Market AnalysisNews

Delivers insights into price action, technical indicators, market forecasts, and future trends. Data-driven analysis helps investors understand market dynamics and identify potential opportunities for informed decision-making.

OpenAI Unveils Its Own Jalapeño Chip: An Accelerator 1.5–2 Times More Efficient Than Nvidia

On August 25, 2026, OpenAI unveiled test results for its custom inference accelerator, Jalapeño. On the public InferenceX benchmark, systems using the new chip delivered 1.5–1.9x more computations per watt at peak throughput and reduced response latency by 1.7–3.6x compared to systems based on Nvidia GB200 and GB300. The chip, rated at 700W nominal power, consumed up to 550W under tested loads, with a block of 128 units reaching 1.7 exaflops in 4-bit precision. OpenAI plans to deploy Jalapeño in its infrastructure by late 2026, marking the first generation of a multi-year platform. Jalapeño was developed in collaboration with Broadcom (for the die and networking) and Celestica (for boards and racks) over nine months. It is specifically designed for OpenAI's own predictable, high-volume inference workloads around its language models, computational kernels, and data movement, unlike Nvidia's general-purpose accelerators. The initiative is backed by an agreement with Broadcom to deploy 10 GW of OpenAI's custom accelerators between late 2026 and 2029. While OpenAI will continue relying on external suppliers like Nvidia for training cutting-edge models and parts of inference, the company aims to control the processor architecture for its largest daily computational stream. This shift addresses the economics of inference: by building chips as internal components, OpenAI avoids paying the market premium associated with Nvidia's high-margin commercial GPUs, directly lowering the cost per query. This move follows a trend where major AI players (e.g., Anthropic with Google TPUs) transition to custom silicon as their inference volume becomes predictable. However, specialization risks reducing flexibility for future AI architectures and creates a single point of failure with manufacturing partners. The challenge for OpenAI will be ensuring Jalapeño remains competitive through multiple future model generations.

cryptonews.ru16h ago

OpenAI Unveils Its Own Jalapeño Chip: An Accelerator 1.5–2 Times More Efficient Than Nvidia

cryptonews.ru16h ago

Why Trillion-Dollar Institutions Hesitate to Go On-Chain? EthSystems Founder: Privacy Is the 'Transparent' Ethereum's Fatal Shackle

"Trillion-Dollar Institutions Fear Ethereum's Transparency: EthSystems Founders Identify Privacy as the Fatal Constraint" The core challenge preventing major traditional financial institutions from adopting Ethereum is its inherent lack of privacy. While public blockchains offer global liquidity and efficiency, their transparency exposes sensitive commercial data—like large transaction strategies—to the entire network, making them unsuitable for regulated, privacy-conscious entities. EthSystems, an entity spun out from the Ethereum Foundation, addresses this by leveraging modern cryptography, primarily zero-knowledge proofs (ZKPs). Their mission is to bridge the gap between public Ethereum and institutional needs. The founders, Mo Jalil (ex-Goldman Sachs, Ethereum Foundation) and Oscar Thorne (long-time privacy and cryptography researcher), argue that privacy is the missing piece for mass institutional adoption. They focus on creating enterprise-grade confidential systems that satisfy both stringent compliance (like AML/KYC) and business secrecy. The problem is not a lack of cryptographic primitives; many exist. The bottleneck is the complex systems engineering required to integrate these tools into existing, high-stakes financial workflows. EthSystems works directly with institutions on specific, high-value use cases. Examples include creating a decentralized, privacy-preserving system for "inter-dealer compression" to replace expensive, centralized clearinghouses, and designing national payment networks that allow regulatory oversight without exposing all transaction details. Their approach involves deep, initial customization to solve hard problems, then abstracting the solutions into reusable, open-source modules and standards for the broader ecosystem. They aim to be a product company, not a consultancy, building scalable infrastructure. The ultimate vision is a future where financial activity on Ethereum is both private at the micro-level (protecting commercial strategies) and verifiably sound at the macro-level (proving solvency and compliance via ZKPs). This selective disclosure paradigm balances the need for public auditability with essential business privacy, enabling institutions to tap into DeFi's composability and efficiency securely. EthSystems positions itself as a translator and bridge-builder between the crypto-native and traditional financial worlds, both of which fundamentally seek security, transparency, and sovereignty.

marsbit2 days ago 02:56

Why Trillion-Dollar Institutions Hesitate to Go On-Chain? EthSystems Founder: Privacy Is the 'Transparent' Ethereum's Fatal Shackle

marsbit2 days ago 02:56

The Battle for Control of the Tracks Enters the Second Half: Banks vs. Crypto, Who Will Have the Last Laugh?

The competition for control over the tokenization infrastructure, or the "rails," is intensifying, moving beyond initial asset listing to dominance over settlement, custody, and regulatory layers. Recent developments signal a shift in power towards traditional finance. Key evidence includes: the formation of the BankChain Alliance by 39 U.S. state banking associations to launch a banking-owned blockchain network; moves by market infrastructure giants like DTCC, ICE, and Citadel Securities to establish their own institutional-grade on-chain systems; the struggle of crypto-native custodians like ZeroHash (re-applying for a bank charter) and Copper (facing a severe valuation drop), highlighting that regulatory "license moats" are now more critical than technical advantages; and the launch of stablecoin USD1 by licensed trust bank BitGo on the permissioned Canton network, showing convergence of stablecoin issuance towards regulated entities. The analysis concludes this is not a simple "banks vs. crypto" battle but a redefinition of the foundational infrastructure. A clear division of labor is emerging: open public chains for DeFi and innovation, while bank-led consortium chains and licensed entities capture institutional settlement, tokenized deposits, and regulated custody. The defining question is no longer *if* an asset is tokenized, but *on which rails* it runs and *who controls* those rails, with regulation and牌照 providing the ultimate backstop.

marsbit08/27 10:16

The Battle for Control of the Tracks Enters the Second Half: Banks vs. Crypto, Who Will Have the Last Laugh?

marsbit08/27 10:16

Morgan Stanley Research Report Analysis: Google at a 12% Premium, Meta at a 30% Discount - Internet Giants' Valuations Diverge

In a volatile end to the summer for internet stocks, Morgan Stanley's latest valuation report highlights a significant divergence in how the market is pricing major tech giants. While the sector's average forward EV/EBITDA multiple trades below historical averages, individual company valuations are splitting, largely driven by perceived AI capabilities. Alphabet (Google) is the standout, trading at a 12% premium to its 3-year average on an EV/EBITDA basis. This premium is attributed to upward revisions for its AI hardware (TPU) sales potential, advancements in its Gemini models, and cloud margin expansion. In stark contrast, Meta trades at a deep 30% discount to its 2-year average, as concerns over social media advertising competition outweigh its progress in AI-powered ad tools. Amazon sits in the middle, trading at a moderate discount. The report notes a broader sector trend: EV/Sales multiples are expanding while EV/EBITDA multiples are contracting, indicating the market is placing a higher premium on profitability over pure revenue growth. Sub-sectors like e-commerce and digital media face the most valuation pressure. Furthermore, adjusting earnings for stock-based compensation (SBC) reveals significantly higher "true" valuation multiples across the board. Morgan Stanley concludes that valuation recovery for the internet sector will depend on upward revisions to profit estimates—driven by catalysts like new AI product cycles and cloud growth stabilization—rather than simple mean reversion. The firm maintains an "Attractive" view on the sector.

marsbit08/27 06:16

Morgan Stanley Research Report Analysis: Google at a 12% Premium, Meta at a 30% Discount - Internet Giants' Valuations Diverge

marsbit08/27 06:16

He Personally Built o1 and o3, but Suddenly Announced: Humans Can Retire Forever

AI researcher Jerry Tworek, former lead of OpenAI's reasoning models (o1/o3), predicts human AI researchers will be largely obsolete within two years. He argues that while AI agents currently excel at execution (reducing experiment cycles from a month to a day), their creative idea generation remains low-quality. True frontier AI development, he claims, is driven by only 30-50 people globally, with everyone else in supporting roles. A core bottleneck is the Transformer architecture, which he believes is suboptimal but has seen only 3-4 serious replacement attempts at OpenAI in seven years due to massive computational and validation hurdles. He highlights a key paradox: you need significant compute to prove a new direction, but you need proof to get the compute. A breakthrough came when he received dedicated GPU resources, leading to scaling reinforcement learning and the o1 model. Now, through his company Core Automation, he uses AI agents to automate costly R&D tasks. In one case, a $100k AI effort optimized a GPU kernel by 60x, matching rare expert-level performance. This acceleration, he argues, will soon automate even the remaining creative and architectural work of the few top researchers. Looking ahead, Tworek envisions a post-work future where humans pursue excellence for its own sake—like ancient philosophers or dedicated athletes—free from the necessity of labor for societal function. He acknowledges the personal challenge of deriving self-worth from work but is actively building the tools that will make his own expert role, and those of his peers, redundant. The prevailing sentiment among researchers, he notes, is a mix of exhaustion and exhilaration in what may be the final, pivotal chapter of human-led AI advancement.

marsbit08/27 00:26

He Personally Built o1 and o3, but Suddenly Announced: Humans Can Retire Forever

marsbit08/27 00:26

Circle CEO: Stablecoins Are at the Internet's 2002 Stage, Will Reach Trillions of Dollars in the Future

Circle CEO Jeremy Allaire, in a Q2 2026 earnings AMA, discussed the current state and future of stablecoins and Circle's strategy. He compared stablecoins today to the internet in 2002, predicting they will grow from hundreds of billions to trillions of dollars. Key points include: * **Current Use Cases**: Stablecoins have achieved product-market fit in digital asset markets (for trading/settlement), as a digital dollar store of value in emerging markets, and for cross-border payments and settlement. * **Future Growth Areas**: Allaire highlighted opportunities in the AI agent economy, merchant payments (especially via QR codes and stablecoin cards), and the convergence of traditional and on-chain finance. * **Circle's Strategy**: Circle aims to grow USDC through global partnerships. Its economic engines will include reserve income, transaction fees from its on-chain payment network (CPN), and its upcoming "economic operating system," Arc. * **Arc's Vision**: Arc, launching its mainnet on September 16, is a stablecoin-native blockchain designed for seamless user and developer experience. It aims to power the future "on-chain" economy where businesses and AI agents operate. * **Global Adoption**: Allaire emphasized that stablecoin adoption is a global phenomenon, driven by regulatory clarity in regions like Europe (MiCA) and the US (GENIUS Act), and will continue regardless of specific US legislation like the CLARITY Act. * **EURC Growth**: Circle's euro stablecoin, EURC, has surpassed €400 million in circulation, benefiting from early preparation for European regulations and existing distribution networks. Allaire expressed confidence in Circle's execution, citing strong team cohesion and the adoption of AI tools, while identifying cybersecurity and global local operations as key areas for continued strengthening.

marsbit08/26 10:12

Circle CEO: Stablecoins Are at the Internet's 2002 Stage, Will Reach Trillions of Dollars in the Future

marsbit08/26 10:12

活动图片