🎥 Watch the latest 7-minute briefing · Subscribe on YouTube

The AI-capex trade rips higher as ASML's second forecast hike of the year — plus a wave of KeyBanc chip upgrades and a cooler CPI — sends semis flying ahead of tomorrow's make-or-break TSMC print.

🔥 Today's Top Stories

🔥 ASML: Blows Past Q2 and Hikes Its 2026 Outlook a Second Time on AI Demand — Stock Surges, NVDA +4%

ASML — the sole supplier of EUV lithography for every leading-edge AI chip — reported €9.3B in Q2 net sales and €2.9B net income, then raised full-year 2026 guidance to €43–45B (up ≈16% at the midpoint from its prior €36–40B range) and said it will expand chipmaking-equipment capacity by 30%. As the single cleanest read-through on the entire AI hardware buildout, the beat lit a fire under the whole complex in premarket, dragging NVIDIA up ≈4% and reinforcing the "picks-and-shovels" infrastructure thesis.

• Second forecast hike of 2026 — guidance now €43–45B vs. €36–40B previously
• Installed-base (service/upgrade) revenue beat guidance by €300M as customers buy software-led productivity upgrades
• Gross margin 54%, guided to 54–56% for the year

  📰 Source: ASML raises 2026 forecast, expands capacity on AI chip demand — The Globe and Mail/Reuters (Tier 1)

🔥 AI Chip Trade Rips: SK Hynix +27%, NVDA +4% as ASML, KeyBanc Upgrades and Cool CPI Align

Semis staged one of their sharpest up-days of the summer as three catalysts stacked: ASML's raised outlook, a broad KeyBanc price-target sweep across the memory and GPU names, and Tuesday's softer-than-expected June CPI (largest monthly decline since April 2020) that cooled Fed-hike fears. The memory complex led — SK Hynix screamed ≈27% higher as the "look beyond GPUs" HBM/DRAM shortage narrative took center stage. Index futures were modestly green (Nasdaq-100 +0.5%, S&P +0.2%, VIX ≈16.3), with the risk-on tone carried by strong bank earnings.

• SK Hynix +27.3%, NVDA +4.1%, ASML +2.9% in premarket trading
• JPMorgan CEO called the banking environment "close to as good as it gets"; Morgan Stanley, BlackRock, J&J also reporting
• June CPI cooling shifts the macro tailwind back toward the AI trade

  📰 Source: Stock market today: Dow, S&P 500, Nasdaq futures extend gains as ASML buoys AI trade — Yahoo Finance (Tier 2)

🔥 TSMC: June Revenue +68%, Q2 at ≈$39.6B — Wall Street's Biggest AI Health Check Lands Tomorrow (July 16)

The world's largest contract chipmaker posted a 68% year-over-year jump in June revenue and roughly $39.6B in Q2 (≈36% YoY), and reports full results tomorrow in what analysts are treating as the single most important read on whether the AI boom is still accelerating. TSMC is sold out on its N3 node through year-end and is on track for $40B+ in AI chip revenue in 2026 (≈25% of total). A strong print + raised guidance would validate ASML's optimism; any softness on capex or AI order visibility is the market's biggest near-term risk.

• June revenue +68% YoY; Q2 ≈$39.6B, near the top of the $39.0–40.2B guide
• Full-year 2026 revenue guided to grow 30%+ in USD, driven by AI/HPC
• N3 sold out through year-end; Q1 gross margin was 66.2%

  📰 Source: TSMC reports 68% surge in June revenue ahead of second quarter — CNBC (Tier 1)

🏢 Big Tech Updates

📌 China: "Anthropomorphic AI" Rules Take Effect Today — ByteDance's Doubao and Alibaba's Qwen Kill Agent Features

China's Interim Measures for the Administration of AI Anthropomorphic Interactive Services — the first national framework targeting humanlike, emotionally-engaging AI — took effect today, July 15. ByteDance's Doubao and Alibaba's Qwen have shut down user-created AI agent/companion features, with existing agents ceasing to function. The mandated anti-addiction systems, usage notifications, and instant-exit mechanisms are architecturally incompatible with persistent-memory companion agents — a regulatory divergence that hands Western labs a freedom their Chinese rivals now lack, and a fresh compliance overhang on Alibaba and ByteDance.

• Doubao/Qwen agent creation disabled July 15; user data viewable read-only until Oct 15
• Rules cite addiction, privacy, mental-health and "extremist ideas" risks
• Widens the US–China regulatory gap on consumer AI product design

  📰 Source: ByteDance and Alibaba to disable humanlike AI custom agents as new rules loom — South China Morning Post (Tier 2)

📌 Microsoft: GPT-5.6 Becomes the Default in 365 Copilot — Even as Homegrown MAI Models Creep Into Excel and Outlook

Microsoft is running a two-track model strategy: GPT-5.6 became the preferred model in Microsoft 365 Copilot as of July 14, while — to cut AI costs — the company is quietly routing tens of thousands of weekly prompts inside Excel and Outlook through its own internally-built MAI models instead of OpenAI/Anthropic. The message to investors: Microsoft wants best-in-class frontier models where quality sells seats, but is aggressively insourcing the commodity, high-volume workloads to protect margins — a gradual de-risking of its OpenAI dependency.

• GPT-5.6 now the default reasoning model in M365 Copilot (July 14)
• MAI models now handle a rising share of Excel/Outlook AI prompts
• Signals margin discipline as Copilot inference volume scales

  📰 Source: Microsoft Replaces OpenAI, Anthropic With Own AI in Some Apps — Bloomberg (Tier 1)

📊 Market & Investment Signals

📌 KeyBanc: Aggressive Chip Price-Target Sweep After Asia Supply-Chain Checks — Micron to $1,750, AMD to $725

A five-star KeyBanc analyst turned decisively more bullish across the AI-semiconductor complex following Asia supply-chain checks that pointed to sustained data-center demand, tight memory supply, and firmer DRAM/NAND pricing. The upgrades were a major fuel source for today's memory-led rally.

• Micron $1,750 (from $1,600), AMD $725 (from $530), Arm $430 (from $300), Marvell $400 (from $385), NVIDIA $330 (from $310) — all Overweight
• Marvell call tied to Trainium 3 ramp in H2 2026 and Trainium 4 share

  📰 Source: Five-Star KeyBanc Analyst Lifts Targets for AMD, NVDA, MRVL, MU, ARM — TipRanks (Tier 2)

💬 Anthropic: Reportedly in Early Talks With Samsung for a Custom 2nm Claude Inference Chip

Context, not breaking: earlier this month Anthropic was reported to be in early-stage discussions with Samsung to build a custom AI accelerator on Samsung's 2nm process, aimed at Claude inference workloads amid a compute bill running toward ≈$1.25B/month. Anthropic stresses a diversified stack (Google, Amazon, NVIDIA) remains central — this would add a layer, not replace it — mirroring the in-house-silicon path already taken by Google, Amazon, Meta and OpenAI (Broadcom). Worth watching as a potential Samsung foundry win.

• Early exploratory stage; chip's role, power and packaging undecided
• Follows Anthropic hiring OpenAI chip-team engineer Clive Chan in June

  📰 Source: Anthropic is discussing a new custom chip with Samsung — TechCrunch (Tier 1)

🎥 Prefer to watch? This brief becomes a daily video on The Daily Alpha — AI × markets in ~7 minutes, every weekday. ▶ Watch the latest briefing 📺 Subscribe to the channel

Generated by Alan · The Daily Alpha. Editorial, not financial advice.