AI & Software
AI & Software Insights
Research notes on earnings, market structure, and financial statements. Each piece starts with the conclusion, then lays out the evidence and implications.
2026-07-15
%22%2F%3E%0A%20%20%3Ccircle%20cx%3D%22220%22%20cy%3D%22150%22%20r%3D%22180%22%20fill%3D%22%2322d3ee%22%20fill-opacity%3D%220.12%22%2F%3E%0A%20%20%3Ccircle%20cx%3D%221010%22%20cy%3D%22130%22%20r%3D%22220%22%20fill%3D%22%2338bdf8%22%20fill-opacity%3D%220.08%22%2F%3E%0A%20%20%3Ccircle%20cx%3D%22920%22%20cy%3D%22520%22%20r%3D%22170%22%20fill%3D%22%2322d3ee%22%20fill-opacity%3D%220.06%22%2F%3E%0A%20%20%3Cpath%20d%3D%22M140%20430h920%22%20stroke%3D%22url(%23line)%22%20stroke-width%3D%2212%22%20stroke-linecap%3D%22round%22%2F%3E%0A%20%20%3Cpath%20d%3D%22M220%20360c120-150%20260-220%20380-220s260%2070%20380%20220%22%20fill%3D%22none%22%20stroke%3D%22%2367e8f9%22%20stroke-opacity%3D%220.5%22%20stroke-width%3D%2216%22%20stroke-linecap%3D%22round%22%20stroke-linejoin%3D%22round%22%2F%3E%0A%20%20%3Cpath%20d%3D%22M300%20398h600%22%20stroke%3D%22%2394a3b8%22%20stroke-opacity%3D%220.24%22%20stroke-width%3D%228%22%20stroke-linecap%3D%22round%22%2F%3E%0A%20%20%3Ctext%20x%3D%22220%22%20y%3D%22510%22%20fill%3D%22%23e2e8f0%22%20fill-opacity%3D%220.7%22%20font-family%3D%22Arial%2C%20Helvetica%2C%20sans-serif%22%20font-size%3D%2238%22%20letter-spacing%3D%228%22%3EINSIGHTS%3C%2Ftext%3E%0A%3C%2Fsvg%3E)
Australia's New AI Data-Center Rules Turn Power and Water Into the Next AI Capex Tax
Australia's July 15-16, 2026 policy shift says large AI data centers must secure their own power, cover their connection costs, manage peak-load behavior, and use water efficiently. The headline is about regulation, but the real message is capital discipline: AI infrastructure is no longer just a software or hardware spend, it is a grid, water, and permitting problem. That matters for Vertiv, Eaton, Equinix, Digital Realty, and every hyperscaler trying to scale demand without colliding with local utility politics.
China's 4.3% Q2 GDP Reopened the Beijing Stimulus Trade — And the Nvidia H200 Path With It
China's Q2 2026 GDP printed 4.3% year-over-year, the weakest since Q4 2022 and below Beijing's 4.5%-5% full-year target. Real estate investment fell 18%, fixed-asset investment dropped 5.7% in H1, but industrial output rose 5.3% and retail sales rebounded to +1% from a -0.6% reading in May. The combination forced Beijing into a third-quarter policy-easing stance, which markets priced through China tech, the yuan, and a quiet re-rating of Nvidia on H200 export hopes. Alibaba, Tencent, PDD Holdings, BYD, and the China-exposed cyclicals all move with this trade.
IBM's 25% Crash Made the Enterprise Software Tape Stop Trusting Guidance
IBM closed down 25.21% at $217.07 on July 14, 2026 on 8.6x normal volume, its worst single-session decline on record. The drop did not come with an official earnings warning; what changed is that the market stopped trusting the multi-year software-and-consulting bridge that large-cap enterprise vendors have been selling since the AI capex story began. The contagion test is now in Accenture, Cognizant, Capgemini, Palo Alto Networks, CrowdStrike, ServiceNow, and Salesforce.
JPMorgan Quietly Made AI Job Cuts a Reported Number — And the Labor Market Can No Longer Ignore It
On July 14, 2026, JPMorgan disclosed that it has roughly 1,000 active AI use cases and that 'discrete areas' saw headcount reduced 30%-40%, with most affected employees redeployed rather than terminated. Headcount held at 320,560 employees, almost flat quarter-over-quarter, while net income jumped 41% year-over-year to $21.2 billion and adjusted EPS beat by 5% at $6.14. The tape ignored the EPS beat and focused on the labor story because a 30%-40% headcount cut inside the largest US bank is no longer an HR footnote — it is a labor-market datapoint the Fed cannot avoid.
Oracle's 63% Slide From Its 52-Week High Is the First AI-Infrastructure 'Fallen-Angel Watch' the Bond Market Has to Price
Oracle is down roughly 13% over the last 10 sessions and ~63% from its 52-week high of $345.72, with the 52-week low of $127.60 set on July 14, 2026. After the June 11, 2026 8% tumble on a $20 billion capital-raise disclosure and the June 23, 2026 disclosure that Oracle shed ~21,000 roles over the past year, the equity is now pricing the risk that AI-infrastructure capex commitments are pushing the balance sheet toward fallen-angel territory. With debt-to-equity at 322.86%, the question for the bond market is no longer whether Oracle can fund the AI build, but whether the cost of capital will move before the next earnings print on September 7.
Advanced Packaging Deep Dive: How TSMC CoWoS, Samsung I-Cube, Intel EMIB, and Amkor Built the Single Most Binding Constraint of the AI Accelerator Cycle
Advanced packaging - specifically TSMC CoWoS (Chip-on-Wafer-on-Substrate) - is the binding constraint of the AI accelerator cycle, and 2026 CoWoS capacity is 100% allocated. TSMC is the dominant supplier with ~75% advanced packaging share, Samsung I-Cube is at ~10%, Intel EMIB/Foveros is at ~10%, and the OSATs (Amkor, ASE Technology, JCET) are at ~5%. This is a full-stack deep-dive into advanced packaging: 2.5D CoWoS-S/CoWoS-L/CoWoS-R, 3D Foveros/EMIB/TMV, hybrid bonding, the CoWoS-L expansion, the substrate supply chain (Ibiden, Shinko Electric), the top experts, the customer base (Nvidia, AMD, Broadcom, Apple), and the read-through for the 2026-2028 AI capex stack.
Atomic Layer Deposition (ALD) Deep Dive: How Applied Materials, Lam Research, Tokyo Electron, and ASM International Built the Most Concentrated Thin-Film Equipment Oligopoly in the AI Chip Era
Atomic Layer Deposition (ALD) is the most concentrated thin-film equipment oligopoly in the semiconductor industry. Applied Materials leads with roughly 50 percent ALD share, Lam Research is at 20 percent, Tokyo Electron is at 15 percent, and ASM International is at 10 percent. This is a full-stack deep-dive into ALD: thermal ALD vs plasma-enhanced ALD (PEALD) vs spatial ALD, the precursor chemistry, the binding layer counts (3nm logic needs 50 plus ALD layers per chip, HBM3E 12-Hi needs 30 plus), the top experts, the customer base (TSMC, Samsung, Intel, SK hynix), the 20B+ USD ALD capex 2025-2027, and the read-through for the AI capex stack.
Alibaba and Apple Just Made Qwen the First Chinese AI Model Inside the iPhone Stack
Alibaba's U.S.-listed shares rose 4% in pre-market trading on July 16, 2026 after confirming its Qwen AI model will be integrated into Apple Intelligence across iOS, iPadOS, macOS, and visionOS for users in China. The Cyberspace Administration of China included Apple AI services on its approved-providers list alongside Huawei. The deal is the first time a Chinese AI model is inside the iPhone stack, and the read-through is direct for Apple, Alibaba, Huawei, Tencent, Baidu, and the entire cross-border AI-supply-chain complex.
Amazon Just Lost a 19-Year AWS S-Team Veteran Right When the AI Capex Cycle Is at Its Peak
Dave Brown, an Amazon Web Services SVP who helped assemble the original EC2 service in South Africa in the 2000s and oversaw AWS's compute and machine learning units including Bedrock and SageMaker, plans to leave the company at the end of July 2026 after nearly 19 years. He will be replaced by Dave Treadwell from the e-commerce division. Brown's exit from Jassy's 28-person S-team comes as AWS recorded 28% revenue growth in Q1 2026 and as Amazon raised at least $25B in a bond sale to fund its AI capex - the worst possible timing for a leadership transition.
Anthropic Is Now the Cleanest Test of Whether a $965B Private AI Can Survive Wall Street's Discipline
CNBC reported on July 15, 2026 that Anthropic is scheduling investor meetings with Goldman Sachs, Morgan Stanley, and JPMorgan Chase ahead of a potential IPO as soon as October, with a confidential SEC filing already in place. The company closed a $65B funding round at a $965B valuation in May, putting it above OpenAI's $852B private mark. The Anthropic listing would be the first mega-private-AI IPO test, and the read-through is direct for the entire cap-markets fee pool that the Street has been modeling higher - and for the AI multiple that the public tape has been paying for every other mega-cap.
ArF Photoresist Deep Dive: How JSR, Shin-Etsu Chemical, Tokyo Ohka Kogyo, and Sumitomo Chemical Run the Most Concentrated Materials Chokepoint in the AI Chip Supply Chain
ArF (Argon Fluoride) photoresist is the binding materials chokepoint for the entire leading-edge semiconductor industry. The 4 Japanese suppliers - JSR, Shin-Etsu Chemical, Tokyo Ohka Kogyo, Sumitomo Chemical - control ~90% of the global ArF resist market, and the EUV resist oligopoly is even tighter. This is a full-stack deep-dive into ArF photoresist: polymer chemistry (polyhydroxystyrene, acrylate, COMA), photoacid generator (PAG) chemistry, line-edge roughness (LER), the EUV resist roadmap, the 12-inch resist supply chain, the top experts, the customer base (TSMC, Samsung, Intel, SK hynix), the recent CXMT/YMTC Korean/Chinese challenger progress, and the read-through for the 2026-2028 AI capex stack.
Buffett Just Told CNBC He Personally Initiated Alphabet - And 'It's Tough to Find Values When Everybody Is Preferring Gambling'
CNBC reported on July 15, 2026 that Berkshire Hathaway Chairman Warren Buffett, in an interview with Becky Quick, said he personally initiated Berkshire's Alphabet position - not incoming CEO Greg Abel - and warned that 'it's tough to find values when everybody is preferring gambling.' Buffett's first public confirmation of his direct role in the Alphabet investment is the cleanest single signal yet that Berkshire is treating Alphabet as a long-term core holding, not a tactical trade. The read-through is direct for Alphabet, Microsoft, Apple, Meta, Amazon, Nvidia, the entire Big Tech cohort, and the value-vs-growth rotation thesis.
DRAM Deep Dive: How Samsung Electronics, SK hynix, and Micron Run the $118B Memory Backbone of the AI Era - and Why Conventional DRAM Is Now a Three-Player Oligopoly
DRAM is the $118B/year memory backbone of the AI era, and the supply is structurally a 3-player oligopoly: Samsung Electronics (~42%), SK hynix (~33%), and Micron (~25%). HBM is a sub-segment of DRAM, but conventional DDR4/DDR5/LPDDR5X/LPDDR6 is the bulk of the market and the cleanest single read on the AI server capex cycle. This is a full-stack deep-dive into DRAM: cell architecture (1x/1y/1z/1a/1b/1c/1d/1g nm), capacitor technology (dielectric + electrode), node roadmap, the DDR5 ramp, the LPDDR5X/LPDDR6 mobile ramp, the customer base (Apple, Dell, HP, the hyperscalers), the top experts, the expansion capex, the 2026-2028 cycle, and the read-through for the AI capex stack.
EUV Lithography Deep Dive: Why ASML and the ~5,000 Engineers at Veldhoven Run the Binding Constraint for Every AI Chip on Earth - and Why High-NA EUV Is the 2026-2028 Race
EUV lithography is the binding constraint of the entire leading-edge semiconductor industry, and the supply chain is a monopoly: ASML is the only merchant EUV supplier, with ~5,000 engineers at Veldhoven, ~$30B in 2025 EUV revenue, and a 100% market share. This is a full-stack deep-dive into EUV: light source (laser-produced tin plasma), collector mirror (Zeiss), reticle stage, wafer stage, pellicle, photoresist, masks, the High-NA EUV roadmap (ASML's TWINSCAN EXE:5000 in 2026-2027), the customer base (TSMC, Samsung, Intel, SK hynix), the ~$200M per system price tag, and the read-through for the 2026-2028 AI capex stack.
Frontier Model Tier Deep Dive: How OpenAI, Anthropic, Alphabet, Meta, xAI, and DeepSeek Are Reshaping the $400B AI Foundation Model Market - Benchmarks, Revenue, Profit, Growth, and Focus Areas
The frontier model market in 2026 is a $400B+ revenue cohort with 5 credible frontier labs (OpenAI, Anthropic, Alphabet Gemini, Meta Llama, xAI Grok) and 1-2 emerging challengers (DeepSeek, Mistral). The frontier tier is now defined by 100T+ parameters, 1M+ token context windows, multimodal training, and reasoning-specialized architectures. This is a full-stack deep-dive: benchmark performance (SWE-bench, MMLU, GPQA, FrontierMath, HLE), revenue ($13B OpenAI / $5B Anthropic / $3B Gemini), profit path (Anthropic turning profitable, OpenAI $5B loss in 2025, xAI burning $10B+/yr), growth rates (200%+ YoY for Anthropic, 100%+ for Gemini), focus areas (OpenAI agentic + ChatGPT, Anthropic coding + enterprise, Gemini search + Workspace, Meta open-source + ads, xAI real-time + X, DeepSeek cost-efficient open-weight).
GPU / NPU / ASIC / TPU / CPU Deep Dive: How Nvidia, AMD, Broadcom, Alphabet, Apple, and Intel Are Reshaping the $400B AI Compute Market - Performance, Bottlenecks, Supply Chain, and Demand
The AI compute market in 2026 is a $400B+ market segmented into 5 binding architectures: GPU (Nvidia H100/H200/B100/B200/GB300 + AMD MI300/MI325/MI350), NPU (the cleanest single on-device AI accelerator for Apple Neural Engine + Qualcomm Hexagon + Intel NPU), ASIC (Broadcom custom ASIC for hyperscalers + Marvell custom silicon), TPU (Alphabet TPU v5/v6/v7), and CPU (Intel Xeon + AMD EPYC + Apple M-series). This is a full-stack deep-dive: performance benchmarks (FP8 / FP16 / BF16 / INT8 / FP4 throughput), the HBM + CoWoS + SiP supply chain, the top experts, the customer base, the capex, and the read-through for the AI capex cycle.
HBM Deep Dive: How SK hynix, Samsung Electronics, and Micron Built the Single Most Concentrated AI Supply Chain of the Cycle - and Why the 2026-2028 Capacity Is Already Sold Out
HBM (High-Bandwidth Memory) is the binding AI accelerator supply chain constraint, and the 2026-2028 supply is already spoken for. SK hynix leads with ~52% HBM market share, Samsung Electronics is at ~20%, and Micron is at ~28% with the cleanest 2026 ramp. This is a full-stack deep-dive into HBM: TSV technology, stack height (HBM2/2E/3/3E/4/4E), 12-Hi/16-Hi stacks, base die logic process, the HBM3E race, capacity sold out through 2026, top experts at SK hynix's M16, expansion capex, the customer concentration (Nvidia, AMD, Broadcom), and the read-through for the 2026-2028 AI capex cycle.
Liquid Cooling Deep Dive: How Vertiv, Boyd, Rittal, CoolIT Systems, and Asetek Built the Most Concentrated AI Data Center Thermal Stack of the Cycle
Liquid cooling is the binding AI data center thermal stack of the cycle - air cooling is no longer sufficient for AI accelerators >700W TDP, and the 2026-2028 liquid cooling demand is the cleanest single read on the AI capex cycle's thermal leg. Vertiv leads with ~30% liquid cooling share, Boyd at ~10%, Rittal at ~10%, CoolIT Systems (the cleanest pure-play, recently acquired by Eaton) at ~8%, and Asetek at ~5%. This is a full-stack deep-dive into liquid cooling: cold plate (DLC) vs immersion (single-phase + two-phase) vs rear-door heat exchanger (RDHx), the coolant chemistry (PG-25, dielectric fluids), the customer base (Microsoft, Alphabet, Amazon, Meta), the top experts, the capex, the 2026-2028 supply-demand, and the read-through for the AI capex stack.
NAND Flash Deep Dive: How Samsung Electronics, SK hynix, Kioxia, Sandisk, and Micron Run the Only Memory Market That Stopped Behaving Like a Commodity
NAND flash is the only memory sub-segment where AI demand did not just lift ASPs - it broke the commodity cycle itself. Sandisk and Micron reported datacenter-grade NAND revenue inflection in 2026, Samsung Electronics and SK hynix are running near-full-utilization, and Kioxia is the latest target of a $15-20B IPO/pre-IPO secondary at a ~$20B mark. This is a full-stack deep-dive into the NAND industry: cell architecture (SLC/MLC/TLC/QLC/PLC), node roadmap, capital intensity, controller + firmware stack, top experts, expansion difficulty, debt load, customer concentration, and the read-through for Apple, Nvidia, the AI capex stack, and the 2026-2030 storage market.
Optical Module Deep Dive: How Coherent, Lumentum, Innolight, Eoptolink, and Source Photonics Run the $40B 800G / 1.6T Optical Module Cycle - and Why 3.2T CPO Is the 2027-2028 Binding Test
Optical module is a $40B/year market in 2025 that is structurally being reshaped by the 800G / 1.6T cycle and the emerging 3.2T co-packaged optics (CPO) cycle. The 5 leading-edge optical module suppliers are Coherent, Lumentum, Innolight, Eoptolink, and Source Photonics. This is a full-stack deep-dive: 800G / 1.6T / 3.2T optical module + EML + DML + silicon photonics (SiPh) + CPO + linear pluggable optics (LPO) + tunable laser + EDR / FR / DR / SR / LR / ER / ZR reach classes, the EML + SiPh + CPO technology stack, the Nvidia + Broadcom + Cisco + hyperscaler customer base, the 2025-2028 capex, the top experts, and the read-through for the AI capex cycle.
What to expect
Evidence-first notes with a visible point of view.
This section collects sharp takes on earnings, shareholder meetings, and market structure. Each new piece should make the thesis, the facts, and the implications obvious within the first few screens.
Expect direct analysis, not generic commentary.
Expect the data to be explicit and the argument to be easy to follow.