Messari:一文速览稳定币市场的主要竞争者

marsbitPubblicato 2024-09-25Pubblicato ultima volta 2024-09-26

Tether上季度破纪录的盈利能力使其跻身tradfi巨头之列。但52亿美元这笔巨额利润也让它成为了那些想要分得一杯羹的新竞争对手的目标。

市场

本文我们来深入快速发展的稳定币世界,涉及到中心化领域和去中心化领域。

稳定币全览

我们将做以下几类的垂直细分:

市场

PYUSD

中心化稳定币往往缺乏透明度,而且往往只在具有明确激励的条件下才会出现高交易量。PYUSD是少数收获信任的10亿美元市值规模的中心化稳定币之一。

市场

USD0

通过空投激励以及与Morpho等DeFi平台的合作整合,像USD0这样更加去中心化的金库支持型稳定币迅速增长。USD0市值约为2.5亿美元,很快就达到了这一目标。

市场

Elixir的deUSD

像USDe这样的合成稳定币使用“多头现货+空头期货头寸”来维持其挂钩。由于基差被压缩,USDe失去了市场份额。但像Elixir这样的新协议旨在通过调整其抵押品支持来改进Ethena模式。

市场

GHO

专注于最大限度地去中心化和最大限度地减少人为干预的稳定币一直以来都没有显出太多的需求。GHO可能是个例外,因为它利用了AAVE上持续增长的活跃用户群。

市场

DYAD

创新的设计通常试图实现某种机制来改进典型的抵押债务。

头寸模式。DYAD就是这样的一种稳定币,它的目标是通过另一种名为KEROSENE的代币利用系统内多余抵押品。KEROSENE允许用户抵押外自己的外源资产铸造更多的DYAD。而且,NFT(NOTE)持有者的KEROSENE越多,他们从流动性池中获得的收益就越多。

市场

这些类别的新稳定币都在收益率、可及性、流动性、稳定性和资本效率方面相互竞争。新设计或对旧设计的调整都会涉及到各种利弊权衡:

市场

每周都有新的稳定币进入市场,稳定币领域格局也在不断发展变化。

Letture associate

The Evolution Path of Physical Bitcoin

The Evolution of Physical Bitcoin Bitcoin's digital nature is its core strength, enabling self-custody and rapid global transfers. However, its intangibility also hinders mainstream adoption. For over a decade, creators have attempted to materialize Bitcoin while preserving its cash-like properties, yielding notable results. Casascius Coins, launched in 2011, were the first and most iconic physical Bitcoin. Creator Mike Caldwell generated private keys offline, printed them on coins, and sealed them with tamper-evident holograms. This model relied on user trust in the centralized issuer. Production ceased in 2013 due to regulatory pressure from FinCEN. RavenBit Coins emerged in 2014 aiming to decentralize minting by letting users generate and apply their own keys. However, this led to trust issues with numerous untrusted minters and insecure key generation methods. In 2016, Coinkite introduced Opendimes—a breakthrough in bearer asset technology. These USB-shaped devices generate and store keys internally. Funds can be received by checking the public key, but spending requires physically breaking the device to extract the private key. While innovative and open-source, its cost (~$20) and form factor limit its use for small, everyday transactions. Satochip's Satodime, a card-shaped device using similar secure chip technology, followed. It supports NFC interaction and comes in various forms. While potentially cheaper in bulk (~13€), it remains a high-security hardware wallet, not a low-cost cash substitute. A fundamental cost barrier exists. For physical Bitcoin to achieve widespread commercial use, hardware costs must drop below $1 to match the production cost of fiat banknotes. Current secure chips capable of running Bitcoin's cryptographic algorithms (like secp256k1) are too expensive. Chips like NXP's NTAG X DNA (~$3) show cost-reduction potential but lack native Bitcoin curve support. Projects like OfflineCash embed chips in banknote-like paper, but face challenges with durability, the need for custom Bitcoin-enabled chips, and the inherent requirement for users to verify balances online—which conflicts with Bitcoin's trustless ideal. Coinkite's Tapsigner, a ~$20 card with a proprietary Bitcoin NFC chip, is seen as a more practical step forward. It functions as a reloadable hardware wallet for contactless payments, solving the "change" problem and focusing on real-world retail integration, a direction also pursued by companies like Cash App and Square. In summary, the journey to physical Bitcoin has progressed from trusted centralized mints (Casascius) to user-generated keys (RavenBit) and finally to self-contained secure hardware (Opendimes, Satodime, Tapsigner). The core challenge remains developing a sufficiently low-cost, durable, and truly trustless physical bearer asset that can function like cash in daily transactions. Current solutions are either too expensive or introduce new trust assumptions, keeping the ideal of ubiquitous physical Bitcoin just out of reach for now.

marsbit48 min fa

The Evolution Path of Physical Bitcoin

marsbit48 min fa

Samsung Relies on Technology Cycles, SK Hynix on HBM, How Did Micron Win a Trillion-Dollar Market Cap?

Micron Technology, the third-largest memory chip maker alongside Samsung and SK Hynix, recently saw its market cap surpass $1 trillion. Founded in 1978 in Boise, Idaho, Micron survived brutal industry cycles while American peers and Japan's memory sector faltered. Its survival is attributed to a dual strategy: leveraging political and legal avenues for critical breathing room, coupled with relentless manufacturing cost control. Historically, Micron sought U.S. government intervention three times. In 1985, it filed an anti-dumping complaint against Japanese firms, leading to the U.S.-Japan Semiconductor Agreement. Ironically, this created an opening for Samsung, which later became its toughest competitor. In 2002, Micron turned "whistleblower" in a DRAM price-fixing investigation, escaping penalties while rivals were fined. In 2017, it sued China's Fujian Jinhua, contributing to its placement on a U.S. entity list, stifling a nascent competitor. However, a major strategic misstep occurred in 2013 with the acquisition of bankrupt Japanese firm Elpida. Integrating Elpida's mobile-DRAM-focused technology diverted resources, causing Micron to miss the critical early decade of development for High Bandwidth Memory (HBM)—the high-performance memory essential for AI chips like NVIDIA GPUs. By the time AI demand exploded in 2022, SK Hynix, which launched the first HBM in 2013, held about 85% of the HBM3 market, leaving Micron with roughly 3%. Micron now faces a triple squeeze. In the high-end HBM market, it lags significantly behind SK Hynix and Samsung. In the mid-to-low end DRAM market, it faces aggressive price competition from China's CXMT. Furthermore, a 2023 Chinese cybersecurity ban on its products slashed its revenue from China, a once-core market, from over 10% to just 7.1% by FY2025, causing it to exit China's data center server business. Beneath its political maneuvering lies Micron's core strength: exceptional manufacturing efficiency and cost control. Decades of engineering have yielded DRAM chips with a smaller cell area than rivals, meaning more chips per wafer and lower unit costs. This efficiency, not subsidies, has allowed it to withstand price wars. While political leverage bought time, Micron is now paying a "time debt" in the HBM race. It is racing to ramp up HBM3E production and develop HBM4, but catching up to competitors who started a decade earlier is a monumental challenge. Its future hinges on whether its expertise in cost control and political strategy can compensate for the lost time in a technology race where early-mover advantage is decisive.

链捕手1 h fa

Samsung Relies on Technology Cycles, SK Hynix on HBM, How Did Micron Win a Trillion-Dollar Market Cap?

链捕手1 h fa

New AMD Paper Overturns Conventional Wisdom: FP4 Training Instability's Cause Is Not Insufficient Randomness

AMD's new research challenges the conventional understanding of FP4 training instability. While reducing precision from FP8 to FP4 promises doubled computational throughput and is supported by new hardware like NVIDIA Blackwell and AMD MI350 series, training large language models natively with FP4 has been notoriously unstable, often attributed to insufficient stochasticity. The paper "Pretraining large language models with MXFP4 on Native FP4 Hardware" demonstrates successful end-to-end FP4 pre-training of Llama 3.1-8B on AMD MI355X GPUs using the MXFP4 format, achieving a 9-10% overall speedup over FP8. Crucially, it identifies the root cause of instability: not randomness, but the accumulation of *structural micro-scaling errors* along the sensitive weight gradient (Wgrad) path. Through controlled experiments, researchers found that quantizing the Wgrad operation to FP4 caused significant convergence degradation. Counterintuitively, common stochasticity-based mitigation techniques like stochastic rounding and randomized Hadamard transforms worsened performance. In contrast, applying a *deterministic* Hadamard transform successfully stabilized training by ensuring consistent error patterns, reducing the extra token cost from 26-27% to just 8-9%. This work has significant implications: 1) It provides a clear diagnostic for low-precision training instability, steering focus towards structural errors. 2) It pushes FP4 from a primarily inference-focused format into the realm of viable training. 3) It leverages the open OCP Microscaling (MX) standard, promoting cross-vendor compatibility. The research marks a critical step towards more economical large model training by further pushing the boundaries of low-precision computation.

marsbit1 h fa

New AMD Paper Overturns Conventional Wisdom: FP4 Training Instability's Cause Is Not Insufficient Randomness

marsbit1 h fa

Trading

Spot
Futures
活动图片