July 27.KawzInvests pointed out that a model of this scale cannot run on an ordinary laptop or even a single server; users need a computing cluster composed of a large number of GPUs to complete model loading and inference. When top-tier open-source models are made available for free, the biggest beneficiaries might not be ordinary users, but rather platforms that offer model hosting and inference services. For example, $DOCN (DigitalOcean) already supports serverless inference services for models like Kimi K2.6. Developers do not need to deploy hardware; they can call the model via API and pay per Token. Additionally, the platform hosts over 70 models and covers GPU leasing, model fine-tuning, and AI Agent development tools.As more large-scale open-source models are released, developers' demand for low-barrier AI infrastructure will continue to increase. Model hosting, inference services, and GPU cloud platforms may become key beneficiaries in the open-source AI wave.
South Korean listed company K Wave Media liquidates 88 BTC, exits the Bitcoin treasury company ranks
According to BitcoinTreasuries data, South Korean listed company K Wave Media (KWM) has sold all of its remaining 88 BTC to repay $6 million in debt. Following the sale, the company's Bitcoin holdings have dropped to zero, exiting the ranks of Bitcoin treasury companies.K Wave Media announced last year that it had secured a $1 billion Bitcoin treasury financing capacity and planned to expand its Bitcoin holdings to 10,000 BTC as soon as possible. However, in May this year, the company redirected up to $485 million of its remaining financing capacity from the Bitcoin treasury strategy to AI infrastructure construction, including data centers, GPU computing power, and related acquisitions.
Different COLDCARD Mk3 devices may generate the same mnemonic phrase, with approximately 4.5 million possible starting states searchable within 3 seconds
Odaily News: Bitcoin News posted on X platform that a new technical analysis released by @KLoaec shows that some vulnerable COLDCARD Mk3 wallets may be generated from only approximately 4.5 million random number generator starting states, which can be searched in about 3 seconds on a single RTX 4090 GPU. Even accounting for additional uncertainty in each wallet's generation method, an attacker could complete the search in about 50 minutes on a single high-end GPU. More critically, this vulnerability could cause different devices to generate identical mnemonic phrases. Assuming 30,000 Mk3 devices, the analysis estimates that approximately 120 pairs of devices could generate the same random number stream. This collision estimate is theoretical but indicates that duplicate mnemonic generation across different devices may be possible.
Aethir Prevents Cross-Chain Bridge Vulnerability Attack and Promises Compensation
Decentralized GPU cloud computing infrastructure platform Aethir confirmed that its Ethereum-related bridge contract was attacked. The team promptly disconnected the affected contract and, in collaboration with major exchanges, blacklisted the hacker’s wallet, limiting losses to under $90,000. Earlier, blockchain security firm PeckShield estimated losses at $400,000. The attacker exploited Aethir’s cross-chain smart contract, AethirOFTAdapter, to transfer stolen funds from BNB Chain to Tron. Aethir stated that its Ethereum mainnet ATH token supply remains unaffected. It plans to release a detailed compensation plan and incident analysis next week and will collaborate with exchanges including Binance, Upbit, and Bithumb to freeze funds. Web3 security platform ZeroShadow is assisting with the investigation. In 2025, Aethir achieved $127.8 million in revenue and deployed over 440,000 GPU containers globally.
Krackpot Launches Browser Game, Offering 6 BTC for Cracking Bitcoin Puzzle 71 with GPU
Bitcoin News posted on X platform, stating that Simon Males has launched Krackpot, a browser-based game that allows anyone to use GPU to attempt cracking the private key of Bitcoin Puzzle 71. The prize is 6 BTC. The probability of cracking depends on the number of GPUs: a single gaming GPU would take approximately 830,000 years, 1 million GPUs would take 1 year, and simultaneous participation from all Steam players would take 1 week. The game runs locally in the browser via WebGPU. If a user finds the private key, 6 BTC will be sent to their designated Bitcoin address, with the remainder going to the developer.
Bank of America: Anthropic leads frontier AI model rankings, token prices ease but GPU rents remain firm
According to TechFlow Research, the frontier AI data tracking report released by Bank of America Securities on August 17 shows that Anthropic leads comprehensively in three major AI benchmarks, with Claude Opus 5 ranking first in the Intelligence Index, Agent Index, and Coding Agent Index, GPT-5.6 Sol following closely behind, Meta MuseSpark 1.2 entering the top ten, and Google Gemini 3.6 Flash ranking outside the top ten. In terms of usage, DeepSeek leads with approximately 30% of the Vercel platform token share, Anthropic accounts for 25% and OpenAI accounts for 16%; but in terms of payment amount, Anthropic leads far ahead with 65%, while OpenAI accounts for only 11%. In terms of pricing, the AI Token Price Index decreased 9% month-over-month in August to $2.21, but still increased 87% year-over-year; GPU rental rates remain strong, with H100 increasing 33% year-over-year to $2.77/hour, DRAM increasing 483% year-over-year, and NAND increasing 432% year-over-year. The research report judges that AI infrastructure demand remains healthy, with open-source model usage growing but payment share still highly concentrated on top closed-source models. BofA believes that the Meta "Watermelon" and Google Gemini 4 releases, token pricing trends and GPU rental trends are
Former Bitcoin Miner IREN Completes First Milestone of $9.7 Billion Microsoft AI Contract
According to Cryptopolitan, former Bitcoin mining company IREN announced the completion of the first milestone of its $9.7 billion, five-year contract with Microsoft — the official delivery of the Horizon 1 facility located in Childress, Texas. The facility is 50 megawatts in scale, equipped with Nvidia GB300 systems and direct chip-level liquid cooling technology, and has received the Exemplar Cloud certification awarded by Nvidia. Under the contract terms, Microsoft will complete GPU acceptance verification within five days; upon approval, IREN can commence monthly billing, with expected annualized revenue of approximately $1.94 billion after all four Horizon facilities go online. IREN plans to complete the deployment of all four phases totaling 200 megawatts by 2026, expand AI cloud computing power to 1.2 gigawatts by 2027, simultaneously exit the Bitcoin mining business, and target a full-year AI cloud revenue run rate exceeding $4 billion.
HIVE Digital Signs $350M Five-Year AI Cloud Services Contract, Deploying 2,016 NVIDIA Blackwell Ultra GPUs
Odaily News - Digital infrastructure company HIVE Digital Technologies' high-performance computing division, BUZZ High Performance Computing, has signed a five-year AI cloud services contract worth approximately $350 million with an undisclosed investment-grade enterprise client. The contract is expected to generate approximately $70 million in additional annual revenue, bringing BUZZ HPC's annualized revenue to approximately $180 million. BUZZ HPC will deploy 2,016 NVIDIA Blackwell Ultra GPUs, utilizing the GB300 NVL72 system, NVIDIA Quantum-X800 InfiniBand networking, and VAST Data storage. The cluster is expected to become operational later this year at Bell's AI Fabric facility in Merritt, British Columbia, Canada, which runs on renewable hydroelectric power and closed-loop liquid cooling technology. HIVE estimates the project's capital expenditure at approximately $185 million, which will be funded through previously announced financing and new equipment debt. The company expects daily revenue of approximately $500,000 from its HPC and AI business once the cluster is fully operational, and plans to achieve $200 million in annualized GPU cloud services revenue by year-end. The company holds approximately 400 megawatts of capacity in Canada, which can support over 120,000 GPUs over the next two years. (Bitcoin.com News)
Serenity: The AI Supply Chain Boom Is Far From Over, With Storage, Packaging, and Power Sectors Poised for Long-Term Demand Surge
Odaily News "White-Haired Stock God" Serenity shared insights on the AI industry chain on the X platform, noting that AI infrastructure demand is driving multiple sectors—including storage, advanced packaging, computing power financing, optical communications, power supply, and electronic components—into a long-term expansion cycle. The AI supply chain remains in a phase of rapid growth.In the storage sector, Serenity cited UBS forecasts indicating that traditional DRAM manufacturers (such as Micron) could see gross margins reach an unprecedented 95% by 2027, potentially even surpassing the gross margin levels of HBM products. Additionally, SanDisk's long-term agreements already cover approximately two-thirds of its 2028 production capacity, with minimum contracted revenue reaching $93 billion. Given its current market cap of around $239 billion, this suggests its future revenue targets could persist for years, making it difficult to simply classify the company as a traditional cyclical stock.On the cloud computing infrastructure front, CoreWeave has signed agreements to use Nvidia A100 GPUs through 2029. This is a positive development for emerging cloud computing companies such as Nebius and Iren, and it also weakens some investors' bearish thesis centered on the rapid depreciation of older GPUs.AI model companies are also continuing to grow at a pace that exceeds expectations. Frontier AI labs are still maintaining extremely rapid growth rates, and a slowdown in growth would actually be a cause for concern. The market projects that Anthropic's 2028 revenue could reach $190 billion to $200 billion.However, advanced packaging and semiconductor infrastructure remain core bottlenecks. The head of advanced packaging at TSMC has stated that in the coming years, the industry may face not only memory shortages but also tight supply of ABF substrates.Serenity concluded that the AI infrastructure supply chain is continuously expanding. From GPUs, storage, and advanced packaging to power, optical communications, and electronic components, every segment is showing a long-term demand growth trend. The AI supply chain is still in a high-speed development stage.
AMD Announces "Day 0" Support for Alibaba's Qwen3.8 27B Model
Odaily News - AMD has announced "Day 0" support for Qwen3.8 27B, the latest-generation model from Alibaba's Tongyi Qianwen (Qwen) series, enabling developers to run this large-scale open-source AI model locally on AMD hardware on the very day of its release.AMD stated that Qwen3.8 27B is a 27B-parameter intensive model suited for local AI development, continuing the Qwen series' optimization focus on code generation, practical work tasks, scientific research, and long-context AI applications. The model can run via the open-source inference framework llama.cpp on AI PCs and workstations powered by AMD processors, or on a single AMD 32GB graphics card, while also supporting AMD hardware platforms with over 24GB of variable graphics memory (VGM) or VRAM capacity.AMD's preliminary tests show that Qwen3.8 27B delivers strong local inference performance on AMD platforms: up to 24.5 tokens/second on the AMD Ryzen AI Max+ 395 processor, and up to 51.8 tokens/second on a single Radeon AI PRO R9700 GPU. The tests were conducted on Windows systems using the llama.cpp Vulkan backend with multi-token prediction (MTP) optimization enabled. AMD noted that actual performance still has room for improvement as further software and model optimizations are rolled out.
Related news
SK Hynix's Optical Interconnect Memory Roadmap: Connecting Larger Shared Memory Pools via Localized HBM
Odaily News, Citrini analyst jukan stated on the X platform that at first glance, SK Hynix's CPO technology roadmap appears to have one axis pointing toward HBM and another toward optical communication. However, the underlying logic is that AI competition is shifting from individual chip performance to data transfer efficiency across the entire system.Over the past few years, HBM has addressed the problem of GPUs being unable to receive data fast enough. By vertically stacking multiple layers of DRAM and placing them next to the GPU, HBM delivers extremely high bandwidth. Yet, as more HBM stacks are placed around each GPU, packaging area, interposer edge space, power supply, and thermal dissipation capabilities are all approaching their limits. Meanwhile, AI clusters have expanded to thousands or even tens of thousands of GPUs. No matter how fast a single GPU computes, the entire system will still be constrained by the "bandwidth wall" if data cannot be efficiently transferred between GPUs and racks.SK Hynix's vision is to extend optical interconnect to memory. Low-latency, high-bandwidth local HBM will remain next to the GPU, while optical fibers connect it to a larger shared memory pool. This approach avoids limiting all memory capacity within a single GPU package and enables horizontal scaling at the rack level.From an investment perspective, this does not mean HBM will be replaced in the near term. What is more likely to emerge is a new memory hierarchy: frequently accessed data remains stored in local HBM, while larger-scale, less frequently accessed data is stored in optically interconnected memory pools, HBF, or SSDs.SK Hynix is repositioning its business, shifting from selling standardized memory chips to co-designing HBM, controllers, advanced packaging, and system-level memory architectures with customers. If this roadmap comes to fruition, SK Hynix could strengthen customer stickiness, increase product added value, and enhance its ability to secure long-term contracts. At the same time, the company may also more proactively address the potential impact of future memory disaggregation on traditional HBM business models.At the supply chain level, areas that may benefit in the long term include silicon photonics chips, optical engines, lasers, fiber coupling technology, and advanced 2.5D and 3D packaging.However, this concept is still in its very early stages and essentially remains just a roadmap. jukan noted that a person involved in TSMC's packaging business whom he interviewed today was completely unaware of this plan.
Nebius plans to issue $4.5 billion in convertible notes to expand AI data centers and procure GPUs
Odaily News: AI cloud computing company Nebius Group (NASDAQ: NBIS) has announced plans to privately offer $4.5 billion in aggregate principal amount of convertible senior notes to qualified institutional investors, consisting of $2.75 billion in notes due 2030 and $1.75 billion in notes due 2034. If all additional purchase options are fully exercised, the offering size could reach up to $5.175 billion.Nebius stated that the proceeds will be used for data center construction and expansion, developing a full-stack AI cloud platform, expanding data center footprint, and procuring key components such as GPUs, as well as for general corporate purposes. The specific interest rate, initial conversion ratio, and other terms of the notes will be determined at the time of pricing.
Kaito Pulse plugin accused of multiple in-depth data collection practices
Odaily News, ultra posted on X platform stating that Kaito Pulse uses hash computation to fingerprint users' GPU rendering, hardware models, and hardware test audio, binding this unique combination to users' X accounts. Additionally, it can capture users' complete browsing history, feed hover time, clicks, and follow status, sending heartbeat packets with activity detection at 30-second intervals. Furthermore, the tool can read subscription plans and quota consumption ratios for Claude and ChatGPT, automatically click into the Usage page in ChatGPT, and automatically access the Positions tab on Binance, resending authenticated requests to read wallet balances, futures positions, profit/loss, and deposit/withdrawal history. Kaito AI announced its new browser plugin product Kaito Pulse yesterday, which aims to bring off-platform activities directly into the native X timeline.
Krackpot Launches Browser Game, Offering 6 BTC for Cracking Bitcoin Puzzle 71 with GPU
Bitcoin News posted on X platform, stating that Simon Males has launched Krackpot, a browser-based game that allows anyone to use GPU to attempt cracking the private key of Bitcoin Puzzle 71. The prize is 6 BTC. The probability of cracking depends on the number of GPUs: a single gaming GPU would take approximately 830,000 years, 1 million GPUs would take 1 year, and simultaneous participation from all Steam players would take 1 week. The game runs locally in the browser via WebGPU. If a user finds the private key, 6 BTC will be sent to their designated Bitcoin address, with the remainder going to the developer.
Bank of America: Anthropic leads frontier AI model rankings, token prices ease but GPU rents remain firm
According to TechFlow Research, the frontier AI data tracking report released by Bank of America Securities on August 17 shows that Anthropic leads comprehensively in three major AI benchmarks, with Claude Opus 5 ranking first in the Intelligence Index, Agent Index, and Coding Agent Index, GPT-5.6 Sol following closely behind, Meta MuseSpark 1.2 entering the top ten, and Google Gemini 3.6 Flash ranking outside the top ten. In terms of usage, DeepSeek leads with approximately 30% of the Vercel platform token share, Anthropic accounts for 25% and OpenAI accounts for 16%; but in terms of payment amount, Anthropic leads far ahead with 65%, while OpenAI accounts for only 11%. In terms of pricing, the AI Token Price Index decreased 9% month-over-month in August to $2.21, but still increased 87% year-over-year; GPU rental rates remain strong, with H100 increasing 33% year-over-year to $2.77/hour, DRAM increasing 483% year-over-year, and NAND increasing 432% year-over-year. The research report judges that AI infrastructure demand remains healthy, with open-source model usage growing but payment share still highly concentrated on top closed-source models. BofA believes that the Meta "Watermelon" and Google Gemini 4 releases, token pricing trends and GPU rental trends are
Former Bitcoin Miner IREN Completes First Milestone of $9.7 Billion Microsoft AI Contract
According to Cryptopolitan, former Bitcoin mining company IREN announced the completion of the first milestone of its $9.7 billion, five-year contract with Microsoft — the official delivery of the Horizon 1 facility located in Childress, Texas. The facility is 50 megawatts in scale, equipped with Nvidia GB300 systems and direct chip-level liquid cooling technology, and has received the Exemplar Cloud certification awarded by Nvidia. Under the contract terms, Microsoft will complete GPU acceptance verification within five days; upon approval, IREN can commence monthly billing, with expected annualized revenue of approximately $1.94 billion after all four Horizon facilities go online. IREN plans to complete the deployment of all four phases totaling 200 megawatts by 2026, expand AI cloud computing power to 1.2 gigawatts by 2027, simultaneously exit the Bitcoin mining business, and target a full-year AI cloud revenue run rate exceeding $4 billion.