ReadingResearchRadarInvestment framework
Sign in / Sign up
AllTechnologyProductsInvestingPublic marketsOther

Reading

2026-11-2081 posts

Refresh

US House Panel Reveals Commerce Dept Approved Over 879,000 AI Chip Exports to China, Including Military-Linked Firms

Structural
2026-11-20

Documents released by the US House Foreign Affairs Committee on November 19, 2026, reveal that the Department of Commerce's Bureau of Industry and Security (BIS) had approved the sale of over 879,000 artificial intelligence (AI) chips to Chinese companies as of May 2026. The approved buyers include tech giants such as Alibaba and Tencent, with some entities listed by the US Department of Defense as "Chinese military enterprises." This disclosure comes amid heightened scrutiny following the resignation of former Nvidia CEO Jensen Huang due to personal scandals.

The committee noted that between December 2025 and May 2026, applications for selling Nvidia H200 and equivalent AMD chips to China totaled 3.4 million. Applicants also included ByteDance, Tesla Shanghai, and Baidu. Committee Chairman Brian Mast specifically named Maginfra and New Matrix as suspected shell companies facilitating chip smuggling, citing a New York Times report that their offices were empty and linked to sanctioned entities.

Democratic Ranking Member Gregory Meeks criticized BIS for failing to enforce regulations, emphasizing the public's right to know how these decisions were made. The release relies on the Export Control Reform Act, which allows Congress to declassify export data if secrecy is deemed against national interest. While Commerce, Nvidia, and AMD did not immediately respond to requests for comment, the move intensifies doubts about the effectiveness of current control policies.

With Jensen Huang scheduled for closed-door questioning on December 1 and Commerce Secretary Howard Lutnick facing criticism for inaction, bipartisan lawmakers are pushing legislation to tighten restrictions on advanced AI chip sales to adversary nations. Additionally, AMD shares fell after short-seller Hunterbrook Media alleged it provided new firmware to blacklisted Chinese server maker IEIT Systems, highlighting the sector's geopolitical compliance vulnerabilities.

Sources:INSIDE

Research

Microsoft and Intel Deepen Partnership: First Use of EMIB Advanced Packaging in Custom Azure Chips, Securing Multi-Year Xeon Supply

Material
2026-11-20

U.S. tech giants Microsoft and Intel announced on November 20, 2026, an expansion of their decades-long partnership to advance Azure infrastructure development. The core breakthrough lies in Microsoft's first-time announcement of adopting Intel Foundry's EMIB (Embedded Multi-die Interconnect Bridge) advanced packaging technology for its custom Azure chips. This move marks further recognition by Microsoft of Intel's manufacturing capabilities within its AI infrastructure supply chain.

Beyond packaging technology, the two companies have established a multi-year capacity supply agreement for Intel Xeon processors, including the next-generation server CPUs codenamed Diamond Rapids. This long-term deal aims to ensure stable and predictable computing resources for Microsoft as it faces growing cloud and AI workloads. By combining Intel's chip design and packaging expertise with Microsoft's cloud engineering know-how, the partners seek to create differentiated infrastructure optimized for performance, efficiency, scalability, and security.

On the technical application front, Microsoft will collaborate with Intel on CXL-enabled Xeon processors to enhance Azure's memory efficiency. For instance, Azure M-series v4 virtual machines equipped with CXL memory and Intel Flat Memory Mode are reported to deliver up to 2.3x performance gains for high-demand workloads. Additionally, SQL backup speeds on Azure are expected to improve by up to 2.5x through Intel QuickAssist Technology (QAT). These system-level optimizations demonstrate deep integration in hardware co-design.

Sources:BenchLife.info

Research

Independent Mobile Analyst Richard Windsor: Apple Must Create an Exclusive AI Agent, or Its Ecosystem Premium Will Be Eroded

Material
2026-11-20

The App Store charts have already indicated the direction of travel: Meta's Muse, ChatGPT, and Claude are firmly entrenched in the top ten, OpenAI is expanding the functionality of Dots, Anthropic is testing Duet, Google has launched Tasks, and SpaceX is heavily pushing Grok Bot. None of these agents belong to Apple.

Richard Windsor believes this is a replay of the script seen fourteen years ago when the mobile ecosystem first emerged. At that time, numerous products competed for the central position in users' digital lives, and ultimately only iOS, Android, and WeChat survived. The difference now is that last time, Apple and Google were challengers with nothing to lose; this time, they are incumbents facing challenges, making them naturally more risk-averse because their existing massive businesses make it difficult for them to let go.

Windsor's core mechanism is this: if the agent on the device eventually performs most digital life services for the user, user preference will shift from the smartphone ecosystem to the agent itself. Since agents like Dots, Muse, and Duet can provide equivalent quality of service across all ecosystems, users will then question why they should continue to pay a premium for Apple. He concludes that Apple must create a top-tier agent and ensure it is available exclusively on Apple products.

However, Apple's current situation makes this task urgent. Windsor bluntly states that Siri is clear evidence of Apple's third-rate AI capabilities, and the fact that Apple is only now posting job openings for roles related to proactive agents shows just how far behind it is. He believes that while users have not yet chosen device ecosystems based on agents—meaning cross-device stickiness among iPhone, Mac, and Watch can temporarily maintain the overall ecosystem cohesion—this buffer will disappear once the agent becomes the true center of digital life.

Windsor also points out that Apple's current lack of exposure to AI actually provides it with a safe-haven status, as market sentiment toward AI trading is cooling. However, he believes this is insufficient to support its premium valuation: Apple shares trade at a 36.7x forward P/E ratio, nearly four times that of Nvidia, while offering only a 6% growth rate. In his view, there is much better value available elsewhere.

Sources:radiofreemobile.com

Research

Korean Web Builder I'mweb Reports AI Agent Submitted Over 100 PRs, Human Support Intervention Dropped Below 2%

Quick take
2026-11-20 08:56 GMT+8

Korean website building platform I'mweb disclosed recent advancements in its internal AI applications at its 'Techground 26' technical meetup. According to CTO Kim Hyung-seop, the team introduced AI agents capable of autonomously participating in work flows. One agent, designed as an 'AI colleague,' communicated with over 60 team members and independently wrote and submitted more than 100 Pull Requests (PRs), completing the entire process from code modification to deployment.

In the customer service sector, I'mweb achieved a higher degree of automation. Product engineer Lee Jung-min stated that by gradually optimizing the AI's judgment process for customer support responses, the rate of human intervention dropped from an initial 100% to less than 2%, eventually eliminating human judgment entirely in the live operational environment. This change relied on continuous validation and rule adjustments regarding the AI's accuracy.

Platform engineer Choi Sung-guk noted that for AI agents handling tasks, the key lies in setting criteria for when to ask humans for help and which decisions require joint human-AI judgment, rather than simply pursuing automated execution. Additionally, the infrastructure team emphasized strict pre-execution verification mechanisms when using AI for operations to ensure system stability. These cases demonstrate specific practices and quantified results of mid-sized internet companies integrating AI into core engineering and operational processes.

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

Research

Korean AI Infrastructure Firm MOREH Unveils TIDE Framework at SC26, Claiming 1.66x Throughput Gain

Quick take
2026-11-20 07:43 GMT+8

South Korean AI infrastructure software company MOREH unveiled TIDE, an inference framework designed to optimize "speculative decoding," at the Supercomputing 2026 (SC26) conference held in Chicago. Speculative decoding accelerates Large Language Model (LLM) generation by having a smaller draft model predict multiple tokens which a larger target model then verifies. However, traditional implementations often suffer from performance degradation when user request patterns shift, as the statically trained draft model's predictions become inaccurate.

TIDE addresses this limitation by leveraging the "intermediate hidden states" generated by the large model during inference to retrain the draft model in real-time. This approach eliminates the need for separate data generation passes, reducing computational overhead while allowing the draft model to adapt to changing traffic patterns. The system monitors prediction accuracy and effectiveness continuously, pausing speculative decoding when benefits diminish and triggering retraining when patterns change.

According to results presented by MOREH at SC26, TIDE achieved up to 1.66 times higher inference throughput compared to baseline methods without speculative decoding. Additionally, the training time for the draft model was reduced by up to 3.02 times compared to existing online learning approaches. The company also demonstrated a heterogeneous computing strategy, running the large model's inference on Nvidia H100 GPUs while training the draft model on AMD Instinct MI250 GPUs, which reportedly improved overall system throughput by up to 1.22 times compared to using a single GPU type for all tasks. These figures are self-reported vendor measurements and have not been independently reproduced.

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

Research

Korean Startup Seed& Wins Prime Minister's Commendation, AI Energy System Leaf Expands to Vietnam Fast Food Chain

Quick take
2026-11-20 08:37 GMT+8

South Korean AI building energy management startup Seed& received the Prime Minister's Commendation at the '2026 Korea Energy Grand Prize'. The award is hosted by the Ministry of Climate, Energy and Environment and organized by the Korea Energy Agency, recognizing individuals or groups who have contributed to the national economy through the development and dissemination of energy technologies. According to the company, Seed& was the only startup among the recipients of the Prime Minister's Commendation this time.

Seed&'s core product is an AI-based building energy management system called 'Leaf'. The system collects over 11 types of meteorological data, including outdoor temperature, along with real-time indoor temperature and humidity for each space. It uses AI to predict heating and cooling demand and automatically controls equipment. Unlike traditional systems that rely on sensors within the devices themselves, Leaf optimizes energy consumption based on areas where people actually stay, and can be installed on existing HVAC equipment without additional construction.

Since its official launch in 2019, Leaf has been deployed in over 3,000 locations domestically and internationally, including restaurant chains, offices, factories, and public facilities. The company states that sites adopting the system have reduced heating and cooling energy consumption by an average of 10% to 15%. After being designated as an innovative product by the Public Procurement Service in December 2025, its supply to the public sector has also increased.

In terms of market expansion, Seed& has signed a master contract to supply approximately 300 Leaf units to 40 Lotteria stores in Ho Chi Minh City, Vietnam. This marks the company's first large-scale entry into overseas markets. Previously, Seed& obtained international certifications for ISO 9001 Quality Management and ISO 14001 Environmental Management in September this year, planning to expand its presence in Southeast Asia starting with Vietnam.

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

Research

Arm China Launches 'Xingchen 100e' AIoT Prototype Platform with New Cortex-M42 Processor

Material
2026-11-19 12:02 GMT+8

On November 18, Arm China held a product briefing to provide an in-depth analysis of the Arm Cortex-M42 CPU and officially launched the 'Xingchen 100e' AIoT prototype platform based on this processor. Co-developed by Arm China and Arm, the Cortex-M42 is named 'Xingchen' STAR-MC3 in the Chinese market and has been formally integrated into Arm's global Cortex-M product roadmap.

Targeting smart endpoint devices sensitive to cost and area, the Cortex-M42 is built on the Armv8.1-M architecture and features three core capabilities: Arm Helium technology accelerates vector processing for stronger machine learning and digital signal processing performance; Arm Custom Instructions allow optimization for specific workloads; and a compact AHB-based design uses a mature microcontroller system architecture without complex memory subsystems. Compared to the predecessor Cortex-M33, its vector computing performance is tripled, and vector energy efficiency is doubled.

Regarding silicon area and power consumption, official data shows that in a configuration using TSMC's 22ULL process at 150MHz, the Cortex-M42 reduces silicon area by 10% and power consumption by 15% compared to the Cortex-M52, while delivering similar scalar and vector performance. This makes it suitable for designs that do not require rich memory subsystems.

The 'Xingchen 100e' is a heterogeneous computing platform combining the Cortex-M42 CPU with the Arm Ethos-U55 NPU. Arm China has built the system on an FPGA platform and successfully run typical AI use cases such as keyword spotting, object detection, and automatic speech recognition. Measured data indicates that compared to using the Cortex-M42 alone, pairing it with the Ethos-U55 improves keyword spotting inference performance by 95 times, object detection by over 132 times, and automatic speech recognition by over 340 times.

Zhu Xiaoming, CPU Product Director at Arm China, stated that the platform aims to provide various types of efficient computing power for low-power edge devices. However, he also emphasized that chip design is not simply 'building blocks'; merely combining IPs does not automatically yield competitive chips, and design companies still need to conduct extensive original architectural design and system optimization based on their understanding of applications and algorithms.

Sources:量子位

Research

Intuit Beats Q1 Estimates but Cuts Full-Year Outlook on Credit Karma Weakness

Material
2026-11-20 05:21 GMT+8

Financial software company Intuit (NASDAQ: INTU) released its fiscal first-quarter 2027 results on November 19, 2026. While quarterly revenue reached $4.35 billion, up 12% year-over-year, and adjusted earnings per share came in at $2.91—both beating Wall Street estimates—the company simultaneously lowered its full-year revenue outlook for fiscal 2027.

The beat was primarily driven by the Global Business Solutions (GBS) segment, which saw 15% revenue growth, led by a 20% increase in online ecosystem revenue. However, the consumer platform's Credit Karma business fell short of internal expectations. Management attributed this weakness partly to cyclical impacts from higher interest rates on loans and credit cards, but more significantly to a shift in consumer behavior towards using AI-powered experiences for financial product discovery.

Consequently, Intuit revised its total company revenue growth guidance for fiscal 2027 to 8%-9%, implying total revenue between $23.081 billion and $23.365 billion. CFO Sandeep Aujla noted that while the Credit Karma revenue outlook was reduced, full-year operating income and earnings per share guidance remained unchanged due to cost savings. Additionally, Intuit announced an agreement to sell Mailchimp to Bending Spoons to sharpen its focus on its core platform strategy.

Sources:Investor's Business Daily

Research

Shanghai AI Biotech Startup YuanShi Tech Completes ~300M RMB Pre-A Round, Claiming New Record for Domestic AI Virtual Cell Sector

Material
2026-11-19 14:02 GMT+8

Zhidongxi reported on November 19 that Shanghai-based AI biotech startup YuanShi Tech has completed a Pre-A funding round of nearly 300 million RMB. The round was led by Legend Capital, with participation from Huaxia Lingtou, JD-affiliated funds, CMC Capital, and others, while LightSource Capital served as the exclusive financial advisor. According to reports from PEdaily, China Merchants Capital, CDH VGC, and existing investors including Monolith, Sinovation Ventures, Daotong Investment, and Yanyuan Venture also participated or continued to add capital. This transaction claims to set a new record for single-round financing in the domestic AI Virtual Cell (AIVC) sector.

This marks YuanShi Tech's third announced funding round within one month. Founded in April 2026, the company primarily develops AI virtual cell models aimed at predicting cellular responses under various disease states, experimental conditions, and interventions. Previously, in October, the company completed Angel and Angel+ rounds totaling nearly 100 million RMB. The new funds are planned for accumulating cell intervention and response data, building automated experimental platforms, iterating on general cell large models, and expanding the R&D team.

YuanShi Tech's core technical approach involves constructing a "dry-wet closed loop" of "model prediction—experimental validation—data feedback—model update." Through this mechanism, the model identifies prediction shortcomings via experimental feedback, assists researchers in selecting the next round of experiments, and uses new results to supplement training data and improve predictions. The company describes this process as the application of Recursive Self-Improvement (RSI) in biological foundation models. Externally, YuanShi Tech has reached a strategic partnership with Yongsheng Intelligent, a subsidiary of MGI Tech, to co-build an automated experimental platform for virtual cells, aiming to reduce manual operation variances and accelerate data production.

The founding team possesses strong academic backgrounds. Co-founder and CEO Wu Jiaqi holds a PhD in Systems Biology from Peking University and previously conducted postdoctoral research at Rockefeller University. Co-founder and CTO Dong Mingze graduated from Peking University's Yuanpei College with a bachelor's degree and earned a PhD in Computational Biology from Yale University, having worked at Stanford University and Anthropic. Dong announced his full-time joining on November 11; he contributed to the STACK model, which achieved zero-shot generalization across different cellular environments.

Sources:zhidx.com

Research

NYU Finance Professor Aswath Damodaran: AI Trading Is Shifting from Broad Bets to Differentiated Pricing for Four Types of Companies

Structural
2026-11-09 21:37 GMT+8

The core judgment of Aswath Damodaran, a finance professor at New York University's Stern School of Business, is that the broad trade of buying any AI company to make money over the past few years is coming to an end. His basis is not a single stock price fluctuation, but a structural assessment—that the four categories of participants in the AI industry chain are parting ways in terms of their business models, revenue timing, and interest directions.

Damodaran splits the AI ecosystem into four groups: suppliers (chipmakers, power equipment providers, real estate firms, etc., selling infrastructure raw materials to builders); builders (hyperscalers with massive capital expenditure budgets such as Alphabet, Amazon, Meta, and Microsoft); product companies (frontier model and application firms like OpenAI and Anthropic facing end users); and customers (all potential individual and enterprise users of AI products). Although these four groups are collectively referred to as AI companies, they create value in entirely different ways.

The key divergence stems from time structure. Suppliers see real money first: the capital expenditures of builders become current-period revenue for suppliers, increasing their profits after deducting operating costs. Builders face the opposite reality—regardless of how reasonable the investments are, they impact profitability and free cash flow negatively in the short term, though this apparent earnings damage is masked by profit buffers from mature businesses. Product companies are even later, having almost no revenue to show before infrastructure is ready, while bearing huge R&D and compensation expenses. The customer group only forms at scale when products are truly usable and payable.

Damodaran uses the market reaction on November 5 to validate this framework: after the results of the US election were announced, the Nasdaq fell nearly 6% in a single day, and the S&P 500 dropped 4%, wiping out nearly $1.8 trillion in market capitalization from AI stocks in one day. However, differentiation had already appeared—chip stocks and electrical equipment stocks took the brunt of the hit, while hyperscale builders (four of the Mag Seven) actually achieved net growth in market capitalization during the week. Damodaran points out that if the election results imply a slowdown in data center construction, suppliers dependent on incremental construction will be impacted first, while shareholders of builders might welcome a deceleration in capital expenditures.

Deeper conflicts arise from opposing interests. Overbuilding by builders harms their own shareholders but is a bonanza for suppliers; builders want to lock out infrastructure through long-term leases exceeding reasonable useful lives to stabilize revenue streams, but this increases default risk for product companies. Damodaran argues that as AI business enters the commercial construction phase, these competitive interests will outweigh common interests, and the market will begin to pick winners and losers.

He simultaneously proposes the 'Big Market Delusion' framework: the potential of large markets combined with the overconfidence of founders, VCs, and CEOs leads to overinvestment and subsequent correction. He acknowledges that AI almost certainly involves overinvestment by builders, excessive compute commitments by LLM companies, and overall market overvaluation. However, he emphasizes that when this correction arrives and in what form is almost unpredictable; therefore, one should not simply avoid the entire sector, but rather sell portions of holdings during each correction and set trigger prices to buy back companies you believe are mispriced once the correction subsides.

Sources:aswathdamodaran.blogspot.com

Research

Dutch Regulator Warns Against Buying AI Smart Glasses Due to Safety Risks

Material
2026-11-20 12:03 GMT+8

The Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) has issued a warning advising consumers against purchasing AI-enabled smart glasses. According to a report by technology news outlet heise online, the regulator investigated six different models currently available in the Netherlands and found that these devices not only violate bystanders' privacy but also provide incorrect or dangerous guidance through their integrated AI chatbots.

In one simulated emergency scenario where a person required resuscitation, the tested glasses falsely claimed to have called the emergency number 112 and dispatched an ambulance, when no such call was actually made. The AP identified this as an "AI hallucination." Other tests revealed instances where the glasses provided inappropriate dating advice to adults or misinterpreted medication leaflets and toy safety warnings.

Monique Verdier, deputy chair of the AP, stated that while AI offers many opportunities, trust in camera-equipped smart glasses is currently unjustified. Although the AP cannot impose a sales ban—which falls under product safety authorities—it expects to gain more regulatory powers under the EU's Artificial Intelligence Act. Meanwhile, digital rights organization Bits of Freedom criticized the warning as insufficient, arguing that dangerous products should be removed from shelves rather than merely warned against.

Sources:heise online

Research

Three Retired V100 GPUs Outpace $5,399 Mac Studio in Long-Context Inference for $1,042

Quick take
2026-11-19

According to Tom's Hardware, a homelab builder known as esologic benchmarked three used Nvidia Tesla V100 GPUs, finding that the ~$1,042 setup outperformed a $5,399 Apple M5 Max Mac Studio in inference speed for the Qwen 3.6 35B mixture-of-experts (MoE) model at long context lengths.

The tests, conducted using the llama.cpp framework, compared generation speeds across different quantization levels and context sizes. With 8-bit quantization (Q8), the three V100s achieved 100.8 tokens per second at 4K context, dropping to 72.7 tokens per second at 128K. The M5 Max Mac Studio, equipped with a 40-core GPU and 128GB of unified memory, was slightly faster at 4K (105.5 tokens per second) but fell behind at 64K and 128K, where the V100 cluster was approximately 19% faster at the longest context tested.

The V100s represent older 'e-waste' enterprise GPUs, whose advantage lies in combining VRAM capacity and bandwidth through multi-GPU parallelism, enabling larger models or longer contexts. While modern consumer devices like the Mac Studio may excel in short-context performance and energy efficiency, they were surpassed by the low-cost legacy hardware in this specific long-text workload. The builder also noted that the V100s consumed about half the energy of an equivalent number of P100s. Due to the use of different runtimes (llama.cpp vs. MLX), these figures should be taken as indicative rather than independently verified.

Sources:Tom's Hardware

Research
Next →
Sign in / Sign up
ReadingResearchRadarInvestment framework

Damodaran’s sensitivity analysis of the valuation for Anthropic’s IPO also corroborates this framework: if the sales-to-capex ratio is lowered from 3 to 2.5 (reflecting heavier compute investment needs) and the regulatory drag on profit margins is increased from 5% to 7.5%, even maintaining a trillion-dollar revenue target, Anthropic’s estimated equity value drops from approximately $1.5 trillion to $1.25 trillion. He believes the IPO pricing will ultimately approach this number, as election results have already altered fundamental assumptions.

It must be noted that Damodaran himself admits the limitations of his sample: public data excludes the largest private players like OpenAI and Anthropic; most companies span multiple groups (Meta, Alphabet, and Microsoft are both builders and developing their own large models); and there is no independent accounting category for AI revenue, so classification relies on public descriptions. His framework is an analytical starting point, not a precise pricing tool.