CRWV and NBIS Surge Post-Earnings! Is AI Computing Power Taking Center Stage?

Publicado a 2026-08-13Actualizado a 2026-08-13

Resumen

Nebius and CoreWeave represent the purest new cloud logic for AI computing power, the NeoCloud thesis. Bloom Energy is the second curve for AI computing power's electricity shortage. The A-share mapping should be analyzed along these two lines.

This wave from Nebius and CoreWeave is the purest NeoCloud logic for AI computing power. Bloom Energy represents the second curve of AI computing power's electricity shortage. The A-share mapping should be analyzed along two lines:

NBIS / CRWV Mapping: AI Computing Power Leasing + IDC + Hardware

Their business model = hoard large quantities of GPUs + build/lease data centers + lease computing power. Q2 single contracts > $1 billion, 70% prepaid, payback period of 1 year and 10 months — it's the sweetest business right now.

The corresponding A-shares are at the very upstream of the computing power industry chain:

Core - Computing Power Leasing / New IDC Players (The NBIS of Computing Power):

• RunZe Technology 300442: The largest intelligent computing center cluster in the Beijing-Tianjin-Hebei region, comparable to CoreWeave's data center + computing power leasing

• GDS Holdings 300383 / OFILM Data 300738 / Kehua Data 002335: Tier-1 city IDCs, having land and power means having computing power

• Capital Online 300846 / QingCloud Technology 688316 / UCloud 688158 / Century Huatong 002602: Pure cloud service providers, already offering GPU cloud services

• Hainan Huatie 603300: Transitioned from construction machinery to computing power leasing, hoarded a large number of H800s

Computing Power Hardware - Who Sells the Shovels:

• Servers/Complete Machines: Foxconn Industrial Internet 601138 / Inspur Information 000977 / Unisplendour Corporation 000938 / ZTE 000063 / Cambricon 688256 / Hygon 688041

• Optical Modules/High-Speed Interconnect, the most elastic link in AI computing power: Zhongji Innolight 300308 / Eoptolink 300502 / TFC Optical Communication 300394 / Yuanjie Semiconductor 688498

- As long as GPU shipments exceed expectations, these companies' earnings will explode first.

Logic: NBIS Q2 cloud business YoY +514%, with room for further price increases, indicates the AI computing power gap is larger than imagined. A-share computing power leasing companies are currently in a state of demand > supply, locking in 3-year contracts in advance.

2. Bloom Energy Mapping: Electricity for AI Data Centers

The direct trigger for BE's 12.56% surge was Nebius stating on the earnings call that the power supply solution for its Vineland, New Jersey data center was changed to Bloom Energy's Solid Oxide Fuel Cell (SOFC), providing on-site power that is stable, quiet, and ultra-low emission.

The biggest bottleneck for US AI data centers now is not GPUs, but the lack of power. The grid can't connect fast enough, making on-site fuel cell power generation the quickest solution.

A-share Corresponding:

Fuel Cells / SOFC / Hydrogen Energy:

• Dongfang Electric 600875: The only domestic company capable of mass-producing SOFC stacks, Bloom Energy's counterpart in China

• SinoHytec 688339 / Weichai Power 000338 / Vision Group 002733 / Snowman 002639 / Broad Ocean Motor 002249: Leading fuel cell system companies, benefiting from both hydrogen energy + data center backup power dual expectations

• Fuel Cell Materials: SinoCera 300285 / Concept

Data Center Power / Thermal Management / Microgrids, the real beneficiaries:

• Power Supply/Microgrids: Kehua Data 002335 / KSTAR 002518 / Sinexcel 300693 / Hopewind Electric 603063 / Jinpan International 688676 • Diesel Generation + Thermal Management: Shenling Environment 301018 / Envicool 002837 / Gaolang 300499 / Ketai Power 300153

Logic: Previously, fuel cells were about the hydrogen energy story; now it's about the story of AI computing power's rigid demand. BE has risen over 180% from its April low, the core reason being its transformation from a hydrogen energy stock to an AI power supply star.

Reference Portfolio Ideas:

For seeking high elasticity, look at Zhongji Innolight, Eoptolink, RunZe Technology, Foxconn Industrial Internet — these are the clear plays mapping NBIS.

For seeking expectation gaps, look at Dongfang Electric, Kehua Data, Shenling Environment — these are the hidden lines for Bloom Energy in A-shares. The market hasn't fully priced in SOFC powering data centers yet.

This round's inclusion of CRWV and NBIS into the Nasdaq 100 marks the official establishment of AI NeoCloud as a main theme, not short-term speculation, but the second wave of computing power demand.

Not investment advice. A-share computing power sector is highly volatile, pay attention to timing.

Lecturas Relacionadas

Crypto valuations could double as protocols link revenue to tokens: Bitwise CIO

Bitwise CIO Matt Hougan believes crypto valuations could double as protocols increasingly use their revenue to fund token buybacks and burns. He argues that crypto beyond Bitcoin is transitioning to a revenue-driven market, where network activity directly boosts native token value, a shift he says investors have not yet fully priced in. Hougan cites protocols like Hyperliquid, Uniswap, Aave, Pump.fun, and Lighter as examples that use fees to repurchase or remove tokens from circulation. He expects more DeFi applications and layer-1 networks to adopt similar mechanisms in the next 12-24 months. This stronger link between protocol revenue and token value could provide conventional valuation metrics, though Hougan notes token holders lack the legal rights of traditional shareholders. He attributes this shift partly to a more permissive US regulatory environment, which is easing securities-law concerns that previously deterred such features.

cointelegraphHace 23 min(s)

Crypto valuations could double as protocols link revenue to tokens: Bitwise CIO

cointelegraphHace 23 min(s)

Oxford and NUS Propose the Next-Generation World Model Direction: The Mental World Model is Here

Researchers from Oxford University and NUS propose Mental World Modeling (MWM), a framework that integrates mental variables like beliefs, goals, and social norms into world state representations, alongside physical variables. They argue that conventional world models, focused solely on physical dynamics, often produce predictions that are physically plausible but behaviorally incorrect in human-centric scenarios, as actions depend on both external and internal mental-social states. To validate MWM, they introduce MENTIS, a modular, training-free benchmark system that forces LLMs to perform explicit reasoning steps: parsing a scene into a coupled physical-mental state, generating agent-specific observations, simulating the joint state transition for candidate actions, and evaluating outcomes. Experiments on 8 LLM-based models show that explicit MWM reasoning significantly improves action prediction accuracy (F1 score rising from 63.3 to 87.9 on average), with the largest gains in social interaction scenarios. A key finding is that the primary remaining bottleneck is the simulation of state transitions. The work formalizes MWM and demonstrates its necessity for systems where behavior depends on variables not inferable from physical scenes alone.

marsbitHace 48 min(s)

Oxford and NUS Propose the Next-Generation World Model Direction: The Mental World Model is Here

marsbitHace 48 min(s)

Wall Street Morning Report: CPI Lands Mildly, U.S. Stocks Stage Deep V-Shaped Reversal, AI Cloud, Storage, and Optical Communications Collectively Surge

Wall Street Morning Report: US stocks staged a deep V-shaped recovery following a mild CPI report, with AI cloud, storage, and optical communication sectors leading the rally. Major indexes were mixed: the Dow dipped slightly, while the S&P 500 and Nasdaq gained, nearing record highs. The July CPI data met expectations, with core inflation hitting its slowest pace since March 2021, further reducing expectations for a September Fed rate hike. Market focus shifted decisively towards AI infrastructure. The Philadelphia Semiconductor Index jumped 2.49%. AI cloud providers NEBIUS and CoreWeave surged 34% and 19% respectively on strong earnings and massive order backlogs. Optical communication stocks like Lumentum and Coherent also posted significant gains. The storage sector rallied broadly, with SK Hynix and Micron among the advancers. Meanwhile, major tech giants saw divergence. Nvidia rose over 3%, while Apple, Microsoft, Amazon, and Meta closed lower. In commodities, gold and silver remained strong, while oil prices retreated as markets awaited developments in the Strait of Hormuz. The US federal budget deficit widened significantly in July, raising long-term concerns. Key upcoming events include US PPI data, Sandisk's investor day, and the SEC's 13F filings deadline, which will reveal institutional holdings.

marsbitHace 48 min(s)

Wall Street Morning Report: CPI Lands Mildly, U.S. Stocks Stage Deep V-Shaped Reversal, AI Cloud, Storage, and Optical Communications Collectively Surge

marsbitHace 48 min(s)

A 22-Year-Old Mathematical Puzzle, Solved by a Union Hospital Intern?

In an astonishing development, a longstanding mathematical problem known as the Crouzeix conjecture, which had challenged experts in numerical linear algebra for 22 years, appears to have been solved by Shanmu Jin, a neurosurgery resident and postdoctoral researcher at Peking Union Medical College Hospital. With no formal advanced mathematical training—his background is in geology and medicine—Jin relied on self-study and, crucially, the AI model GPT-5.6. The conjecture, proposed by mathematician Michel Crouzeix in 2004, concerns a fundamental constant (speculated to be 2) bounding the relationship between matrix norms and polynomial values over numerical ranges. It is critical for applications in matrix function analysis and numerical methods. Despite dedicated efforts by leading mathematicians, the best proven constant had only been reduced to 2.414. Jin approached the problem using sophisticated prompt engineering with GPT-5.6. He adapted a known prompting strategy, isolating the AI from external resources to force original reasoning, employing multiple divergent "sub-agents" to explore different paths, and enforcing rigorous adversarial review of proposed proof steps. After about 16 hours of autonomous, unsupervised operation, the AI produced a novel and elegant proof. The key insight involved a clever "sampling strategy" that reduced the problem to a simple positivity condition—a solution described as elegant and unexpected by experts. The proof was verified by the conjecture's originator, Michel Crouzeix, and other specialists like Alex Townsend and Anne Greenbaum, who expressed astonishment at its validity. Jin has made the entire process open-source, including the prompt, drafts, and formal verification code. Remarkably, just eight days after Jin's preprint appeared, mathematicians Emiel Lorist and Felix Schwenninger published an independent, concise 5-page proof using a different approach, also developed with the aid of ChatGPT 5.6. Jin welcomed this complementary work. This event marks a potential turning point, demonstrating how AI can enable experts from non-traditional backgrounds to solve deep theoretical problems and dramatically accelerate scientific discovery, heralding what some are calling a new "golden age" for interdisciplinary research.

marsbitHace 1 hora(s)

A 22-Year-Old Mathematical Puzzle, Solved by a Union Hospital Intern?

marsbitHace 1 hora(s)

Metaplanet CEO shuts down Bitcoin sale speculation after $322M transfer

Metaplanet CEO Simon Gerovich has clarified that a recent transfer of 5,014 BTC (worth $322 million) was a routine custody operation and not a sale. The company's holdings remain at 43,000 BTC, cementing its position as the third-largest publicly traded Bitcoin treasury firm and the largest in Asia. Despite sitting on an unrealized loss of approximately $1.4 billion, Metaplanet continues to expand its Bitcoin strategy. It has established a ventures arm with a $25 million pledge to invest in Japan's crypto infrastructure and aims to hold 100,000 BTC by the end of 2026 and 210,000 BTC by the end of 2027.

cointelegraphHace 1 hora(s)

Metaplanet CEO shuts down Bitcoin sale speculation after $322M transfer

cointelegraphHace 1 hora(s)

Trading

Spot