News linked to both this project and an event.
Indian AI company Sarvam has launched Token Factory, offering inference for Chinese open-source models such as DeepSeek, Kimi, GLM, and Qwen on GPUs within India. As a beneficiary of this sovereign AI initiative, the company is now caught in a debate over whether to develop its own models or act as a distributor.
Odaily reports: According to Axe Compute (NASDAQ: AGPU)'s Q3 2026 financial report, the company's third-quarter revenue was $7.2 million, up 125% quarter-over-quarter; operating income was $6.8 million, net income was $11.5 million, and diluted earnings per share was $0.52. As of September 30, the company's cash and cash equivalents increased to $147.7 million, with an adjusted EBITDA loss of approximately $2.3 million.On the digital asset front, Axe Compute holds approximately 2.6 billion Aethir (ATH) tokens, with digital assets and related receivables totaling approximately $31.6 million in value. The company recorded approximately $10.5 million in digital asset gains in the third quarter, primarily from fair value changes in ATH holdings and related receivables.On the business side, the company has completed the acquisition of a Georgia-based computing cluster equipped with 2,304 NVIDIA B300 GPUs, at a transaction cost of approximately $124.4 million. Currently, Axe Compute has signed contracts in the United States and Europe totaling over $3 billion in value, and expects an annualized revenue run rate exceeding $700 million once the related projects are fully deployed.
Perplexity launches Frontier Research, aiming to build an AI system capable of autonomous research, backed by over 10,000 GPUs.
Solmate plans to provide AI compute services via an asset-light model and is expected to support payments in USD or stablecoins on the Solana network, while currently negotiating GPU compute agreements.
Odaily Report: The Solana Foundation has issued an announcement stating that Solana Confidential Compute, an open-source tool developed in collaboration with the Nito team, includes an SDK and on-chain verification program and is now open in beta on Solana devnet. The tool is designed for building private AI applications and verifying compute hardware. The mainnet version has not yet been given a release date and has not been audited.Confidential virtual machines generate signing keys, and hardware reports bind the keys to recent Solana slots; after on-chain programs verify them, they record the keys, hardware types, and expiration times. AI computation is performed off-chain, and outputs are signed by the key. After verifying the signature, applications can query the on-chain records, and conversation content and private inputs are not made public. The beta supports Intel TDX and AMD SEV-SNP hardware attestation, while NVIDIA GPU attestation is still under development. The Solana Confidential Chat example encrypts prompts after verifying the environment in the browser and does not send them if verification fails; the application server can see visitor IPs, and Nito can see accounts, models, request times, and ciphertext sizes. Nito has integrated the tool into its private AI platform.
Odaily Report: NVIDIA CEO Jensen Huang's sudden resignation has sparked concerns within South Korea's AI industry regarding GPU supply and AI infrastructure cooperation. In response to market worries about potential disruptions to next-generation GPU supply and AI partnerships, the South Korean government assessed and stated that GPU supply under existing contracts is not expected to be affected by this management change. Currently, approximately 13,100 NVIDIA B200 GPUs have been deployed, and plans to subsequently introduce next-generation GPUs such as Vera Rubin are also expected to proceed as originally scheduled.It is reported that NVIDIA has partnerships with South Korean companies including NAVER, Hyundai Motor Group, SK Telecom, and Kakao Mobility in areas such as AI data centers, autonomous driving, and robotics. The market believes that the short-term impact on existing contracts will be limited, but the pace and priority of future new investments, technical support, and strategic cooperation will still depend on the decisions of the new management. (Yonhap)
Odaily — Megaspeed International, the company involved in the Jensen Huang incident, has issued an official announcement stating that it places high importance on its obligation to comply with U.S. export control laws, that compliance is the company's top priority, and that it is committed to constructive and transparent cooperation with the U.S. government.Megaspeed has cooperated with an inspection by the U.S. Department of Commerce's Bureau of Industry and Security (BIS) and stands ready to fully cooperate with other relevant U.S. government agencies. To further strengthen its compliance commitment, Megaspeed has engaged a prominent U.S. law firm to lead a comprehensive inventory of the company's graphics processing units (GPUs) and to independently verify the company's export control compliance practices. Megaspeed stated that it welcomes appropriate government oversight and will demonstrate through concrete actions that it operates in a responsible and transparent manner and complies with applicable U.S. laws.It is understood that Megaspeed is a cloud computing and data center company registered in Singapore and an important customer of Nvidia. The New York Times previously reported, citing trade data, that between June 2024 and February 2026, Megaspeed's affiliated company in Malaysia procured advanced AI servers and other equipment, including Nvidia chips and related products. Huang Le previously served as Megaspeed's chief executive officer and board chairman and was one of its main controllers; she has appeared alongside Jensen Huang multiple times at public business events.
Analyst Jukan posted on X that Nvidia chips remain in short supply, but to sustain the scarcity premium, the company needs to continue supporting emerging cloud service providers (Neoclouds). If the new CEO takes a more cautious approach to this business model, it could put pressure on GPU spot prices, reduce cloud service providers' profit margins, and push them to turn to alternatives such as AMD or Google TPU, thereby weakening Nvidia's market dominance and supply chain control.Jensen Huang has been working hard to maintain Nvidia's growth flywheel, which not only helps expand chip demand but also ensures compute utilization for emerging cloud service providers. He reportedly hoped to gradually step back once Neoclouds were able to independently order next-generation GPUs, but it remains uncertain whether the successor CEO is willing to wait. According to The Wall Street Journal, the current CEO appears to lack such patience.Previously, Nvidia (NVDA.O) interim CEO and Chief Financial Officer Colette Kress had repeatedly raised objections to high-risk deals pushed by Jensen Huang. People familiar with the matter said Kress may adopt a more prudent operating strategy than Huang. She had previously opposed Nvidia's plan to provide up to $250 billion in credit support to OpenAI; the deal was ultimately split, and Nvidia currently only committed to providing about $100 billion in credit support. With data center projects delayed and cloud computing deals canceled, Kress had already begun diversifying Nvidia's financial risk exposure to a small number of customers.
According to Yonhap News Agency, South Korea's Ministry of Science and ICT held an operational committee for its Memorandum of Understanding (MOU) with AMD on November 4 in Seoul to review the implementation status of the agreement signed by both parties this July. During the meeting, Lee Do-gyu, director of the Information and Communications Policy Office at the ministry, and Keith Strier, Senior Vice President of Global AI Markets at AMD, along with other attendees, discussed proposals to integrate AMD's CPUs and GPUs with domestically produced Korean NPUs to build heterogeneous AI computing infrastructure. Both sides also reviewed the progress of initiatives such as the establishment of an AMD AI research center, support for universities, and AI talent development programs, and discussed specific implementation plans and subsequent timelines.
Runware stated that the manufacturing time for a single module will be reduced from three weeks to one day, with installation and commissioning completed within 48 hours after delivery to the customer site. Customers can either rent computing capacity or provide their own GPUs, paying monthly module and operational fees while power and network costs are billed separately. The company said it has secured $1 billion in letter-of-intent commitments since launching external sales, covering more than 20 megawatts of capacity.
Odaily reports: Bitcoin mining company IREN has announced the completion of a $363 million financing facility arranged by Wingspire Equipment Finance, secured by AI computing equipment that is already operational and generating revenue, to refinance recently purchased GPU servers and networking equipment. The facility was provided by six lenders, with Eldridge Capital Management and Silver Point Capital making major funding commitments. The financing is primarily secured by the relevant equipment and supported by multi-year customer contracts, with specific interest rates and terms not yet disclosed.This refinancing will help IREN recover capital already invested in operational infrastructure for further expansion. IREN is transitioning from Bitcoin mining to AI cloud services. Its AI cloud business revenue for the fiscal year ending June 30, 2026, reached $128.8 million, up from $16.4 million in the previous fiscal year; Bitcoin mining revenue still reached $578.2 million. The company reported a net loss of $702.6 million for the same period, including $638.8 million in non-cash asset impairments primarily related to retired Bitcoin mining machines. (theenergymag)
According to Investing.com, NASDAQ-listed IREN Limited (IREN) has completed a $363 million equipment financing deal with Wingspire Equipment Finance to refinance GPU servers and networking systems deployed in data centers.
According to TechCrunch, NVIDIA participated in the $14 million Series B financing round for Indian startup Simplismart, which was led by Dallas Venture Capital with participation from Accel, Shastra VC, and Micromax. The transaction values the four-year-old company at approximately $100 million, including the new investment. Simplismart states that its software sits between AI models and compute infrastructure, automatically adjusting model deployment and management to optimize response time, request throughput, and compute utilization, enabling customers to better leverage their existing GPU resources. The company noted that its software helped AI video generation platform InVideo reduce GPU service costs by 56%. Simplismart CEO Amritanshu Jain told TechCrunch that the company’s annualized revenue run rate grew 14 times over the past six months to reach an eight-figure amount. This figure is based on recent customer usage rather than contractual recurring revenue, and excludes revenue passed through to infrastructure providers. Although the company serves more customers in India, the United States accounts for more than half of its total revenue.
Odaily reports: Investment firm Seawolf Capital says it is shorting cloud computing company CoreWeave. The firm was founded by three traders known for the book and film "The Big Short," and co-founder Vincent Daniel said the short position began with an analysis of CoreWeave's business model.Seawolf Capital points out that CoreWeave rents out AI compute powered by high-end GPUs, a model similar to businesses that purchase specific assets and then lease them out. The firm believes the economic lifespan of GPUs may fall short of oil drilling equipment, aircraft, or shipping equipment, and that competing chip compute supply is rapidly increasing.Vincent Daniel said CoreWeave needs to continuously invest large amounts of capital to purchase high-priced AI chips. Based on its assessment of the company's 2027 free cash flow gap and balance sheet leverage, the company may need to raise tens of billions of dollars in new funding on favorable terms. Seawolf Capital says its short thesis is focused on capital availability and cost, not a bearish view on the overall AI trade. (Business Insider)
In his article "The Trifecta," Arthur Hayes proposes establishing a compute trading market for AI agents through Flop Network, using the FLOP token as a payment medium, and developing forwards, futures, and perpetual contracts based on spot compute prices. Hayes states that the market currently lacks a unified global benchmark for compute pricing, and existing rental price indices for GPUs, CPUs, and TPUs, along with AI model token pricing data, have limitations. Flop Network plans to calculate compute prices based on the latency, computational workload, and transaction fees of inference tasks, and will form a reference benchmark through an open market.
decentralized cloud computing network Akash Network has published an article announcing the selection of Solana as its blockchain migration target. Akash provides decentralized GPU and AI computing power rental services and previously operated an independent blockchain based on Cosmos SDK. In October 2025, Akash announced plans to deprecate its existing independent chain and transition to a shared security architecture, with both the Solana and Ethereum ecosystems under consideration at the time.
According to Reuters, Nvidia has secured a CPU order from New York-based AI cloud computing startup Modal. Previously relying solely on Intel and AMD chips for its AI agent workloads, Modal plans to transition to Nvidia's Vera central processing units. The company stated that early tests show Vera can double its AI agent workload capacity compared to existing Intel and AMD hardware. Modal CEO Erik Bernhardsson said the firm will maintain a hybrid chip deployment strategy, noting that next-generation chips from Intel and AMD may narrow the performance gap. Nvidia stated that Vera is the next-generation data center CPU designed for AI factories, marking another step in expanding its business beyond GPUs. Nvidia also added that OpenAI, Anthropic, and Oracle plan to deploy Vera.
Odaily News: LayerZero Labs, in collaboration with the a16z Crypto research team, Carnegie Mellon University, and Imperial College London, has released new research findings on Aerie, aimed at improving the aggregation efficiency of post-quantum cryptographic signatures.According to the introduction, Aerie is built upon the previously launched lattice-based polynomial commitment scheme Akita. On a MacBook Pro equipped with an Apple M4 Max chip, using 16 CPU cores without GPU acceleration, it can aggregate 32,768 Falcon-512 signatures in 0.65 seconds, with peak memory usage of approximately 2.1 GiB.LayerZero stated that compared to the previous Orthus scheme, Aerie achieves approximately 1.5x faster proof generation, 38x faster verification, and approximately 1.7x smaller proof size. Additionally, Aerie supports aggregation of ML-DSA signatures standardized by the National Institute of Standards and Technology (NIST), with specific performance data to be released later.This research aims to provide a technical pathway for blockchain to achieve low-latency, post-quantum secure consensus without requiring specialized hardware.
According to CNBC, Microsoft officially released its third-generation in-house AI accelerator chip, the Maia 300, on October 12, announcing plans to use the Maia series as the primary architecture for running AI models to gradually reduce reliance on NVIDIA GPUs. The Maia 300 is slated for production in 2027, with each data center rack capable of hosting 96 chips (doubling over the previous generation), increasing AI token processing volume per rack by seven times and improving token processing efficiency per megawatt by 40% compared to GPUs. Microsoft aims to deploy multi-gigawatt-scale Maia chip clusters, with Microsoft 365 Copilot expected to be the first to benefit from the performance improvements. Analysts at Evercore ISI estimate that by Q1 2027, Maia chips will account for approximately 10% of Microsoft’s AI chip spending. Notably, Maia accelerators are currently reserved exclusively for internal use and are not available for external leasing; NVIDIA and AMD will continue to participate in Microsoft's model training operations.
Starknet has published an introduction to its Quantum-Safe Bitcoin (QSB) scheme, stating that StarkWare researchers have verified a quantum-safe locking mechanism on the Bitcoin mainnet that requires no soft fork, no new opcodes, and no changes to consensus rules. The first QSB output was spent on August 26, 2026, in Bitcoin mainnet block 964,199, with the associated GPU computation cost reduced from approximately $320 to about $50. QSB builds its locking mechanism through hash computation, but each transaction still requires substantial computation, and moving assets into the scheme still requires a traditional transaction that carries quantum risk. The transaction does not comply with Bitcoin Core's default relay rules, and there is currently no accompanying wallet, address format, or recovery process. StarkWare stated that QSB is an emergency research solution, and that the long-term solution still requires the introduction of post-quantum signature mechanisms via a soft fork. Starknet's full quantum resistance also depends on the progress of Ethereum's underlying upgrades.