Daily Highlights
- CoreWeave bond yield hits 15%, stock plunges 15% in a day
- Epic delays Disney shooter, pivots to AI-first development
- Anthropic pushes IPO to November at ~$2T valuation
- ByteDance Seed sets up standalone safety and alignment unit
- Qualcomm CEO says AI giants plan to enter smartphones by 2028
TODAY'S HEADLINES
[CoreWeave bond yield hits 15% as AI-infrastructure debt fears trigger a global chip selloff] On Oct 14, the yield on CoreWeave's 2032 bonds touched 15% intraday and closed at a record 14.5%, while the stock crashed 15% to $65.78, its worst day since February, drawing a one-word "yikes" from Elon Musk. CEO Michael Intrator told employees the market was being "irrational" and that contracted backlog fully covers debt obligations. The company has raised over $35 billion this year, and the bond turmoil is now driving a global reassessment of AI infrastructure financing. (Source: Bloomberg)
[Epic delays its Disney shooter and pivots to an 'AI-first' production model] On Oct 12, Bloomberg reported that Epic Games has delayed its shooter with Disney featuring Marvel and Star Wars characters - originally set for March 2027 with no new date - and told the team to go "AI first": no more hand-written code, with all art generated through full AI workflows as developers manage output from tools such as Anthropic's Claude. CEO Tim Sweeney confirmed the "huge pivot," saying the goal is to let every team build "MORE and BETTER stuff" with AI. (Source: Bloomberg)
NEW PRODUCTS & TECH
[Qualcomm CEO: multiple AI giants plan their own phones, aiming to run 100B-parameter models on device] Qualcomm CEO Cristiano Amon said in a Fireside Alpha interview that several leading AI companies are planning to enter the smartphone market, with one proposing to keep a 100-billion-parameter model running continuously on a phone by 2028 and already scouting hardware to support it. Amon did not name the company but said the moves show AI firms are no longer content to be behind-the-scenes model suppliers and aim for the front of the device. (Source: AIBase)
[Anthropic's first overseas data center heads to Osaka as NTT DATA talks enter final stage] Anthropic is in final-stage talks with NTT DATA over its first data center in Japan in Osaka, with a definitive agreement expected within the current fiscal year, Japanese media reported. The facility is planned at roughly 500 MW with investment in the trillions of yen, partially online as early as 2028 - extending the compute-hungry lab's footprint to the western Pacific for Asia-Pacific enterprise customers and model training. (Source: AIBase)
IPO & FUNDINGS
[Anthropic pushes IPO to November, targeting roughly $2 trillion valuation and up to $100B raise] According to Reuters and the Wall Street Journal, Anthropic has moved its IPO timeline from October to November so the prospectus can include a full third quarter; Reuters-reviewed filing documents show annualized revenue climbing from about $65 billion in July to over $100 billion. At a roughly $2 trillion valuation, the listing would surpass SpaceX and become the largest IPO in history. (Source: Reuters/WSJ)
[Tencent reportedly prices $5 billion bond with proceeds mainly for AI capex] Tencent has issued a $5 billion bond, with the company saying proceeds will go primarily toward AI capital expenditure, replenishing ammunition for model training and compute infrastructure, according to Taiwan's Economic Daily. The AI arms race among Chinese internet giants is spreading from equity funding into the bond market. (Source: Economic Daily)
OPEN SOURCE & RESEARCH
[Mistral Large 4 open weights on a countdown: Europe's 1T-parameter flagship lands end of October] France's Mistral AI confirmed that open weights for its trillion-parameter flagship Mistral Large 4 - about 1T total parameters, roughly 52B active per token, natively multimodal - will drop by the end of October, with Hugging Face showing an Oct 31 ETA. The model is currently served only through a guardrailed preview API while Mistral red-teams it with cybersecurity agencies and state authorities, positioning it as Europe's "third way" between US closed models and China's open-weight camp. (Source: Mistral blog/TechCrunch)
COMPANY & PEOPLE
[ByteDance Seed sets up standalone model safety and alignment unit, a first among Chinese frontier labs] According to Securities Times, ByteDance's Seed team announced in an internal email a new "Seed Model-Alignment & Preparedness" department led by Yan Lin, reporting directly to Seed head Wu Yonghui, with the former Responsible AI team and part of the Chat team moving in. The unit will focus on agent-model safety and guarding against frontier risks such as CBRN; ByteDance says it is the first Chinese frontier AI lab to make safety and alignment a standalone department, elevating it from a supporting function to core infrastructure. (Source: Securities Times)
POLICIES & DATA
[AI financing panic sweeps global markets: Philadelphia chip index down six straight sessions, KOSPI halts after 8% plunge] Squeezed by CoreWeave's bond turmoil and a hotter-than-expected September CPI, the Philadelphia Semiconductor Index fell 4.3% on Oct 14, its sixth straight decline wiping out about $1.6 trillion in market value; the selloff spread to Asia on Oct 15, with Korea's KOSPI falling more than 8% intraday and triggering a 20-minute trading halt, SK Hynix and Samsung both down over 9%, and the Nikkei closing down nearly 5%. UBS says the move looks like an expectation reset rather than a fundamental reversal, but market tolerance for AI-trade leverage is narrowing fast. (Source: Bloomberg/Cnyes)
Powered by AOYII.com
