News linked to both this project and an event.
Odaily News NVIDIA has launched an AI model routing system called NeMo Switchyard, designed to help developers dynamically allocate AI Agent tasks across multiple large models, reducing costs and latency while maintaining performance.NVIDIA stated that building AI agents does not necessarily mean relying on a single large model. Different models have their own advantages in reasoning capability, response speed, and operating costs. For example, classification tasks may be suited for lightweight models, while complex reasoning requires more powerful frontier models. If all requests are sent to the largest model, costs and latency will increase; conversely, using small models for everything may degrade the quality of complex task completion. NeMo Switchyard uses an intelligent routing mechanism to automatically select the most appropriate execution model among multiple specialized and frontier models based on factors such as task requirements, model capabilities, cost, latency, and system status. The system allows developers to switch between different model providers and model versions without changing the application architecture.NVIDIA explained that NeMo Switchyard offers multiple routing strategies, including a training-free LLM classifier router, Stage Router, Escalation Router, and an adjustable routing model trained on real workload data. Among these, the Escalation Router prioritizes assigning tasks to low-cost models and escalates requests to stronger models when it detects increased task complexity, persistent errors, or stalled execution, thereby achieving a balance between performance and cost.NVIDIA stated that in related tests, NeMo Switchyard significantly reduced AI Agent operating costs while maintaining a high task completion rate by distributing tasks across different models. For example, compared to using only frontier models, the escalation-based routing approach reduced costs by approximately 74% in LangChain multi-turn agent testing, with only 7% of requests requiring calls to frontier models.Additionally, NVIDIA has partnered with companies such as Cognition, Nous Research, Ramp, LangChain, LiteLLM, and Kong to integrate NeMo Switchyard into AI Agent development workflows and enterprise application infrastructure.
According to Yonhap News, Naver D2SF announced the completion of a follow-on investment in physical AI data startup NdotLight (엔닷라이트). This round of financing was led by the Korea Development Bank, with a total size of 15 billion Korean won. This is also Naver D2SF's third investment in the company following the Pre-A round in 2021 and the Series A round in 2022. NdotLight independently developed the 3D data generation solution TRINIX, which can automatically generate high-precision 3D data containing physical properties (mass, friction), joint structures, and collision range information, and is deeply integrated with NVIDIA's simulation platform Omniverse, achieving mass supply of large-scale high-quality 3D simulation datasets.
According to CNBC, Nvidia has signed a memorandum of understanding with Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs, and KKR to jointly establish a financing platform for Nvidia customers, aiming to mobilize over $500 billion in third-party capital for hyperscale data center construction and Nvidia hardware procurement. Nvidia CEO Jensen Huang characterized this as the first time AI chips have become an "investable asset class," stating they possess revenue-generating capabilities, long service lives, and can be transferred across customers, while analogizing compute infrastructure to electricity and the internet. BlackRock CEO Larry Fink defined the project as the "next future of financial engineering" following the securitization of mortgages in the 1970s, and stated that more funds would be raised as soon as possible. Goldman Sachs CEO David Solomon revealed that this collaboration was initiated by Jensen Huang. Currently, some funds have already been raised; the parties will provide financing support for GPUs and data centers through institutional credit, insurance capital, and private capital, helping end users complete AI infrastructure construction without tapping their own balance sheets.
NVIDIA plans to invest up to $3 billion in power infrastructure developer Lancium, with an initial investment of $2 billion to acquire approximately 20% equity, and an additional $1 billion may be added subsequently if milestones such as grid connection are met. Lancium is backed by Blackstone Group, and its "Clean Campus" located in Abilene, Texas is the site of the first phase of NVIDIA's "Stargate" AI supercomputing data center project. The company is currently valued at approximately $10 billion and is evaluating an IPO in 2027.
According to the latest trading data disclosed by Ark Invest, funds under Cathie Wood sold a total of 109,492 shares of Palantir (PLTR) on August 4 and 5, valued at approximately $17 million based on the latest closing price.Palantir previously reported second-quarter revenue of $1.94 billion, up 93% year-over-year, and raised its full-year guidance, driving the stock up nearly 30% in a single day after the earnings release. In addition, Cathie Wood has recently increased positions in Circle (CRCL), SpaceX (SPCX), and NVIDIA (NVDA), while reducing holdings in Shopify (SHOP) and Roblox (RBLX). (The Street)
According to Reuters, Australian AI infrastructure company Firmus completed a $2 billion equity financing, pushing its post-money valuation to over $10.5 billion, nearly doubling the $5.5 billion valuation from its previous funding round in April. The round saw continued participation from NVIDIA and Coatue Management, with funds under Blackstone and Jane Street also providing support.
To address HBM memory supply shortages, NVIDIA is considering reducing the specifications of the next-generation Rubin Ultra GPU. According to sources cited by The Information, NVIDIA has internally tested at least three versions of the Rubin Ultra GPU with lower HBM configurations over the past few weeks, requiring HBM capacity lower than the initially announced specifications. Possible downgrade paths include scaling down from HBM4E to HBM4, and from 12Hi HBM to 8Hi HBM. This move aims to advance product launch plans under supply constraints.
Odaily News Jiaxian Communication related personnel responded to Tongyu Communication's clarification announcement, stating that in March 2026, NVIDIA's AI-RAN technical team visited Jiaxian Communication, and both parties exchanged views on the AI-RAN technology roadmap. During the exchange, NVIDIA recommended that Jiaxian Communication use the CUDA Aerial open-source platform for AI-RAN base station system R&D. Since then, Jiaxian Communication has initiated pre-research and validation of AI-native 5G and 6G base station technologies based on the CUDA Aerial ecosystem, and related work has been steadily progressing for nearly six months.The aforementioned personnel stated that both parties have now established a regular technical working group to maintain communication on related technology R&D and validation. More than ten technical personnel from NVIDIA, from regions including Mainland China, Hong Kong, and Singapore, are participating in the working group. (Jiemian)
Odaily News: AI research startup Mirendil has entered into a multi-year partnership agreement with Google Cloud to secure large-scale computing resources in support of its "Self-Improving AI" research and development. Under the agreement, Mirendil will gain access to TPU and NVIDIA GPU computing resources provided by Google Cloud, as well as managed AI training clusters, to develop AI systems capable of continuously optimizing their own capabilities. It is reported that Mirendil is focused on advancing "Recursive Self-Improvement" AI, in which AI systems enhance their own performance through iterative refinement, self-learning, and optimization. This direction is also a research area of interest among some of the top AI laboratories today.Benham Neyshabur, co-founder and CEO of Mirendil, revealed that the total value of the agreement exceeds $100 million, roughly equivalent to half of the $1 billion valuation seed funding round the company completed at the end of June. (TechCrunch)
According to Fortune magazine, Musk stated on the earnings call following SpaceX's initial public listing that SpaceX, leveraging the engineering advantages of rocket scientists, will dominate the AI computing power competition, comparing it to "the New York Yankees versus a minor league team." SpaceX's Q2 AI revenue reached $2.6 billion, a quarter-over-quarter increase of 213%, with capital expenditures reaching $18.4 billion. The company has signed $14.1 billion in cloud service contracts, with clients including Anthropic and Google, and announced it will fully adopt the NVIDIA Vera Rubin architecture, targeting 20 gigawatts of computing power deployment by the end of 2027.
Odaily News: At the company's first earnings call, SpaceX CEO Elon Musk stated that SpaceX's AI services will run "exclusively" on NVIDIA systems in the future.Musk noted that the company considers NVIDIA's Vera Rubin architecture to be the "best AI computing architecture." SpaceX expects its computing capacity to exceed 2 gigawatts (GW) by the end of this year, with plans to increase it to nearly 10 gigawatts by the end of next year.Additionally, SpaceX plans to deploy NVIDIA Vera Rubin NVL72 rack-scale systems both on the ground and in space, using them for the "Starmind" satellite project, with related satellites expected to begin launching next year.
Odaily News: NVIDIA CEO Jensen Huang announced on X that the company is launching Alpamayo 2 Super, a cutting-edge open reasoning model designed for the autonomous driving sector. He stated that the next phase of AI development will shift focus from software intelligence to robotics, and autonomous vehicles will become a key gateway to the era of robotics.Huang noted that Alpamayo 2 Super can not only "see" the road environment but also understand and reason about complex scenarios, thinking before taking action. It can serve as the core AI model for robotaxis, autonomous trucks, shuttle buses, delivery vehicles, agricultural robots, and future large-scale mobile robot systems.It is reported that NVIDIA is making the model available for commercial use under the OpenMDW-1.1 license, allowing developers and enterprises to research, fine-tune, and deploy it. Huang stated that open models will help enhance the safety and reliability of autonomous driving systems.
Odaily News: Bitdeer announced that its subsidiary Tydal Data Center AS has signed a 16-year data center hosting and services agreement with Volta Tydal AS, which will provide 121 MW of IT load capacity (approximately 133 MW total power) at the Tydal AI/HPC campus in Norway, all to be deployed with NVIDIA GPUs to serve a leading AI laboratory. The total expected contract payments during the base term are approximately $4.7 billion, with an 8-year renewal option attached, bringing the potential total contract value after renewal to approximately $8 billion. (Stocktitan)
According to Bloomberg, AI cloud infrastructure company Volta Infra Holdings announced the completion of $300 million in venture financing, achieving a valuation of $2.4 billion, co-led by Andreessen Horowitz and Altimeter Capital, with participation from NVIDIA and Michael Dell, founder of Dell Technologies. Additionally, Volta secured support from a $5 billion customer financing pool designed to help small and medium-sized AI enterprises lower the barrier to purchasing NVIDIA high-end chips. The company also disclosed it has signed a $10 billion, six-year cloud computing service contract with an unnamed leading AI developer. The contract will be fulfilled in partnership with Bitdeer, delivered via its 133 MW data center in Norway. Volta was co-founded earlier this year by former Brookfield Asset Management infrastructure executives Ricard Boada and Sofia Gumuzio. It has currently secured 1 GW of data center power resources and plans to develop new sites in Texas and Wyoming, aiming to deploy multi-gigawatt compute capacity before 2030.
Odaily News NVIDIA has released Vera CPU storage benchmark results, showing that its NVIDIA BlueField-4 STX storage processor can significantly improve encryption, compression, integrity verification, and data recovery performance in AI-native storage, helping enterprises address the growing data processing demands of the Agentic AI era.NVIDIA stated that as AI agents perform knowledge retrieval, call tools, manage long-term memory, and handle larger context windows, storage systems are no longer just simple data read/write components, but have become a critical link in the AI inference pipeline. Large volumes of data need to be encrypted, compressed, verified, and recovered along the storage path, tasks typically handled by the CPU, which can become a performance bottleneck for AI infrastructure. Benchmark results show that the BlueField-4 STX storage processor equipped with the Vera CPU delivers performance improvements over a comparison x86 CPU across multiple storage tasks: AES-128 encryption performance improved by up to 1.43xAES-128 decryption performance improved by up to 1.29xReed-Solomon data recovery performance improved by up to 3.26xCRC32C integrity check performance improved by up to 3.67xCompression performance improved by up to 3.29xDecompression performance improved by up to 1.72xCompression + encryption multi-stage storage pipeline performance improved by up to 3.21xNVIDIA said that while traditional CPUs typically require adding cores, power consumption, and cooling costs to scale storage processing capabilities, AI-native storage needs to maintain low latency under higher concurrency and larger data volumes. By improving per-unit CPU resource processing capability, Vera helps storage systems support more AI agent workloads without significantly increasing infrastructure costs.
Odaily News: NVIDIA has officially released Video Codec SDK 13.1, offering developers multiple video encoding and decoding upgrades, including AV1 layered reference mode, zero-copy transcoding architecture, precise frame-level search, enhanced hardware decode statistics, and a new Docker development environment, further improving performance and efficiency in AI video processing, streaming, and large-scale content distribution scenarios. On the decoding front, SDK 13.1 adds per-macroblock decode statistics for H.264/H.265 without requiring additional CPU parsing overhead.NVIDIA stated that as demand for high-quality video streaming, generative AI video tools, remote collaboration, and large-scale content distribution continues to grow, Video Codec SDK 13.1 will help developers build more efficient, lower-latency, and more scalable video processing pipelines.
HIVE Executive Chairman Frank Holmes stated that the company's cluster of 504 NVIDIA B200 GPUs at Bell Canada's Manitoba AI fabric generates approximately $2.90 per GPU per hour in revenue; by comparison, HIVE's Bitcoin mining equipment generates approximately $0.12 per hour. HIVE's total revenue for fiscal year 2026 reached $298 million, up 158% year-over-year; digital currency revenue from Bitcoin mining grew 164%. During the same period, average hash rate stood at 22.2 EH/s, up 290% year-over-year, accounting for approximately 3% of the Bitcoin network's total hash rate, and the company mined 2,885 BTC. HIVE's BUZZ HPC division, which houses its AI and high-performance computing business, generated revenue of $19.5 million, up 94% from $10 million in the prior fiscal year. The company also signed GPU cloud agreements worth approximately $220 million with Bell and AI company Cohere, and raised $75 million through a note issuance to fund AI infrastructure expansion. HIVE is building a 320-megawatt AI data center in the Greater Toronto Area, with plans to eventually house more than 100,000 GPUs. The company stated that if the facility becomes fully operational in the second half of 2027, it could generate approximately $360 million in annualized recurring revenue.
Odaily News Four.Meme has announced the launch of the Stock Meme feature. Projects launching tokens on Four.Meme can now choose bStocks US stock tokens as the trading liquidity pool. The first batch supports NVDAb (NVIDIA), with more US stock assets to be added gradually.This feature is jointly launched by Four.Meme, bStocks, and Binance Wallet with the support of the BNB Chain ecosystem, making it one of the first officially supported US stock liquidity pools on BNB Chain. Four.Meme will collaborate with bStocks to provide liquidity support for each bStocks pool on the platform. Additionally, select Stock Meme projects may be featured in the Binance Wallet US Stock Meme section.
According to the official announcement, to meet users' diversified investment needs, Bitget's US stock token rToken has listed rJMKE (Jersey Mike's Subs). As of now, the Bitget platform supports a total of 608 US stock tokens. It is reported that rTokens, identified by the letter r + stock ticker (e.g., NVIDIA as rNVDA), are issued by Bitget's licensed RWA protocol Reality, directly connecting to global liquidity pools such as Nasdaq and NYSE through cooperation with compliant broker Alpaca. Its features include: underlying assets 1:1 reserved and custodied by licensed custodians, stock dividends distributed 1:1 in token form, support for synchronized mapping of corporate actions (stock splits, etc.), and holdings can serve as joint margin for unified accounts and USDT-margined contracts, allowing users to flexibly manage funds while holding global stock assets.
According to TechFlow Research, Morgan Stanley released a research report on July 27, quantifying for the first time the Incremental Return on Invested Capital (ROIC) of Generative AI investments. The report constructed three estimation frameworks: the ROIC for hyperscale cloud service providers' GPU leasing business is approximately 31%, the ROIC for proprietary infrastructure model API business is approximately 46%, and the ROIC for third-party compute API business is approximately 25%. Under base case assumptions, a single 1 gigawatt (GW) data center is configured with approximately 410,000 NVIDIA GB300 GPUs, with a utilization rate of 75% and an hourly leasing price of $8.5. The combined capital expenditure of the three major cloud giants is expected to exceed $1.4 trillion. Morgan Stanley maintains an Overweight rating on Microsoft, Amazon, Meta, and Google. The report points out that as AI moves from the training phase to the inference phase, demand for GPU compute power will continue to grow. Providers with self-built compute infrastructure will achieve considerable profits by leveraging their pricing power in an ecosystem where compute is scarce. If Morgan Stanley's calculations hold true, the hundreds of billions of dollars in AI capital expenditure will shift from being perceived as "costs" to "growth assets".