gpu

市场上同时出现两种声音：一边是GPU租金与算力租赁价格仍紧，说明训练与推理需求没有消失；另一边是部分AI token/调用价格在竞争中回落，用户侧成本在改善。乍看矛盾，实则可能是“基础设施仍紧、应用层价格战已开”的分层。
对二级市场的启示是：卖铲人与应用层不能混为一谈。算力紧支撑设备与云基建叙事；token降价则考验应用公司的毛利与差异化。本周ASML、台积电的表述，会给这条分层再加一层证据。
我的看法：别用单一价格信号判断整条AI链。分清你持有的是算力、模型、还是应用，各自的风险点不同。
交易上减少一刀切加减仓，按环节拆开管理。
分层看清之后，配置会更干净：算力与设备看利用率与订单；模型公司看企业付费与毛利；应用层看留存与单位经济。混着买，就容易在错误的环节上加错仓。
短线若再出现「算力紧但应用降价」的标题组合，不要本能地一刀切减仓，先问自己持有的是哪一层。
以上内容仅供参考，不构成投资建议。
#AI #算力 #GPU

Nvidia's ($NVDA.US ) stock continues to climb primarily due to its near-monopoly on the high-performance computing hardware powering generative AI.
Key drivers include:
• Unprecedented Data Center Demand: Cloud providers and AI developers continue to buy Nvidia’s GPUs faster than suppliers can manufacture them, consistently beating revenue expectations.
• The CUDA Software Moat: Developers build AI software on Nvidia’s proprietary CUDA architecture, making it difficult and costly for customers to switch to competing chips.
• High Pricing Power & Margins: Because its hardware is the industry benchmark, Nvidia commands high gross margins across its enterprise and data center segments.
• Capital Returns: Aggressive share buybacks and rising dividend payouts continue to attract institutional investors
#NVDA #GPU #AI

Training and running AI models requires significant computing power, and access to that power is typically controlled by a small number of large cloud providers. Janction aims to change this by aggregating idle GPU resources from around the world and making them available through a blockchain-based marketplace.
This article explains what Janction is, how its technology works, the role of the JCT token, and its listing on Binance Alpha and Binance Futures.
What Is Janction? Janction is a distributed AI computing network that describes itself as the first Layer 2 to provide verifiable, synergic, and scalable AI services. It is built on Arbitrum, an Ethereum Layer 2 scaling solution, and is incubated by Jasmy Corporation, a Japanese data infrastructure company.
Instead of relying on centralized cloud providers such as AWS or Google Cloud, Janction aggregates GPUs from multiple sources, including personal computers, data centers, and other hardware, and makes that capacity available to AI teams and developers.
Janction is designed as a scalable web of blockchains that supports distributed AI multi-party collaboration and resource sharing. It provides underlying infrastructure for computing power trading, data trading, resource scheduling, verification, consensus, and security. Key elements such as AI models, GPU computing power, data feeding, and data labeling are all integrated into Janction for co-processing.
How Does Janction Work? Janction addresses a few core technical challenges that arise when building a distributed, trustless AI computing network:
Resource scheduling and management The Janction cluster GPU pool is a system for managing and allocating multiple GPU resources to support large-scale parallel computing. The system handles task scheduling, load balancing, resource security and isolation, fault recovery, and cost optimization. A key component is the colocation and arithmetic routing layer, which manages how GPU clusters are mixed and matched for different workloads.Data acquisition Janction can source data from both on-chain and off-chain environments. The platform compensates data providers fairly for their contributions, supporting a range of data types required for AI model training and inference.Proof of workload To build a credibly neutral computing network, Janction needs a way to verify that AI computation has been performed as promised. Its proof-of-workload mechanism provides on-chain verification of computing tasks, enabling financial incentives to be distributed fairly to GPU providers. This is conceptually related to how zero-knowledge rollups verify computation without revealing underlying data.Incentive and gaming mechanism Multiple roles participate in Janction's computing tasks, including data providers, data annotators, GPU providers, and AI model providers. Janction uses the Vickrey-Clarke-Groves mechanism (used in game theory) and workload correlation functions to price contributions fairly and ensure that all participants are incentivized to complete their required tasks.Privacy As data privacy regulations tighten globally, Janction integrates privacy-preserving techniques to enable AI model training and inference without creating data silos or exposing personal information. Jasmy's Sovereign Data Protocol is designed to align with the EU’s General Data Protection Regulation (GDPR).
Janction Use Cases Janction currently supports the following AI workloads through its GPU network:
AI image generation: text-to-image generation for designers, advertisers, and social apps.
Speech-to-text and text-to-speech: automated subtitles, AI customer service, podcast narration.
Video enhancement and processing: 4K/8K upscaling, noise reduction, slow-motion smoothing, and old film restoration.
Object detection and image analysis: security surveillance, industrial defect detection, and autonomous driving support.
Private LLM hosting: organizations can run ChatGPT-style AI models privately within their own infrastructure, reducing costs and meeting compliance requirements for regulated industries such as finance, healthcare, and government.
What Is JCT? JCT is the native token of the Janction network. It functions as the medium of exchange for computing power transactions, incentivizes GPU providers, and enables governance participation. The total supply of JCT is 50 billion tokens.
The token distribution follows a deflationary model with a structured vesting schedule designed to limit early sell pressure. The initial circulating supply at launch was approximately 12.7% of the total supply (around 6.35 billion tokens), unlocked for liquidity, community, and airdrop purposes. Full token circulation is expected by 2028.
JCT On Binance Alpha and Binance Futures Janction (JCT) was listed on Binance Alpha on November 10, 2025, and Binance Futures also launched the JCTUSDT Perpetual Contract with up to 40x leverage on the same day.
FAQ What problem does Janction solve? Janction addresses the high cost and centralized control of AI computing power. Training and running AI models typically requires expensive GPU infrastructure managed by a small number of cloud providers. Janction aggregates idle GPUs from multiple sources into a shared marketplace, aiming to reduce AI compute costs significantly while distributing the economic benefits to hardware contributors.
What is the Janction GPU Marketplace? The Janction GPU Marketplace is a decentralized platform where GPU owners can contribute their idle computing capacity in exchange for JCT token rewards. AI developers and teams can then access this pooled capacity for inference, training, and other AI workloads. The marketplace aims to provide unlimited GPU capacity at lower costs by aggregating hardware from many independent providers.
What blockchain is Janction built on? Janction is built on Arbitrum, an Ethereum Layer 2 network. Using Arbitrum allows Janction to process high transaction volumes at low fees, which is necessary for the frequent micropayments made to GPU node operators. Jasmy Corporation's data infrastructure also integrates with Janction for privacy-preserving AI data handling.
What is JCT used for? JCT is the native currency of the Janction network. It is used to pay for GPU computing tasks, reward data providers and node operators, and participate in protocol governance. The token has a total supply of 50 billion and launched on Binance Alpha in November 2025.
How is AI computation verified on Janction? Janction uses a proof-of-workload mechanism that provides on-chain verification of AI computing tasks. This allows the network to confirm that GPU providers have performed the work they were paid for, and to distribute rewards accordingly. The system is designed to be trustless, meaning no central authority needs to verify the computation.
#GPU #JCT #JCTUSDT
$BTC
$ETH
$JCT