LesenForschungRadarAnlageframework
Anmelden / Registrieren
AlleTechnikProdukteInvestierenBörseSonstiges

Lesen

2026-11-1392 Beiträge

Aktualisieren

US Stock Market Macro Sensitivity Fracture: Independent Research Attributes Shift to Economic Regime Change, Not AI Concentration

Thema · 美股宏观敏感度断裂Wesentlich
2026-11-13 19:49 GMT+8

In a paper published in November 2026, independent researcher Mohammad Bagher Karimi examined a specific question: Has the sensitivity of the US stock market to real yields, breakeven inflation rates, oil, gold, and the dollar fundamentally changed as the S&P 500 has become increasingly concentrated in a few large technology firms? Using weekly data from 2010 to 2026 and an index reconstructed from public sources, he systematically tested this hypothesis.

Karimi identified three structural breaks, occurring respectively in early 2013, early 2020, and mid-2023. The confidence interval of the last break includes late 2022, but no break was estimated at the time of ChatGPT's release. This temporal distribution poses an initial challenge to the AI concentration hypothesis: if AI were the primary cause, the breaks should be more closely clustered around the release and diffusion of large language models.

More critical evidence comes from his precise decomposition of index beta coefficients. Karimi decomposed the change in the overall index's sensitivity to macro factors into two sources: the mechanical effect brought about by changes in index weights, and the changes in the betas of individual constituent stocks. The decomposition results showed that the contribution of weight changes was minimal. The vast majority of sensitivity changes were experienced jointly by nearly all companies, rather than being a phenomenon unique to top holdings. Specifically, after 2022, the entire market lost its positive exposure to breakeven inflation rates, which is a universal phenomenon rather than a characteristic of a few tech stocks.

The real-yield betas of large companies did rise relative to average stocks after 2022, but Karimi further tested the AI intensity indicators in 10-K filings prior to 2022 and found that they had no additional explanatory power after controlling for firm size. In other words, there is no observable link between whether a company heavily uses AI and changes in its macro sensitivity that goes beyond the effects of scale.

Karimi's conclusion is that the evidence points to a macroeconomic regime shift, rather than concentration or AI, as the main source of the change in sensitivity. This research has its boundaries: it relies on an index constructed by a single researcher and specific weekly estimation windows, so the exact timing of the three breaks may be influenced by the sample period and estimation methods. 'Regime shift' itself is a descriptive label—Karimi's evidence indicates that the change was not driven by concentration, but he did not define what specific macro mechanisms drove this shift. For readers, the value of this paper lies in using decomposition methods to rule out concentration as the primary explanation, thereby redirecting attention back to macroeconomic conditions themselves.

Quellen:capitalspectator.com

Forschung

Howden Survey: One in Three UK Workers Expect 'AI Dividend' Pay Raises from Productivity Gains

Thema · HowdenWesentlich
2026-11-13 19:05 GMT+8

A recent survey of over 2,000 employees by UK insurance group Howden has found that one in three British workers expect higher pay if their employer is able to achieve significant productivity gains or cost savings through the use of artificial intelligence (AI). Howden describes this expectation as an 'AI dividend'.

The study indicates that while UK workers are generally willing to use AI at work, they do so on the condition that its benefits are shared with them and not just retained by the company. Beyond higher pay, those impacted by AI also expect shorter working hours (16%), greater job security (15%), and bonuses (13%) if their company manages to unlock AI's benefits.

However, acceptance is accompanied by anxiety. Forty-nine percent of workers worry they could be expected to complete even more work because of time savings, and another 50% fear AI could soon be used to assess their performance. Additionally, 53% are worried they won't have time to learn new skills on top of existing workloads, and half fear AI could be introduced faster than they can understand it.

Cheryl Brennan, MD of Howden Employee Benefits & Wellbeing, stated that simply providing access to AI training isn't enough; workers need full support, including dedicated learning time. She emphasized that the workforce is central to the AI-productivity journey and that a misalignment in expectations must be addressed to improve the chances of success.

Quellen:TechRadar

Forschung

Global Study of 9,034 Buybacks Challenges Dividend Supremacy: How Tax Drag and Low Reinvestment Rates Alter Capital Allocation Judgments

Wesentlich
2026-11-13 16:05 GMT+8

"Dividends account for the lion's share of stock market returns"—this phrase is repeated so often in investment circles that it has become common sense. However, it rests on a fatal premise: the assumption that investors reinvest every penny they receive back into the same stock. In his 2014 Credit Suisse report, Michael Mauboussin pointed out that this assumption does not hold up in reality. Dividends are taxable events; investors receive amounts that have already been subject to tax deductions. More critically, historical data shows that private investors actually reinvest only about 10% of their dividend income into the relevant stocks. This means the narrative of "dividend-driven returns" is largely a product of the theoretical world rather than the actual experience of investors.

This does not mean buybacks are automatically superior to dividends. The global buyback study published by Manconi, Peyer, and Vermaelen in 2018—which examined 9,034 buyback announcements across 31 non-US markets between 1998 and 2010—offers a more precise conclusion: on average, buyback announcements by non-US companies are indeed followed by significant short-term and long-term excess returns. However, the authors simultaneously emphasized that "not all buybacks are created equal." Excess returns are significantly larger for small-cap, undervalued value stocks, and are more pronounced in markets with lower liquidity and poorer efficiency.

This brings us back to Warren Buffett’s judgment in 1984: when a company with an excellent business and comfortable financial position finds its shares trading far below intrinsic value, no alternative action can benefit shareholders as surely as repurchases. Decades later, the hard numbers from Manconi et al. provide empirical support for this intuition. However, the article also points out an important omission in Buffett’s statement: buybacks benefit shareholders who continue to hold. For investors looking to exit, buybacks may not be the most effective way to realize value.

The other side of buybacks also needs to be faced squarely. The US Securities and Exchange Commission (SEC) once found that insiders sold five times as much stock in the eight days following a buyback announcement as on ordinary days, and the number of insiders disposing of shares after repurchase announcements was twice the normal level. The Atlantic magazine’s August 2019 long-read article, "The Stock-Buyback Swindle," was based on such evidence, accusing buybacks of becoming a tool for CEOs to enrich themselves at the expense of other shareholders. SEC Commissioner Rob Jackson stated bluntly: Boards of directors that allow buybacks to occur without transparency regarding CEOs selling into them raise serious questions about the board’s leadership.

Therefore, the key issue is not whether buybacks are good or bad in general, but the specific conditions: Is the company truly undervalued? Does management have the sincerity to continue execution? Is the board transparent about insider transactions? For companies with reasonable or even high valuations, buybacks may simply be using shareholders' money to buy back their own stock at high prices. Conversely, for companies whose share prices remain below intrinsic value—particularly small-cap value stocks in the UK market that have been persistently undervalued—failing to buy back shares is the true capital allocation failure.

Quellen:Undervalued Shares

Forschung

AI Company Appier Doubles Q3 Operating Profit, Raises Full-Year Guidance for Second Consecutive Quarter

Wesentlich
2026-11-13 19:56 GMT+8

Appier, an AI-native company centered on Agentic AI-as-a-Service (AaaS), released its fiscal year 2026 third-quarter earnings report. Quarterly revenue reached JPY 13.9 billion, growing 21.8% year-on-year. Operating profit hit JPY 2.1 billion, a 104.7% increase from the same period last year, significantly exceeding previous Q3 guidance. Revenue, gross profit, gross margin, and operating profit all set historical highs. Based on strong core business growth and profitability, Appier raised its full-year 2026 guidance for the second consecutive quarter, increasing the annual operating profit target to JPY 5.9 billion and the projected cash dividend per share to JPY 2.50.

Core business revenue for the third quarter was JPY 13.4 billion, up 27.1% year-on-year, surpassing the original 25% growth target. The gross margin improved to 61.3%, higher than 60.1% in the second quarter, representing a 7.2 percentage point increase year-on-year. Appier attributed the margin expansion to continuous Agentic engineering improvements in R&D efficiency and differentiated algorithms enhancing customer return on investment. Regionally, the US and EMEA accounted for 21% of total revenue, with core business revenue growing 49% year-on-year. Northeast Asia, comprising 71% of total revenue, saw core business revenue grow 33% year-on-year, driven by strong performance in its Ad Cloud business.

In terms of operational efficiency, after fully adopting Agentic AI, quarterly gross profit per employee rose 45% year-on-year to JPY 11.34 million, a historical high. Excluding currency effects, average quarterly revenue per core client (ARPC) increased 10% year-on-year to JPY 5.72 million. The ratio of operating expenses to gross profit fell to 78%, a historical low, indicating that productivity gains from AI over the past three years are accumulating. Despite headwinds from exchange rates, R&D investment remained stable, with the company emphasizing that the improvement stems from operational efficiency rather than reduced investment scale.

Cash flow performance improved significantly, with Q3 core free cash flow reaching JPY 1.9 billion, six times the level of the same period last year (up 501.3% year-on-year). The core free cash flow margin was 13.5%. Core free cash flow for the first three quarters turned positive at JPY 1.25 billion. Appier expects full-year 2026 core free cash flow to turn positive for the first time, funding company growth and shareholder returns with its own capital.

Dr. Chih-Han Yu, CEO and Co-founder of Appier, stated that the standard for measuring AI should be the actual value it creates. He noted that as customers succeed, Appier succeeds, describing this as a self-reinforcing AI growth model that requires minimal capital input, which is why the company has consecutively raised its full-year profit guidance.

Quellen:INSIDE

Forschung

South Korean Live Commerce Tech Firm Mobidoo Announces 'Always On' Strategy, Claims Up to 80% Cost Reduction via AI

Thema · MobidooKurzfassung
2026-11-13 15:27 GMT+8

South Korean live commerce solution provider Mobidoo announced its new "Always On" strategy at its annual event "Source Day 2026" on November 12. CEO Lee Yun-hee stated that the company aims to transform traditional scheduled live selling into an always-available purchasing model by leveraging artificial intelligence and the creator economy.

Central to this strategy is the launch of "AI Live." Users simply input a product URL, and the AI automatically generates scripts and videos, acts as a virtual host, and answers customer questions during the broadcast via an AI chatbot. Lee Min-hyun, Head of Solutions Division at Mobidoo, noted that preparing a traditional live broadcast takes at least 10 days and costs approximately 3 million KRW (about $2,200 USD) on average, whereas AI Live can reduce production costs by up to 80%.

According to Mobidoo's self-reported data, the platform produced and aired approximately 1,000 AI live broadcasts over the past three months, averaging 11 per day, with 54 million cumulative views and 200,000 product clicks. The company also built internal AI infrastructure incorporating over 20 video and image generation models linked with text-to-speech systems, allowing automatic selection of the most suitable model for each task.

Beyond AI technology, Mobidoo emphasized expanding market reach through creators. The company has recruited over 3,500 micro and nano-creators and launched "Source Link," a platform for tracking sales performance and settlements. Mobidoo plans to introduce a self-service platform connecting brands and creators early next year, enabling creators to directly find brand products and obtain sales links.

Mobidoo disclosed that its sales grew by more than 25% year-over-year through the end of October this year, and it has achieved monthly break-even profitability. The company stated it will further increase investment in AI video and the creator sector next year to fully implement the "Always On" strategy.

Quellen:바이라인네트워크

Forschung

South Korean AI Firm Konan Technology Unveils Defense Intelligence Fusion Platform ARGO-X

Thema · Konan ARGO-XWesentlich
2026-11-13 12:49 GMT+8

South Korean artificial intelligence company Konan Technology unveiled its all-source information fusion analysis platform, "Konan ARGO-X," for the first time at the "5th Konan Technology Defense AI Summit 2026" held in Seoul on November 13. CEO Kim Young-seom stated that as unmanned systems and drones reshape the battlefield, speed of judgment has become key combat power, and technology must truly work in the field to help soldiers.

ARGO-X is a platform that uses AI to integrate information scattered across various sources and analyze relationships and context to support intelligence analysts' judgments. It connects assets such as Imagery Intelligence (IMINT), Signals Intelligence (SIGINT), Geospatial Intelligence (GEOINT), Open Source Intelligence (OSINT), and Human Intelligence (HUMINT). The goal is to identify conflicting evidence and information gaps, reducing the burden on analysts who manually search and compare vast amounts of data.

Chief Operating Officer Yang Seung-hyun outlined the company's dual-axis strategy of "Intelligence + Autonomy": one axis involves intelligence fusion and command decision support via ARGO-X, while the other includes unmanned autonomous device AI such as AI pilots, counter-drone AI, and space edge AI. According to the company, ARGO-X is currently developed based on open-source data and is being built through the South Korean Ministry of National Defense's rapid commercialization support project, with an official release planned for December.

Quellen:바이라인네트워크

Forschung

KubeVirt Graduates from CNCF, Unifying VMs and Containers on Kubernetes

Thema · KubeVirtWesentlich
2026-11-13 08:31 GMT+8

The open-source project KubeVirt announced its graduation from the Cloud Native Computing Foundation (CNCF) at KubeCon + CloudNativeCon North America, becoming one of the foundation's top-tier projects. This milestone signifies broad recognition of Kubernetes' ability to manage both traditional virtual machines (VMs) and modern containerized applications on a unified platform.

Founded by Red Hat in 2016, KubeVirt was designed to allow Kubernetes to orchestrate VMs similarly to how it orchestrates containers. As artificial intelligence (AI) drives demand for compute resources, enterprises face the challenge of maintaining legacy VM systems while deploying new AI infrastructure. Fabian Deutsch, Director of Engineering for OpenShift Virtualization and Modernization at Red Hat, stated that maintaining two separate systems was never the goal; the vision is to unify all compute workloads under a single control plane.

CNCF graduation indicates that the project has met strict standards for maturity, adoption, and governance. Contributors now include Nvidia, Microsoft, AMD, and Intel. Deutsch noted that AI is changing the value proposition of virtualization, with growing needs for passing GPUs, CPUs, and memory directly into VMs, and enhanced security through isolation models.

For enterprise users, KubeVirt's graduation reduces technology selection risk. It allows organizations to leverage native Kubernetes APIs to manage mixed workloads without completely re-architecting their data centers. This progress reflects the cloud-native ecosystem's evolution from simple container orchestration toward broader unified compute resource scheduling.

Quellen:SiliconANGLE

Forschung

TSMC Confirms Employee Data Leaked to External Site via Cloud Provider, States No Evidence of Internal System Breach

Thema · 台积电云端数据泄露Wesentlich
2026-11-13 09:04 GMT+8

On November 12, TSMC stated that personal information of some current and former employees was unauthorizedly posted on an external website. The company emphasized that there is currently no evidence showing its internal systems were breached, and it has notified affected individuals to provide necessary assistance.

According to INSIDE, TSMC explained that the leaked data matched records within a certain cloud service provider's system and was unrelated to the company's production or customer data. TSMC is cooperating with this cloud provider, investigative authorities, and a third-party cybersecurity firm to clarify the incident, though it did not disclose the supplier's name or the specific fields of the leaked information.

Prior reports indicated that hackers named TSMC in underground forums selling employee personal data. The seller claimed to have 6.9 GB of data covering approximately 46,000 employees, priced at 3 BTC, and threatened to release all data if no buyer was found within a week. These figures are solely the seller's claims and unconfirmed by TSMC; the actual number of affected individuals and scope of data remain pending further investigation.

The key aspect of this incident is that the data leak occurred not within TSMC's own systems but through outsourced cloud services. When enterprises entrust HR or employee data to third parties for storage, they effectively transfer part of their protective responsibility; if a supplier fails, it is still the enterprise's own employees who suffer.

Quellen:INSIDE

Forschung

Unit 42 Discloses Autonomous AI Cyberattacks on Semiconductor and Defense Targets Using Cairn Framework

Strukturell
2026-11-12 23:00 GMT+8

Threat intelligence firm Unit 42, part of Palo Alto Networks, published a report on November 12, 2026, detailing a cyber espionage campaign tracked as CL-STA-1238. The activity, conducted by Chinese-speaking attackers between August and September 2026, utilized the open-source AI penetration testing framework Cairn combined with Chinese large language models Zhipu GLM 5.2 and DeepSeek V4 Flash/Pro to autonomously execute 25 operations against at least 19 organizations.

Unlike previous reports focusing on Asian government targets, this campaign concentrated on advanced semiconductor manufacturing supply chains, U.S. defense policy institutions, and human rights and civil society organizations. Unit 42 assesses with moderate confidence that the targeting aligns directly with established Chinese state intelligence priorities.

Technical analysis reveals that attackers bypassed model safety guardrails by injecting a Chinese-language preamble into the system prompts, asserting that all tasks were authorized under security testing or CTF scenarios and instructing the model never to refuse. Each operation also generated an AUTHORIZATION.md file to simulate client-approved engagement. Neither Zhipu GLM nor DeepSeek refused these framings.

The AI agents independently planned and executed multi-stage tasks ranging from reconnaissance and asset mapping to privilege escalation and lateral movement. Individual operations lasted from 20 minutes to over 50 hours without attacker-imposed time limits. The report confirms unauthorized access or data retrieval at 10 targeted organizations, including the acquisition of dozens of valid credentials, exposure of over one million personal records, retrieval of military and government identities, and creation of persistent administrator accounts in production systems.

Despite the scale of automation, the vulnerabilities exploited were known issues such as authentication bypasses, insecure direct object references, weak passwords, and unpatched public CVEs. Unit 42 emphasizes that while autonomous AI increases the volume and coverage of exploitation, it does not increase complexity; traditional identity hygiene and software patching remain effective defenses.

Quellen:Unit 42

Forschung

AI Research Institute Epoch AI Estimates: Huawei’s Largest AI Data Center Has Only 7% of the Compute Power of SpaceX’s Colossus 2

Wesentlich
2026-11-13 00:03 GMT+8

China possesses abundant electricity, and Huawei is constructing AI data centers on a massive scale. However, an estimate released on November 12, 2026, by Epoch AI researchers Ben Cottier and Lynette Bye indicates that electricity is not the bottleneck—chips are. Using commercial high-resolution satellite imagery and modeling cooling equipment, they calculated the actual compute capacity of Huawei’s flagship AI data center in Horinger, Inner Mongolia, finding it to be approximately 81,000 H100 equivalents, merely 7% of the capacity of SpaceX’s Colossus 2 data center in Tennessee.

The methodology behind this estimate is noteworthy. The researchers did not rely on any compute figures published by Huawei. Instead, they started with satellite imagery: observing four buildings in the West and South zones equipped with dense cooling facilities (cooling towers and air-cooled chillers), they estimated the IT power capacity of these buildings at 286 megawatts. Buildings in the North zone had noticeably sparse cooling equipment and were excluded from the AI compute estimate. They then combined this with the sales timeline for Huawei Ascend chips, assuming a four-month lag from sale to deployment, to calculate that approximately 105,000 Ascend 910C chips and 8,000 910B chips were currently operational.

The key performance conversion comes from public disclosures: the Ascend 910C performs at approximately 0.76 H100 equivalents, while Nvidia’s B200 and B300 chips perform at about 2.5 H100 equivalents. This means that for every watt of power, US data centers can produce 2.5 times more AI compute than Chinese ones. This is the direct consequence of US export controls—not restricting China’s access to electricity, but restricting its access to high-performance chips.

Comparing Horinger with Colossus 2 makes the gap even more intuitive. Colossus 2 has approximately 14 times the compute capacity of Horinger. The researchers also made forward-looking projections: Horinger has at least three new buildings planned, one of which is already under construction and expected to come online by February 2028. By then, Horinger’s total compute could grow by over 70%, reaching approximately 141,000 H100 equivalents. However, Colossus 2 is expected to reach at least 1.8 million H100 equivalents during the same period, meaning Horinger’s share will not improve and may even shrink.

This estimate has specific implications for understanding China’s AI capabilities. The researchers note that data center compute is typically divided into multiple densely interconnected clusters, and total capacity cannot necessarily be merged to train a single model; however, large R&D experiments can still run even if the compute at Huawei Horinger is split into many clusters. Just 10% of Horinger’s current estimated compute (approximately 8,100 H100 equivalents) is sufficient to train a DeepSeek-V4-Pro-scale model within a month; if all compute were merged for a single training run, a GPT-6 Astra-scale model would take one year.

Of course, this estimate has clear boundaries. The 90% confidence interval ranges from 36,000 to 141,000 H100 equivalents, a span of nearly four times, reflecting uncertainties in assumptions such as chip model mix, deployment timing, and building usage. Colossus 2’s capacity itself is not an officially confirmed figure. The researchers also acknowledge that the overall frontier of Chinese AI data centers may not fall further behind—Huawei’s Wuhu campus is growing slightly faster than Horinger and is projected to reach 151,000 H100 equivalents by December 2027. But regarding single-facility comparisons, the Horinger case clearly demonstrates that, given sustained export controls, China’s AI compute constraints are unlikely to ease within the next two years.

Quellen:epochai.substack.com

Forschung

US Data Center Power Demand Outstrips Supply by 10x, Threatening AI Boom Halt by 2030

Thema · 美国数据中心电力瓶颈Wesentlich
2026-11-12 22:13 GMT+8

According to TechRadar, a new report by insurance broker Howden states that US data center power demand currently outstrips supply capacity by ten times. If new power sources are not brought online, this bottleneck could bring the AI boom to a grinding halt by 2030.

Data shows that since 2000, US data center power demand has increased by 19% year on year, while actual electricity grid expansion has only grown by 2% annually. In 2025, data centers accounted for 5% of total grid capacity, up from just 0.1% in 2000. To continue expansion without increasing grid usage, up to 58% of power demand will need to be sourced from alternatives.

Major AI companies are taking various measures to address this gap. Anthropic recently revealed it is investing $3.4 billion in onsite gas turbine generators to supply power for new sites. However, this makes the company heavily reliant on natural gas supplies, and the noise and emissions from these turbines have raised health concerns among local residents.

Other companies are turning to nuclear power. The Palisades Nuclear Plant in Michigan, Crane Clean Energy Center in Pennsylvania, and Duane Arnold Energy Center in Iowa are set to receive a total of $4.4 billion in federal grants to be recommissioned, with Microsoft and Google first in line to receive energy from these sites. Additionally, Amazon plans to purchase over 960 megawatts of energy from small modular reactors (SMRs) in Iowa at a cost likely exceeding $1 billion, though these sites are yet to be approved and may not be operational until the mid-2030s.

While alternative solutions are being pursued, grid expansion is typically funded through ratepayers, leading experts to predict that power costs in some states could rise by as much as 50% by 2030. Meanwhile, building new nuclear plants is expensive and time-consuming, requiring significant investment to meet safety standards, meaning continued AI infrastructure expansion still requires substantial capital and time buffers.

Quellen:TechRadar

Forschung

SMIC Q3 Revenue Rises 6.1% to $3.19 Billion, Gross Margin Hits 29.9%

Thema · 华为麒麟芯片Wesentlich
Verifiziert 2026-11-13 18:45 GMT+8

Semiconductor Manufacturing International Corporation (SMIC), a leading wafer foundry in mainland China, released its unaudited results for the three months ended September 30, 2026, on November 13, 2026. The company reported sales revenue of $3.19 billion for the third quarter, up 6.1% quarter-over-quarter and 33.9% year-over-year. The gross margin reached 29.9%, an increase of 4.6 percentage points from 25.3% in the second quarter, primarily driven by higher capacity utilization and changes in product mix.

In terms of capacity, SMIC's monthly production capacity increased from 1,096,500 units (equivalent to 8-inch standard logic) in the second quarter to 1,116,750 units in the third quarter. Total wafers sold reached 3,033,683, with a capacity utilization rate of 97.3%. Capital expenditure was $1.094 billion, lower than $1.836 billion in the previous quarter.

For the fourth quarter, management expects revenue to grow 4% to 6% quarter-over-quarter, with a gross margin guidance of 26% to 28%. The company stated it will continue to expand capacity based on customer demand and build a leading position through high-value-added product portfolios.

Quellen:static.cninfo.com.cn

Forschung
Weiter →
Anmelden / Registrieren
LesenForschungRadarAnlageframework

The author of this article expects that fewer investors will automatically favor dividends over buybacks when facing clearly undervalued companies in the future. He believes that since the United States—the world’s most mature capital market—has already taken the lead in shifting toward a more active buyback culture, boards in persistently undervalued markets like the UK will eventually follow suit. The real question is: Can today’s decision-makers learn to allocate capital more intelligently, or must shareholders replace them first?