LecturaInvestigaciónRadarMarco de inversión
Iniciar sesión / Registrarse
TodoTecnologíaProductosInversiónBolsaOtros

Lectura

2026-11-17124 publicaciones

Actualizar

Man Group: Japanese Banks Are No Longer Cheap, Contrarian Opportunities Shift to AI Hardware and Heavy-Asset Industrials

Tema · 曼集团日本AI与工业轮动Material
2026-11-17 07:00 GMT+8

Investors often treat 'value' (buying cheap stocks) and 'contrarian' (buying unpopular stocks) as the same strategy, equating 'growth' with 'momentum'. For the most part, these pairs do overlap—cheap stocks are typically out of favor, and high-growth companies usually enjoy strong price momentum. But when they diverge, the best opportunities often emerge. Nick Wilcox, Managing Director of Discretionary Equities at Man Group, believes that the Japanese stock market is currently experiencing both of these separations simultaneously.

Over the past five years, large Japanese banks have served as a textbook case: transforming from deep value into a crowded momentum trade. In 2021, large Japanese banks traded at just half their net asset value (a 0.5x price-to-book ratio), long shunned by investors. As inflation returned and the Bank of Japan began raising interest rates, bank profits improved, prompting investors to flock to the sector. By late October 2026, these banks had risen to nearly 2x book value, accounting for over 40% of the weight in Japan’s large-cap value index. With Japanese interest rates now rising to a more normal level of approximately 1.5%, the easy gains from this re-rating appear to be behind us. Wilcox also points out that banking is one of the more obvious white-collar industries that AI could eventually disrupt.

Meanwhile, hardware companies within Japan's AI supply chain—including optical fiber cable, power equipment, specialty semiconductor, and flash memory chip manufacturers—have experienced the reverse process. At the height of the AI rally in June, these companies accounted for roughly 18% of the weight in Japan’s TOPIX Index; during the pullback in November, their weight dropped to approximately 11.5%. When the first wave of selling hit in July, buying the dip was still a growth bet, as valuations had not yet fallen to traditionally cheap levels. But by November, indiscriminate selling had pushed several profitable hardware suppliers down to deep-value multiples: leaders in optical cabling and automotive chips fell to 10 to 14 times expected earnings, while memory chip companies, after dropping up to 70% from their June peaks, compressed to 2.5 times consensus earnings for the current year and 1.7 times for next year. Wilcox emphasizes that even if profits at these memory companies were to fall by three-quarters next year, they would still sit on single-digit earnings multiples.

The core of Wilcox’s argument is not to predict where AI will go, but to highlight the asymmetry implied in current pricing. The market worries that US tech giants may slow the pace of data center spending, but a more measured rollout could actually lengthen the hardware cycle, giving AI revenues time to catch up with capital investment. Amidst this uncertainty, he suggests pairing cheap hardware suppliers with neglected 'old economy' value stocks to hedge against either scenario.

He identifies two categories of mispriced 'old economy' assets. The first category consists of what he calls 'HALO' businesses—companies with heavy assets and low obsolescence rates, such as steelmakers, shipping lines, and mining equipment manufacturers. If AI construction continues, grid upgrades and data centers will require vast quantities of copper, steel, and energy equipment; if the AI bubble deflates, these asset-intensive businesses trade at low multiples and are highly unlikely to be disrupted by software. The second category is Japan’s automotive sector. Concerns over US tariffs have pushed export-oriented automakers into the deepest value territory in the Japanese market—once cash and cross-shareholdings are stripped out, their core businesses are priced almost at zero. However, Wilcox argues that pricing these companies solely as tariff victims ignores their decades of accumulated factory automation capabilities; one leading automaker plans to deploy 400,000 robots across its factories next year.

Fuentes:man.com

Investigación

Rockwell Automation Launches EdgeAI Industrial PCs with NVIDIA GPUs for On-Premises Workloads

Resumen rápido
2026-11-17 20:00 GMT+8

US industrial automation giant Rockwell Automation (NYSE: ROK) announced on November 17, 2026, the launch of new GPU-enabled industrial computing products designed to expand its ability to run demanding AI and machine learning workloads closer to production. The initial offerings include the 6300B-SB1 EdgeAI 4000 industrial PC powered by an NVIDIA RTX PRO 4000 Blackwell GPU, and the 6300B-SB1 EdgeAI 2000 equipped with an NVIDIA RTX 2000E Ada Generation GPU.

These industrial computers provide additional processing capacity for compute-intensive applications at the supervisory layer, targeting use cases such as advanced machine vision, model predictive control, predictive maintenance, and small language models that offer real-time advisory support to operators. Rockwell states that the new hardware integrates with its existing FactoryTalk Optix HMI and visualization software, allowing manufacturers to process data within local facilities to meet requirements for low latency or limited cloud connectivity.

Matheus Bulho, Senior Vice President of Software & Control at Rockwell Automation, noted that this strategic partnership aims to bring intelligence where decisions are made—close to equipment and operational context—helping industrial systems transition from fixed, deterministic automation toward more intelligent, adaptive operations. The new line complements rather than replaces Rockwell's established control and visualization architectures, many of which already support AI capabilities without discrete GPUs, giving manufacturers flexibility in choosing their computing architecture based on workload and environment.

Fuentes:prnewswire.com

Investigación

GoDaddy Launches Airo AI Builder: An Agentic Website Tool for SMBs

Resumen rápido
2026-11-17 18:49 GMT+8

On November 17, 2026, GoDaddy launched Airo AI Builder, described as an "agentic operating system" designed to help small and medium-sized businesses (SMBs) turn ideas into websites quickly. According to TechRadar, the tool goes beyond initial setup, allowing users to continuously refine and rebuild their sites through a simple conversational interface, supporting everything from portfolios to online stores and booking systems.

At launch, Airo AI Builder supports over 90 third-party integrations, including Slack, Notion, and HubSpot, focusing on tasks like database access and customer relationship management. GoDaddy's research indicates that service-based businesses accounted for a significant portion of early site creations, such as health and body (13%), home and property (12%), and business services (11%). These sectors typically lack extensive tech teams, making effective website building a challenge.

For e-commerce, GoDaddy offers its own Payments platform and has partnered with Stripe to allow users to collect payments. Selina Bieber, VP of marketing-led growth at GoDaddy, stated that Airo is positioned to help users take the first step and keep improving as their customers' needs and their business evolve.

Fuentes:TechRadar

Investigación

Schneider Electric's Mark Deguara: Liquid Cooling Readiness Must Be Verified Per Row; Hall-Level Headroom Cannot Be Directly Promised to AI Tenants

Tema · 施耐德电气液冷就绪度Material
2026-11-17 22:40 GMT+8

Imagine a new cloud tenant requesting a colocation provider in Western Sydney, Australia, to deploy a row of liquid-cooled GPU racks within an occupied data center. On paper, the site has spare floor space and backup electrical capacity. However, before quoting rack density or delivery dates, the operator must confirm whether the target row can support the new load while ensuring that existing tenants continue to receive their contracted services. In an assessment method published by Mark Deguara, Head of Technology Innovation at Schneider Electric, in November 2026, it is pointed out that liquid cooling readiness is specific to a particular rack row, not a general indicator for the entire hall.

This judgment changes the way capacity commitments are verified. Mark Deguara requires operators to confirm six dimensions per row: the sustained and peak electrical loads of the racks, the proportion of heat captured by the liquid loop versus the remaining air-cooling needs, the coolant temperature, flow, and pressure requirements, the complete electrical path from the grid connection to the rack distribution (including UPS and protection settings), floor loading and installation access (including the impact of CDU maintenance space on the number of leasable rack positions), and retrofit economics. In the Western Sydney case study, the central chiller plant may have overall headroom, but the branch serving the target row lacks sufficient flow during peak demand—a constraint that only becomes apparent through per-row verification.

Heat rejection verification must be conducted under local summer design conditions and planned maintenance states. Direct-to-chip cooling captures heat from high-power components via cold plates, while liquid-to-liquid Coolant Distribution Units (CDUs) transfer this heat to the facility water system. Operators must confirm the available temperature, flow, and differential pressure at the connection point while existing tenants are drawing cooling, and compare the CDU supplier's water quality requirements with the building-side filtration conditions. For rows without access to building cooling water, Heat Dissipation Units (HDUs) reject heat back into the room air; in this scenario, the room cooling system must simultaneously handle the heat discharged by the HDU and the residual heat from the cold plates. In mixed halls, rear door heat exchangers can intercept hot rack exhaust where the existing water system and rack configuration support them.

Retrofit economics depend on shared infrastructure and tenant deployment plans. Mark Deguara recommends comparing three options—targeted retrofit, dedicated hybrid zones, and independent expansion—using the same load and service conditions. If pipe routing, floor loading, or the outage required for grid integration makes live retrofitting unfeasible, prefabricated modular IT pods serve as an alternative option. Highly customized AI-fit facilities are difficult to repurpose after the lease expires, which impacts residual value assumptions. Acceptance records should capture the tested capacity, operational limits, and responsibility boundaries on both sides of the rack connection; the tenant agreement defines services based on this foundation rather than on the hall's nominal figures.

Fuentes:Schneider Electric Blog

Investigación

CoreWeave Claims First Multi-Rack NVIDIA Vera Rubin Production Results in MLPerf, Citing 2.7x Speedup for DeepSeek-V3

Tema · CoreWeaveMaterial
2026-11-17 00:15 GMT+8

US-based AI cloud infrastructure provider CoreWeave published a blog post on November 16, 2026, stating that it has become the first cloud provider to deliver multi-rack NVIDIA Vera Rubin NVL72 cluster results in the MLPerf Training v6.1 benchmarks from a production environment. CoreWeave reported that it trained the DeepSeek-V3 671B Mixture-of-Experts (MoE) model to target quality in 17.65 minutes using 256 NVIDIA Rubin GPUs across four Vera Rubin NVL72 racks. In comparison, its NVIDIA GB200 NVL72 submission at the same GPU count took 48.40 minutes, representing a 2.7x improvement in time to train.

The results were submitted under the Closed division’s Available category of MLCommons. CoreWeave emphasized that these performance figures reflect its full-stack engineering capabilities, including high-performance networking configuration, topology-aware workload placement, and communication collectives optimization. At multi-rack scale, cross-rack traffic is handled via a RoCE-based NVIDIA Spectrum-X Ethernet fabric (leveraging 102.4 Tb/s Spectrum-6 networking) and NVIDIA ConnectX-9 SuperNICs to minimize communication overhead and keep GPUs productive. Additionally, CoreWeave reported single-rack Vera Rubin NVL72 performance on Llama 3.1 8B pretraining, reaching target quality in 7.16 minutes, which was 1.9x faster than its 72-GPU NVIDIA GB200 NVL72 result in the same round.

It is important to note that CoreWeave explicitly states that the speedup multiples are derived by CoreWeave from published MLCommons results and have not been verified by the MLCommons Association. While CoreWeave claims its Vera Rubin NVL72 clusters are running in production today, the "first" designation is also based on CoreWeave’s analysis of public submissions as of November 16, 2026. For readers tracking AI compute deployment, this signals a shift from next-generation NVIDIA hardware specifications to actual large-scale production application, though specific performance advantages await third-party reproduction or broader industry benchmark comparisons for further confirmation.

Fuentes:wf.coreweave.com

Investigación

a16z Invests in First Person Labs, Betting on Privacy-First Personal AI Context Models

Material
2026-11-17 23:10 GMT+8

Venture capital firm a16z announced on November 17, 2026, an investment in startup First Person Labs. The company is developing a product called "Second," designed to provide continuously updated personal context models for AI agents.

According to a16z's announcement, Second's core architectural feature is that its background engine runs on the user's device by default, making the generated personal model inspectable and portable. Users can choose the intelligence powering it and control which applications and agents can access this data. This design aims to solve the problem of personal data being stranded inside individual apps while emphasizing user control over sensitive digital life data.

First Person Labs was co-founded by Ken Liu and Erik Chi. The announcement notes that Ken Liu previously focused on AI and data privacy research at Stanford University and Google DeepMind, studying what models remember and how information can be protected. Erik Chi worked on network security, privacy, and censorship resistance at the University of Michigan. Together, they built the Open Anonymity Project to provide more private access to AI.

The investment did not disclose specific amounts, round stage, or valuation terms. In its statement, a16z noted that as personal agents become useful in everyday life, whoever controls the user's context model will have considerable influence over agent behavior, arguing that users should be in total control of it.

Fuentes:a16z.news

Investigación

Voyage Labs Raises $12 Million Seed Round for Specialized AI Travel Assistant

Tema · Voyage LabsResumen rápido
2026-11-17 23:00 GMT+8

Voyage Labs, an AI startup founded by former Google Maps Product Director Manik Gupta, announced a $12 million seed funding round led by Kleiner Perkins, with participation from SignalFire and SV Angel. The company plans to use the funds to expand supplier partnerships and grow its engineering team.

The company's flagship product, Atlas, is a personal AI travel planning assistant currently available in private beta. Unlike most products that include travel planning as a feature within general-purpose agents, Atlas focuses exclusively on end-to-end travel services, from inspiration to booking and management. Co-founder and CTO Yogesh Jain noted that trip planning is a project with its own lifecycle spanning days or weeks, making long-running agent technology a natural fit for the use case.

Despite facing stiff competition from general-purpose AI assistants like Gemini, Gupta believes that travel and outdoors represent a large enough vertical to require a focused product solution. Voyage Labs is betting that by integrating real-time price monitoring, multi-week conversational memory, and final flight/hotel booking capabilities, it can deliver a more seamless user experience than generalist models.

Fuentes:TechCrunch

Investigación

EMC Clarifies Nvidia Revenue Share Never Exceeded 20%, Refuting Media Forecast of 25-30%

Tema · 台光电子Material
2026-11-16

Taiwan-based copper-clad laminate manufacturer EMC Materials Co., Ltd. (stock code: 2383.TW) issued a major announcement on November 16, 2026, to clarify recent media reports regarding its business relationship with Nvidia. Previously, Economic Daily News cited a report from Taishin Securities Investment Advisory predicting that Nvidia would account for 25% to 30% of EMC's total revenue in 2026.

In the official statement, EMC explicitly stated that Nvidia's contribution to its revenue has "never exceeded 20%." The company emphasized that the significant discrepancy between the reported figures and reality could mislead investors, prompting the clarification. EMC also noted that it does not provide financial forecasts and characterized the media reports as speculation, urging investors to rely solely on announcements made through the Public Information Observation Platform for accurate financial and operational information.

Fuentes:emctw.com

Investigación

Executive Coach Meerbott: The First 90 Days After Funding Is a Dangerous Window for Founder Judgment

Material
2026-11-17 20:00 GMT+8

The moment the funding clears, the founder has just completed the most difficult sales job of their career. Immediately following this, the board shifts from patience to having a timeline, numbers that were once mere expectations become commitments, and the individual is required to make the best decisions of their career immediately while operating on sleep deprivation. Executive coach Kelly Meerbott writes in Crunchbase News that she has coached approximately 1,500 leaders, including venture-backed founders; this pattern repeats so frequently that she considers the first 90 days post-funding a dangerous window.

The mechanism Meerbott cites comes from published neuroscience research. Amy Arnsten, in a review in Nature Reviews Neuroscience, wrote that even quite mild acute uncontrollable stress can cause a rapid and dramatic loss of prefrontal cognitive abilities. The prefrontal cortex is responsible for weighing options and maintaining plan stability. Experiments by Porcelli and Delgado in Psychological Science showed that volunteers stressed with ice water became more cautious about potential gains but more willing to take risks regarding potential losses. Meerbott maps this pattern onto founders post-funding: they begin to hesitate on the bets that justified the current round, while becoming willing to gamble everything on anything that feels like a threat.

These decision-making errors rarely feel reckless because the nervous system is treating uncertainty as an emergency. Meerbott describes three specific manifestations: In hiring, a CEO offers a generous package to a head of sales simply because they know the person, mistaking familiarity for certainty; in expansion, a new market originally planned for 12 months is compressed into 3 months—the data hasn't changed, only the waiting has become unbearable; in co-founder relationships, one party falls silent due to being overwhelmed, while the other interprets that silence as distrust. She cites Harvard Business School researcher Noam Wasserman, noting that 65% of high-potential startups fail due to co-founder conflict.

Meerbott proposes four rules to protect judgment during this window: Do not finalize senior hires, market entries, or equity changes in the meeting where they are first proposed within the first 90 days; write down criteria and sleep on it; before major decisions, ask yourself your pressure level (1 to 10), adding a second conversation if the score is high; schedule 30 minutes weekly for co-founders to discuss each other's well-being rather than business matters; and give the board a better question—'Which decision are you about to make that you wouldn't make if you were well-rested?' She advises boards to start paying attention the week funds clear, as composure may indicate stability, or it may mask exhaustion.

Fuentes:Crunchbase News

Investigación

DeepMind Spinout Isomorphic Labs Raises $5B Series C to Advance AI Drugs to Clinical Trials

Material
2026-11-17 22:50 GMT+8

Isomorphic Labs, an AI drug-discovery platform spun out of Google’s DeepMind in 2021, has announced the completion of a $5 billion Series C funding round. The investment was co-led by Greenoaks and Sequoia Capital, with participation from new investor Andreessen Horowitz and existing backers including Alphabet’s venture divisions GV and Capital G, Thrive Capital, MGX, Temasek, and the UK Sovereign AI Fund.

The new capital will be used to continue developing the company's AI drug-design engine and to expand into more types of medicine. Isomorphic Labs utilizes a suite of simulation AI models to explore various areas of chemistry to assist in drug design. Building on DeepMind’s breakthrough AlphaFold protein-folding AI model, the company stated that its program has yielded “positive preclinical results” in oncology and immunology. It has also struck deals with pharmaceutical giants such as Eli Lilly, Novartis, and Johnson & Johnson.

This raise follows a $2.1 billion Series B round completed in May. Notably, this is Isomorphic Labs' first fundraise since founder and CEO Demis Hassabis left his position as head of DeepMind to focus on drug discovery, moving into a chairman role at the lab instead. Hassabis and former DeepMind director John Jumper won the Nobel Prize in Chemistry for their work on AlphaFold, a system capable of predicting the 3D structure of proteins using only their genetic sequence.

Isomorphic Labs did not disclose its new valuation. Bloomberg reported earlier this month that the lab was seeking to raise funds at a valuation of approximately $50 billion.

Fuentes:TechCrunch

Investigación

First Ant Lingbo Embodied AI Challenge Concludes: Developers Showcase New Synergy Between General LLMs and Embodied Foundation Models

Tema · 蚂蚁灵波VLA模型Material
2026-11-17 19:18 GMT+8

On November 16, the first Embodied AI Challenge hosted by Ant Lingbo concluded in Zhangjiang, Shanghai. After online preliminaries, 15 teams advanced to a 48-hour offline final. All teams used the same open-source Vision-Language-Action (VLA) model, LingBot-VLA 2.0, along with the Songling COBOT MAGIC robot and unified on-site computing power, to complete three real-world physical tasks: vegetable and fruit placement, cosmetics organization, and dual-arm cleaning.

The competition results revealed a key trend in current embodied AI development: success no longer depends solely on who collects more data, but on how developers cleverly utilize the capabilities of existing foundation models. The winning team, TARS, achieved an 82-point score by prioritizing task difficulty and introducing large models to identify target objects and distractors. By occluding irrelevant areas, they allowed the robot to focus on the current task, achieving a 100% success rate in the vegetable and fruit placement task. Additionally, TARS compressed single-step training time from approximately 15 seconds to 3.4 seconds, gaining valuable time for rapid on-site iteration.

The third-place team, SST, demonstrated a "less but precise" strategy. They collected only 13 additional data points but integrated DeepSeek during inference to understand task instructions and used visual segmentation models to locate target objects before handing off actions to LingBot-VLA 2.0, ultimately scoring 65.87 points. This approach concretely illustrates the synergy between general-purpose large models and embodied foundation models: DeepSeek solves the understanding problem of "what to do," while LingBot-VLA 2.0 handles the execution problem of "how to do it." SST also froze some pre-trained parameters to avoid disrupting the model's original general capabilities with small amounts of supplementary data.

Other teams adopted different technical routes. Second-place EVA chose continuous data augmentation, converting failure cases into training data and improving their self-tested success rate from about 10% to 73.83% after seven iterations. The resource-limited FDU_SONIC team deeply analyzed the 647 trajectories provided by the organizers, adjusting data based on real-world test failures, and secured fifth place. These differences indicate that as foundation models become stronger, developers need less redundant construction of underlying capabilities and can focus on solving specific real-world problems.

However, the competition also exposed realistic challenges facing embodied AI. The team "Today Also Check Things Well," coming from the field of large model evaluation, found that although they validated their solutions in simulation environments, a discrepancy of about 5 centimeters between simulated and real tabletops caused their designed actions to fail on-site. This directly demonstrates that precise reproduction in digital environments cannot automatically bridge the distance between simulation and reality; robots still face uncertainties such as lighting, friction, and mechanical errors in the physical world.

Fuentes:GeekPark

Investigación

MongoDB Acquires Firebolt to Integrate Real-Time Analytics Engine for AI Agents

Material
Verificado 2026-11-17 22:34 GMT+8

US database company MongoDB announced on November 17, 2026, via an 8-K filing and a blog post that it has acquired Firebolt Analytics Inc. The acquisition aims to integrate Firebolt's high-performance real-time analytics engine directly into MongoDB's Intelligent Data Platform, enabling applications and AI agents to make decisions based on live operational data without building complex data pipelines.

MongoDB stated in its blog that traditional architectures separate operational databases from data warehouses, creating a "data-movement tax"—the cost and latency associated with copying, synchronizing, and maintaining additional systems. As AI agents require continuous, high-concurrency queries, this separation becomes a bottleneck. Firebolt's engine is designed for low latency and high concurrency, capable of processing fast-changing data on object storage.

Benjamin Wagner, CEO of Firebolt, noted that joining MongoDB means bringing their technology to where millions of developers already build applications. Mosha Pasumansky, CTO of Firebolt, emphasized that the engine retains the benefits of columnar storage and vectorized processing while supporting MongoDB's native flexible document model.

Currently, approximately 75% of Fortune 100 companies rely on MongoDB's platform. Following the acquisition, MongoDB plans to allow customers to enable real-time analytics for their data in the coming months. Existing Firebolt customers will continue to receive the same support and service. Specific financial terms of the transaction were not disclosed in the filings.

Fuentes:sec.gov

Investigación
Siguiente →
Iniciar sesión / Registrarse
LecturaInvestigaciónRadarMarco de inversión

Wilcox acknowledges that the value style has performed strongly in Japan for five years, and with the bank recovery largely complete, the next five years are unlikely to be equally smooth. However, he believes that in a market dominated by fast money and momentum, greed and fear tend to stretch further in both directions—removing valuation ceilings on the way up and breaking through valuation floors on the way down. For investors willing to step outside traditional style boxes, this volatility is creating opportunities across the spectrum, from overlooked steel and industrial companies to profitable AI hardware names.

The above views are attributed to Wilcox. Data is sourced from Bloomberg (as of October 29, 2026). Wilcox himself admits that there is no need to predict the specific form of AI in 2030 to see the asymmetry in current pricing, but he does not provide quantitative scenario analysis for a slowdown in hardware demand.

As a vendor methodology, this article does not provide independent verification data proving that following this process actually prevents over-commitment or reduces delivery risks. The value of Mark Deguara's argument lies in transforming liquid cooling readiness from a vague assumption of hall-level headroom into a verifiable, definite capacity for specific rack rows under normal operations and planned maintenance—enabling operators to identify constraints and price accordingly before quoting, rather than discovering bottlenecks only after delivery.

At the concurrent Ant Lingbo First Developer Conference, CEO Zhu Xing stated that embodied intelligence is still in its early stages. He noted that progress in general large models accelerates industry development, but robots must ultimately prove their adaptability, speed, and cost feasibility in real-world scenarios. He pointed out that the industry does not yet know which type of data scaling will lead to a capability leap, meaning embodied AI has not yet found its own "Scaling Law." The significance of this competition lies in enabling developers to experiment around the same foundation model, bringing back successful methods, lessons from failures, and real-world problems, which is valuable for driving the convergence of technical paradigms.