〔Day Digest〕AI Demand Converts Into Revenue: Vertiv Posts $4.313 Billion and Moore Threads Revenue Jumps 1,773.92% — Cash Flow, Memory Inflation and Capex Durability Now Divide the Trade

0. Weekly Arc

Sourcing quality is uneven: India's securities-regulator story appears in three near-identical briefs, while the Legora and Anthropic items are headline-only Google News entries ^{[1][2][3][4][5][6][7]}. Against that backdrop, Vertiv beat revenue and adjusted-EPS expectations, while Moore Threads and Tianfu Communication posted rapid revenue growth and positive attributable net profit ^{[8][9][10][11]}. The counterweight is cash and capacity: Google and Microsoft cash flow, PC shipment contraction, DRAM/HBM pricing and the durability of AI capital spending remain the unresolved variables ^{[12][13][14][15]}.

1. Infrastructure Earnings

• **[NEW] Vertiv:** Q3 revenue was $4.313 billion versus $3.760 billion expected, adjusted EPS was $2.21 versus $1.83 expected, and Q4 revenue guidance was $4.313-$4.513 billion versus $4.308 billion expected ^[8].
• **[NEW] Moore Threads:** Q3 revenue reached RMB1.552 billion (+1,773.92% y/y); attributable net profit was RMB49.2604 million and net profit excluding non-recurring items was RMB20.0724 million, with both turning profitable year on year and quarter on quarter ^[9][10]. Across the first nine months, attributable net profit was RMB37.6973 million, but net profit excluding non-recurring items remained a RMB131 million loss; R&D expenditure was RMB1.340 billion, up 55.72% y/y ^[9][10]. Moore Threads attributed the acceleration to strong AI demand for high-quality compute and full-function GPUs, plus faster adoption of products such as MTTS5000 across cloud-service providers, token-service providers, AI startups, operators, research institutions and industry customers ^[10].
• **[NEW] Tianfu Communication:** Q3 revenue was RMB1.923 billion (+31.50% y/y), attributable net profit was RMB879 million (+55.30%), and net profit excluding non-recurring items was RMB879 million (+55.78%); nine-month revenue was RMB4.751 billion (+21.25%), attributable net profit was RMB2.083 billion (+42.18%), and adjusted net profit was RMB2.065 billion (+44.26%) ^[11]. Tianfu Communication attributed the growth to accelerating global AI activity and data-center construction, which sustained demand for high-speed optical devices ^[16]. The calculated Q3 attributable net margin rose to 45.72% from 38.71%, while receivables and inventory increased, operating-cash-flow growth trailed profit growth, and the company increased R&D spending, strategic stock-up and equipment prepayments ^[16].

2. Memory, Servers and Capex

• **[ESCALATED] Cash-flow test:** Microsoft and Google are due to report Wednesday after the close; the market appears to favor Microsoft, Google's 2026 and 2027 free cash flow is expected to be negative, and Microsoft's free cash flow is expected to decline in the fiscal year ending next June before returning to growth ^[12].
• **[ONGOING] Server-PC split:** Bernstein's Q3 preview, citing IDC and Gartner, says global Q3 PC shipments fell 20-22% y/y and 9-14% q/q, the worst third-quarter sequential performance in more than 20 years; Taiwanese ODM notebook shipments fell 24%, 38% and 20% y/y in July, August and September, with Q3 down 17% q/q against a pre-pandemic average of about +7% and down 27% y/y ^[14].
• **[ONGOING] Memory squeeze:** A Goldman Sachs report relays TrendForce's expectation that DRAM prices will keep rising through H1 2028 despite weak consumer-electronics demand, with long-term-agreement coverage potentially 60-70%; TrendForce forecasts 2027 blended HBM ASP up 122% y/y to $3.70 per Gb and global HBM revenue up more than 230% to above $200 billion ^[13].
• **[NEW] Capex debate:** Ling Chen, an investment manager at Harmony Huiyi No. 1 TMT Innovation Select Fund, expects AI to move over the next 6-12 months from thinking and reasoning to executing digital tasks, opening digital-employee and personal-assistant markets ^[15]. He views concern about a 2028 capex cliff as overstated, acknowledges AI has not yet achieved self-funded cash generation and needs financing, and expects application pricing to move from per-seat toward outcome-based models ^[15].

3. Enterprise and Consumer AI

• **[NEW] Boss Electric:** Q3 revenue was RMB2.289 billion and net profit was RMB260 million; AI kitchen appliances accounted for 27% of overall kitchen-appliance revenue, up from 24% in the first half, while the Shishen App and mini-program had served more than 9.2 million cumulative users ^[17]. Q3 average monthly users reached 670,000, up 63.4% y/y; the AI-dialogue user share reached 31%, about 6.6 times the start-of-year level, and monthly dialogue users rose 127% y/y ^[17].
• **[NEW] India:** India's securities-market regulator chair said the regulator is considering domestic AI models for securities trading ^[1][2][3].
• **[NEW] TokenRouters:** Xunce Technology launched a one-stop exchange and matching platform for real-time AI data and dedicated tokens, integrating datasets, models, applications, a scenario library and intelligent customer service, with the goal of delivering privately deployable domain-specific small models at lower cost and in shorter cycles ^[18].
• **[NEW] Legora (headline-only):** A Google News headline says Legora is building an agentic operating system for legal work; no operating details are provided in the supplied material ^[4].

4. Capital Formation and Physical AI

• **[NEW] Zhaofeng:** Controlled subsidiary Zhaojü Intelligent won the robot innovation service platform project at a bid price of RMB299.92 million, but no formal contract had been signed by the announcement date, leaving execution uncertain ^[19].
• **[NEW] Tianzhun Technology:** General Manager Cai Xiongfei said the company's Constellation Physical AI Tool Chain provides cross-chip edge deployment and model migration for embodied AI and mainstream open-source models including autonomous-driving BEV, end-to-end and VLA models, covering edge inference and task engines to compress typical 3-6 month deployment periods to less than one month ^[20].
• **[NEW] Unitree and Sequoia China:** The companies launched the Shanyu Venture Plan to invest in early-stage embodied intelligence, robotics applications and their upstream and downstream key technologies, along with other frontier hard-tech areas ^[21].
• **[NEW] Micro-nano Core:** The company completed more than RMB1 billion in Series C2 funding, following RMB1 billion in Series C1 in September, taking two-month funding above RMB2 billion and six-month funding to RMB3 billion ^[22]. Three first-generation chips are expected to enter customer testing in early 2027; the company says its SRAM compute-in-memory architecture delivered 4-6 times the compute density and 5-10 times the energy efficiency of a conventional von Neumann architecture in prior tape-out tests, but the public materials do not provide a complete comparison chip ^[22].
• **[NEW] Black Sesame Intelligence:** The company is repositioning from smart-driving chips to a broader embodied-intelligence and edge-AI inference platform, framing a strategic path from a roughly $10 billion smart-driving market through roughly $100 billion of embodied intelligence toward roughly $1 trillion of physical-AI edge inference ^[23].

5. Risk Signals and Falsifiers

• **[NEW] Anthropic (headline-only):** An NPR headline examines a plan for Anthropic to go public, a Wall Street Journal headline says Elizabeth Warren demands answers about existential risks, and a Hacker News headline says an unreleased Anthropic model chains flaws into working Chrome, Safari, Windows and Linux exploits ^[5][6][7].
• **[NEW] MIRI:** President and former Google engineer Nate Soares estimates AI has a greater-than-50% probability of causing human extinction in under a decade and says he stopped contributing to his U.S. 401(k) for ten years because of that risk ^[24]. Soares argues governments should treat superintelligent AI as at least a nuclear-weapons-level threat and calls for pausing AGI training in favor of narrower applications such as cancer research or safer autonomous driving ^[24].
• **Falsification test:** The near-term checks are the free-cash-flow disclosures in Microsoft and Google results, the server-versus-PC split in Intel and AMD results, conversion of Zhaofeng's project award into a signed contract, and customer testing of Micro-nano Core's chips in early 2027 ^{[19][12][14][22]}. The infrastructure-demand thesis would weaken if cash generation deteriorates as memory costs rise and PC shipments keep contracting; the physical-AI commercialization thesis would weaken if announced projects fail to become executed contracts or tested silicon ^[19][14][22].