Daily Highlights
- Anthropic's S-1 draft reveals $80.7B first-half net loss
- SpaceX posts first profit as public company, AI revenue up 830%
- DeepSeek fully open-sources its DeepInfer inference framework
- Amazon to invest 16B euros in German AI and data centers
- Qualcomm reportedly planning global layoffs
TODAY'S HEADLINES
[Anthropic's S-1 draft leaks: $80.7B first-half net loss and up to $750B in compute commitments] According to an S-1 draft reviewed by The Wall Street Journal, Anthropic posted $16.3 billion in first-half revenue but an $80.7 billion net loss, driven by accounting charges on convertible financing instruments and roughly $9 billion of infrastructure spending. The company has locked in over $600 billion of compute commitments, a figure that could reach $750 billion if some $150 billion of deals under negotiation land, and its IPO may slip past the US midterms in November. (Source: WSJ)
[SpaceX posts its first quarterly profit as a public company as AI revenue soars 830% to become its biggest business] After the close on Oct 20, SpaceX reported third-quarter revenue of $14.58 billion, up 161% and ahead of estimates, with net income of $4.4 billion - its first profitable quarter since going public in June. The "Super Intelligence" AI unit grew revenue 830% to $7.56 billion, crossed half of total revenue for the first time and turned a $431 million operating profit; Starlink reached 13.6 million subscribers. Musk also said SpaceX and Tesla have each formed special committees to evaluate a merger. (Source: Reuters)
NEW PRODUCTS & TECH
[Samsung is testing a robot that works 24 hours straight, eyeing a CES debut] Samsung Electronics' RX business office is validating a robot that can run around the clock at industrial sites in Seoul: it performs repetitive manufacturing and logistics tasks such as hauling cargo while continuously learning its surroundings and work patterns, with Samsung and partner staff monitoring it in three shifts. Samsung is studying whether to enter it at CES among several candidate products. (Source: Sina Finance)
[Waymo plans to grow its robotaxi fleet fivefold to 20,000 vehicles by end of next year] Waymo plans to expand its driverless fleet to five times its current size by the end of next year. Scaling the fleet is a critical step for robotaxi commercialization, and dispatching, remote assistance and safety systems will all face much larger-scale tests. (Source: ITHome)
[Oracle and AstraZeneca partner to recruit clinical trial patients with AI] The two companies signed a non-binding agreement to use AI and real-world data to identify patients eligible for clinical trials, speed up the generation of medical evidence and fold findings into routine care, starting with oncology, cardiovascular, respiratory, immunology and rare diseases, with shared governance and effectiveness metrics planned. (Source: Jiemian News)
IPO & FUNDINGS
[Amazon to invest over 16B euros in Germany in 2027, expanding data centers and AI infrastructure] On Oct 20, Amazon announced at the Berlin Summit on Investing in Germany that it will invest more than 16 billion euros (about $18 billion) in the country in 2027, covering data center and AI infrastructure expansion, high-speed satellite internet via Amazon Leo, logistics projects across Germany and a major Prime Video content slate. (Source: Sina Finance)
[Neolix closes nearly $400M Series E, the largest single autonomous-driving raise in China this year] Neolix has raised nearly $400 million in Series E funding, reportedly led by China's national AI industry investment fund, the largest single round in the country's autonomous-driving sector this year. The company builds unmanned delivery vehicles, and the proceeds are expected to fund scaled manufacturing and multi-scenario deployment. (Source: ITHome)
[Nvidia in talks to invest hundreds of millions in Australian AI cloud Firmus] According to Cailianshe, Australian AI cloud provider Firmus is seeking at least $2 billion in private funding after a failed IPO attempt, and Nvidia is in talks to put in several hundred million dollars - it had originally earmarked a substantial sum for the IPO. (Source: Cailianshe)
OPEN SOURCE & RESEARCH
[DeepSeek fully open-sources its DeepInfer inference framework, compatible with Nvidia and Ascend] DeepSeek announced on Oct 20 that it has fully open-sourced its internal inference framework DeepInfer on GitHub, alongside a production-grade token encoder-decoder component, deepseek-recipe, aligned exactly with its official API logic. DeepInfer bundles high-performance kernels with a deeply optimized inference engine, supporting distributed inference, speculative decoding, SWA Bounded Replay and structured output across both Nvidia and Huawei Ascend platforms; it already powers high-throughput, low-latency serving for DeepSeek V4.1 Flash. (Source: 36Kr/Jiemian News)
[AI agents can spontaneously develop a shared vocabulary without human language] A large-scale study by teams at the Vrije Universiteit Brussel AI Lab and the University of Namur shows that even with no human-taught language and no preset communication rules, AI agents interacting repeatedly will self-organize a shared, mutually understandable vocabulary, published in the latest issue of Computational Linguistics - fresh evidence for multi-agent communication and language-emergence research. (Source: Cailianshe)
COMPANY & PEOPLE
[Qualcomm reportedly planning global layoffs as CEO says expansion has changed talent needs] Qualcomm is reportedly preparing worldwide job cuts, with CEO Cristiano Amon saying the company's business expansion has changed its talent requirements. The chipmaker is in the middle of diversifying from handsets into PCs, automotive and edge AI, and its workforce structure is expected to shift with that strategy. (Source: ITHome)
[Jensen Huang: AI development cannot slam the brakes, but it must wear a seatbelt] Speaking in London, Nvidia CEO Jensen Huang opposed slowing AI development with abrupt halts or heavy-handed regulation, arguing instead for a "seatbelt" approach: smarter safety standards, monitoring mechanisms and incident-prevention measures. His remarks come as the White House "Super Intelligence" accord pushes industry self-policing and governments draft their own oversight frameworks. (Source: Cailianshe)
POLICIES & DATA
[UN launches new initiative on military AI, with first international forum on Thursday] The UN Institute for Disarmament Research will launch a new initiative on the military uses of artificial intelligence on Thursday, creating the first international forum for key players to tackle the issue. UNIDIR head Robin Geiss warned at a UN press briefing that "the margin for human judgment is increasingly narrow, including in the nuclear domain," and said global competition for AI leadership is a serious obstacle to guardrails. (Source: AP)
[IDC: agentic PCs are redefining the PC, shipping 20.8M units by 2030] IDC said on Oct 20 that agentic PCs are beginning to redefine competition in the PC market and projects 20.8 million units shipped by 2030, as AI shifts from add-on feature to core PC capability. (Source: ITHome)
[TrendForce: Q4 DRAM contract prices expected to rise another 20%] Market research firm TrendForce, citing multiple reports, expects DRAM contract prices to climb about 20% quarter over quarter in Q4 2026 - a marked slowdown from roughly 80% in Q2 and 40% in Q3, but the market remains in a tight-supply, strong-demand cycle lifted by AI. (Source: TMTPost)
Powered by AOYII.com
