LesenForschungRadarAnlageframework
Anmelden / Registrieren
AlleTechnikProdukteInvestierenBörseSonstiges

Lesen

2026-11-1298 Beiträge

Aktualisieren

AI Startup millionways Launches Thorsten-4 Model Trained on Psychological Data, Raises $6M

Thema · millionwaysKurzfassung
2026-11-12 22:00 GMT+8

New York-based AI startup millionways launched its new AI model, Thorsten-4, on Thursday, which claims to offer insight into human intent and motivations. The company also told Semafor it has raised $6 million in US funding, with investors including football stars Jason Kelce and DK Metcalf.

The Thorsten-4 model was born out of a research project in Germany in which dozens of psychologists interviewed more than 7,000 people about their personal lives and aspirations. That data — which the psychologists labeled by hand to capture what people’s words revealed about their motives and thinking — was used to train the model.

Millionways cofounders Martin Cordsmeier and Max Weidemann argue that understanding is missing from regular AI models, which are trained on internet text. They see the model as particularly useful in sales and executive hiring, where a search firm could prompt it to evaluate an interview transcript to gauge how a candidate will behave under pressure. “We can add a level of human understanding that the language models cannot get to by themselves,” Weidemann said, adding that the model should be “a decision-making support tool rather than the ultimate decision-maker.”

Quellen:semafor.com

Forschung

Norway's Sovereign Wealth Fund Discloses $870.1 Billion US Stock Portfolio Heavy on AI Infrastructure

Thema · 挪威主权财富基金美股持仓Wesentlich
Verifiziert 2026-11-12 22:54 GMT+8

Norway's sovereign wealth fund, managed by Norges Bank, has filed its amended Form 13F-HR/A with the U.S. Securities and Exchange Commission (SEC) for the quarter ended September 30, 2025. The filing discloses approximately $870.1 billion in reportable US equity securities. It is important to note that this figure represents only the market value of specific stocks held in US markets as of the quarter-end snapshot; it does not reflect the fund's total Assets Under Management (AUM), which remains undisclosed in this form.

The portfolio reveals substantial exposure to the artificial intelligence sector. Key holdings include significant positions in Alphabet Inc. (approximately $35.9 billion across Class A and C shares), Amazon.com Inc. (approximately $28.2 billion), Apple Inc. (approximately $47.3 billion), Broadcom Inc. (approximately $21.9 billion), and Advanced Micro Devices Inc. (approximately $3.6 billion). Other notable tech and AI-related holdings include Intel Corp., Marvell Technology Inc., and Cloudflare Inc.

13F filings are quarterly snapshots required from large institutional investors, reflecting holdings at quarter-end rather than real-time trading activity or complete asset portfolios (excluding shorts, detailed derivatives, and non-US assets). Despite the lag, Norges Bank's disclosure of such a massive portfolio provides concrete quantitative evidence of how top-tier sovereign capital is positioned within the AI infrastructure ecosystem.

Quellen:sec.gov

Forschung

AI Agent Management Platform Kindo Raises $30M After Reporting 800% YoY Revenue Growth

Thema · Kindo AI代理管理Wesentlich
2026-11-12 22:00 GMT+8

AI agent management platform Kindo announced on November 12, 2026, that it has raised $30 million in a funding round co-led by Crosspoint Capital and Forgepoint Capital, bringing its total capital raised to $57.6 million. According to Tech Funding News, Kindo reported approximately 800% year-on-year revenue growth in 2026. The new capital will be used to expand its go-to-market organization and accelerate development of its enterprise AI Control Center and Deep Hat models and add-ons.

As businesses deploy AI agents across more departments, controlling what these systems can access, change, and spend is becoming a growing challenge. Kindo aims to provide a unified view of agent activity, data access, costs, and outcomes. Its platform includes gateways to govern and audit calls made by agents to connected data sources, tools, and approved AI models, along with identity controls to extend existing organizational permissions. Additionally, Kindo offers a purpose-built agent harness that allows teams to directly intervene or take over an agent when something goes wrong.

Beyond security and compliance, the cost of running AI agents is also a major concern for businesses. Kindo's AI Control Center gives finance teams visibility into token consumption, agent runs, and their outcomes, supporting cost tracking by employee, agent, department, and project, as well as setting budgets and hard limits. Its Deep Hat add-ons are designed to reduce these expenses further, for example by filtering repetitive events, adapting smaller models to specific tasks, and converting repeatable workflow steps into code to eliminate model costs.

Quellen:Tech Funding News

Forschung

Nvidia CEO Jensen Huang Resigns; CFO Colette Kress Named Interim CEO

Strukturell
2026-11-12 22:10 GMT+8

Nvidia confirmed on Thursday that Jensen Huang has resigned from his position as president and chief executive officer. The decision comes just days after Huang admitted to a romantic affair with an alleged Chinese spy, sparking a U.S. national security crisis. According to regulatory filings, Nvidia's board accepted Huang's resignation on Wednesday.

In a transitional move, Collete Kress, Nvidia's current chief financial officer and executive vice president, will take over as interim CEO until the company finds a permanent successor. Kress will continue to serve as the company’s CFO. Additionally, the board appointed Harvey Jones, a managing partner at Square Wave Ventures and former CEO of Synopsys, as the company’s first-ever chair of the board of directors. Jones has been on Nvidia’s board since 1993.

Following the announcement, Nvidia’s shares fell more than 11% in premarket trading on Thursday. Since Friday, the company’s stock has dropped dramatically, resulting in a market cap loss of more than $300 billion.

“Jensen’s contributions to NVIDIA are profound. NVIDIA is one of the most consequential technology companies in the world. The Board and the company thank Jensen for his immeasurable contributions to the company and the industry,” the company said in a statement.

Quellen:TechCrunch

Forschung

AI Startup Qvyon Launches with $8M Funding and Physical AI Validation Platform Mimic Core

Thema · QvyonKurzfassung
2026-11-12 22:00 GMT+8

US-based AI infrastructure startup Qvyon Technologies Inc. announced on November 12, 2026, that it has secured $8 million in total funding and officially launched its core product, Mimic Core. The round was led by MaC Venture Capital, with participation from BDC Capital, BKR Capital, and Spotlight VP. According to SiliconANGLE, Qvyon aims to provide pre-deployment safety validation services for autonomous vehicles and robots.

Mimic Core is a software-as-a-service platform that transforms real-world camera, lidar, and radar sensor data into simulated environments. By reconstructing 3D structures and time-sequence snapshots of scenes, the platform allows engineers to generate various extreme or dangerous conditions—such as a car suddenly changing lanes during a rainstorm or a blocked sensor array—in virtual settings to test AI model responses. This approach avoids the high cost and risk of repeatedly collecting field data in real scenarios.

Frantz Saintellemy, co-founder and CEO of Qvyon, stated that his team previously experienced the difficulty of proving system readiness while building AI for vehicles. He emphasized that companies need to test what happens in their own vehicles and factories using their own recordings. Currently, Mimic Core is available for a 14-day guided evaluation via Qvyon Cloud and also supports self-hosted solutions. The company plans to release subsequent products, including Mimic Validation OS and Mimic Learn, over the next two years to further help customers document test results and optimize AI performance.

Quellen:SiliconANGLE

Forschung

Korean AI Startup Enhance Launches MOPai with PwC to Automate Salesforce Marketing Setup via Computer Use Agents

Kurzfassung
2026-11-12 14:14 GMT+8

Korean enterprise AI agent startup Enhance has announced a collaboration with global consulting firm PricewaterhouseCoopers (PwC) to launch "MOPai," an AI-driven marketing operation platform designed to automate complex workflows in enterprise marketing.

A core feature of MOPai is its "AI-native CRM." It utilizes Enhance's Computer Use Agent (CUA) technology, allowing the AI to read screens and manipulate software interfaces like a human. In practice, the AI analyzes market and customer data to plan marketing campaigns, then directly logs into Customer Relationship Management systems such as Salesforce to automatically input sending conditions and campaign parameters. Human employees are only required to approve the final output.

The platform also includes features for Generative Engine Optimization (GEO), helping brands improve their visibility and citations in AI search services like ChatGPT or Perplexity. PwC contributes industry consulting expertise, while Enhance provides ontology and AI agent technologies. The two companies plan to expand their cooperation to overseas market analysis, managed services, and AI commerce in the future.

Quellen:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Forschung

Why Nvidia’s Role as the 'AI Central Bank' Relies on Jensen Huang’s Personal Credit

Strukturell
2026-11-12 21:09 GMT+8

Over the past three years, Nvidia has provided approximately $300 billion in financial support to customers, alongside investments in startups. Manuel Pinto, Head of Research at XTB Spain, describes this role as a "Nvidia put": the market does not believe Nvidia will prop up its stock price, but rather that it may use its balance sheet and negotiating power to support strategic projects when financing becomes scarce. This expectation has permeated the capital structure of AI infrastructure—OpenAI secured a lease guarantee of up to $105 billion for its Ohio data center project; CoreWeave accepted both a $2 billion investment from Nvidia and a stake exceeding 12%; and six financial institutions partnered with Nvidia to build a compute financing platform aimed at raising over $500 billion in third-party capital. Jensen Huang himself stated on X that Nvidia retains the option to backstop up to $125 billion of these deals.

Bruno Schneller, Managing Partner at Erlen Capital Management, argues that while these arrangements are grounded in Nvidia’s balance sheet and technological position, Jensen Huang’s personal credibility has been instrumental in forging some of the industry’s most ambitious partnerships. He suggests that while existing contractual obligations would not vanish with a leadership change, the critical question is whether a successor would possess the same appetite for equity investments or financing guarantees. Counterparties may demand greater transparency, tighter terms, and more formal assurances under new leadership. A more conservative approach to capital allocation could be positive for Nvidia’s balance sheet itself, but negative for parts of the AI ecosystem that have become dependent on its financial support.

Pinto adds that if Nvidia’s backing becomes less predictable, investors and creditors may demand higher risk compensation, which could translate into more expensive financing, slower project progress, and increased pressure on leveraged operators. The market has already begun pricing this in: Nvidia’s stock fell approximately 8% in Thursday’s premarket trading, while CoreWeave dropped about 7%, despite having surged 25% the previous day following strong earnings. Analysts interpret this synchronized decline as a reassessment of financing certainty, though it does not independently prove causality.

The applicability of this argument rests on structural drivers—hyperscaler capital expenditure, model training, inference, and accelerated computing—which will not disappear due to leadership changes. Large technology companies with strong balance sheets are better positioned to absorb such adjustments. The real pressure falls on mid-tier operators who rely on Nvidia’s credit endorsement to access low-cost capital but lack investment-grade ratings. Schneller proposes a practical test: whether existing projects continue to execute on schedule and whether new financing and capacity agreements can be concluded on comparable terms under different leadership. In this sense, Huang’s potential departure serves not just as a test for Nvidia’s leadership, but as a stress test for the entire financing model underpinning the AI investment cycle.

Quellen:invezz.com

Forschung

UAE's Largest Bank First Abu Dhabi Plans $4.1 Billion Rights Issue Backed by Sovereign Fund Mubadala for Postwar Growth

Thema · 第一阿布扎比银行配股融资Wesentlich
2026-11-12 21:19 GMT+8

First Abu Dhabi Bank (FAB), the United Arab Emirates' largest lender, has announced plans to raise 15 billion dirhams ($4.1 billion) through a rights issue to fund its "next phase of growth." The bank told analysts it sees higher organic growth opportunities rather than acquisitions ahead, positioning itself for a postwar investment boom in the country.

Abu Dhabi sovereign wealth fund Mubadala, which owns a nearly 40% stake in FAB, is backing the deal. This follows other banks in the emirate raising money through equity sales, including Abu Dhabi Commercial Bank's 6 billion dirham raise last year and Abu Dhabi Islamic Bank's ongoing 1.75 billion dirham offering.

Banks in the region have been tapping equity markets more frequently as loans grow faster than new deposits. They also face higher provisioning and impairments due to war-related risks. FAB's strategic pivot signals a preference for strengthening the balance sheet to support domestic demand over external mergers and acquisitions.

Quellen:semafor.com

Forschung

AI Optical Design Startup Opdo Raises $2.3M Pre-Seed Led by Angular Ventures

Thema · Opdo AI光学设计Kurzfassung
2026-11-12 21:08 GMT+8

Los Angeles and Berlin-based deeptech startup Opdo announced it has raised $2.3 million in pre-seed funding, led by Angular Ventures with participation from Inflection and Tiny.vc. The capital will support engineering team expansion, tooling, and initial steps toward in-house production.

Opdo's core product consists of AI agents embedded within existing optical design software such as Zemax and CODE V, rather than replacing these traditional tools. According to co-founder and CEO Stephan Kuehr, this integration has reduced customers' product development and manufacturing cycles from months to weeks. Kuehr previously founded manufacturing automation company 3YOURMIND and worked with CTO Yousef Arzhangnia at xolo, a maker of optical element production machines.

Despite the funding, Opdo currently has only five employees and has not named any specific customers. An article on their site mentions collaborative work with individuals from the University of North Carolina, Google, and Amazon. The niche market remains small; one study estimates the global optical design software market at $237 million in 2025, projected to reach $398 million by 2031.

Quellen:Tech Funding News

Forschung

Yakov Filippenko, Founder of AI Flexible Staffing Platform Intch: How App Store Removal Destroys Subscription Revenue and Why a Fallback Must Be Built in Advance

Thema · Intch下架事件Wesentlich
2026-11-12 20:00 GMT+8

Yakov Filippenko, founder and CEO of the AI flexible staffing platform Intch, described how an App Store removal event impacted the revenue structure of a subscription-based business within weeks. The removal first cut off the download entry point for new users; then, paying customers flooded customer service asking if the disappearance of the app meant the company was a scam; subsequently, existing users replacing their phones could not reinstall the app, leading to a continuous loss of the customer base. The financial impact was more direct: Apple immediately froze funds from the previous three months and shut down all subscription renewals. Filippenko pointed out that for a company relying on subscription revenue, this means monthly recurring revenue could drop to zero.

He cited data from Apple's 2025 Transparency Report: throughout the year, 166,899 apps were removed, and only 423 were restored after appeals. Intch itself experienced a 30-day appeal window, ultimately receiving a generic rejection reply that provided no information. Filippenko believes that appeal results are unpredictable, but businesses must continue to operate during the waiting period, meaning the appeal process itself cannot serve as a contingency plan.

His response was to decouple from the App Store two years before the crisis occurred: he built a web version and designated it as the primary payment channel. Filippenko stated that this choice also freed them from Apple's commission cuts and enabled an iteration speed of ten releases per day. Pavel Durov, founder of Telegram, has also promoted web access to bypass store restrictions, but Filippenko emphasized that the key is not post-removal remediation, but rather testing whether alternative channels truly work while the app is still downloadable.

After the removal event, every key metric at Intch fully recovered. Based on this, Filippenko posed a specific question to investors: when evaluating the growth of a subscription-based company, what proportion of revenue depends on continued access to a single platform? If the platform decides it can interrupt this revenue stream, has this exposure been reflected in valuation discussions?

It should be noted that this argument is based on the self-reported experience of one company. Filippenko did not provide industry benchmark data to prove that web alternatives restore performance for all App Store-dependent businesses, nor did he discuss under what scale the costs of building and maintaining independent payment channels become economically reasonable. His core assertion—that platform dependency is infrastructure risk and fallbacks must be built before removal—is a testable operational judgment, not a universally validated general rule.

Quellen:Crunchbase News

Forschung

Quadric CMO Says RISC-V for AI Acceleration Is a Bad Idea; Breker CEO Argues Control Is the Correct Positioning

Wesentlich
2026-11-12 16:01 GMT+8

Steve Roddy, chief marketing officer at AI chip company Quadric, made a sharp judgment in a Semiconductor Engineering roundtable discussion: using RISC-V for AI acceleration is a "completely bad idea." His argument is not based on RISC-V's inherent lack of capability, but on the cascading consequences of architectural choices—once you create a partition between a programmable processor and an external accelerator, you simultaneously create three problems: data movement, compiler complexity, and interconnect verification.

Roddy's mechanistic reasoning is specific. CPUs excel at pointer chasing and unpredictable thread switching—with 40 threads running, you have no idea what will execute in the next clock cycle. But AI inference does not follow this workload pattern: you know what needs to be done millions of cycles in advance. Therefore, most CPU hardware logic is useless for inference. More critically, while DSPs are closer to what is needed, they are vector machines (vector-by-scalar or vector-by-vector), whereas the most challenging operations in AI are matrix-by-matrix. Thus, his conclusion is that if you start from a CPU to build an accelerator, you are essentially using the wrong starting point to solve a problem that shouldn't exist, then spending billions of dollars to make up for that initial error.

However, Dave Kelf, CEO of verification company Breker, provided counter-evidence from actual projects. He observed that large AI accelerators—whether SIMD or SIMT architectures—are traditionally controlled by complex state machines, which he describes as "hopeless." Four companies he has worked with (both large and small) are replacing state machines with RISC-V processors to drive their accelerators. The reasons are practical: RISC-V allows adding custom instructions for efficient load/store operations to reduce partition latency, and integrating vector or matrix multiplication instructions to handle residual computation when a parallel lane in a SIMT array diverges.

Kelf cited a specific case: a customer had a SIMT array where occasional divergence occurred in some parallel lanes during massive parallel math operations. When divergence happened, they used the RISC-V control processor to handle the residual computation for that lane, while the rest of the channels continued running on the accelerator. This was not heavy inference on the RISC-V—it was just "enough math processing to keep the whole machine running." He acknowledged this required significant engineering effort, but the results were "quite successful."

The consensus of the roundtable discussion thus became clear: RISC-V’s value in AI systems is not to perform inference calculations itself, but to serve as an orchestration layer—using custom instructions to manage data flow between accelerators, reduce partition latency, and handle edge cases. Roddy also conceded that "you can get value out of tweaking and tuning a processor for the control function," though he did not believe it should be used for heavy inference.

But the discussion revealed a frequently overlooked layer: correct instructions do not guarantee a working chip. Ewald Liess of SignatureIP pointed out that when multiple RISC-V cores coexist with an accelerator in the same system, cache-coherency corner cases and traffic bursts can lead to interconnect deadlocks. An instruction may execute perfectly at the ISA level, but if the required data burst is unavailable, the chip will still fail. Francisca Tan of Siemens EDA added the dimension of observability—tracing custom instructions, system-level performance monitoring, and embedded analytics are indispensable during silicon debugging.

Quellen:semiengineering.com

Forschung

US cybersecurity firm Cloudflare researches MOLE post-quantum anonymous credential mechanism for AI agent traffic management

Kurzfassung
2026-11-12 08:00 GMT+8

US-based cybersecurity and cloud computing company Cloudflare recently disclosed that it is researching a new post-quantum secure anonymous credential mechanism called MOLE (Moderation Of unLinkable Endorsements). The research aims to address the threat of future quantum computers with cryptographic attack capabilities potentially breaking traditional encryption technologies, which could allow attackers to forge tokens and bypass security controls such as rate limiting.

The existing Privacy Pass technology allows users to prove they have passed verification while protecting their privacy, but websites can still identify the source of the token issuance. Cloudflare's MOLE attempts to further hide the user's endorsement source: the verification service can confirm that the endorsement comes from a trusted institution without identifying exactly which party provided the endorsement, subsequently issuing a credential for the user to access the website.

Technically, Cloudflare is researching the combination of multivariate cryptography and VOLE-in-the-Head (VOLEitH) zero-knowledge proof technologies to build an anti-quantum anonymous credential system suitable for MOLE. Beyond basic security protection, Cloudflare has also listed AI agent traffic management as a key future application direction for MOLE, hoping to use it to reduce the situation where massive automated AI requests occupy website resources and affect the experience of ordinary users.

However, the technology is currently in the research and development stage. Preliminary tests show that although some solutions can complete a single operation within milliseconds, post-quantum cryptographic technologies generate larger signature and verification data, with some operations requiring the transmission of several KB of data. If applied to actual web services in the future, besides verification speed, the communication burden caused by frequent verification will need to be carefully considered.

Quellen:iThome

Forschung
Weiter →
Anmelden / Registrieren
LesenForschungRadarAnlageframework

The practical implication for chip architects from this discussion is that RISC-V’s customization capabilities have clear applicability boundaries in the AI context. It is suitable for orchestration, control, and handling edge cases; it is not suitable for being a matrix inference engine. But even within its "suitable" range, between correct instructions and a usable system lie three verification hurdles: cache coherency, interconnect deadlocks, and performance observability.