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.
2026-07-03
2026-07-02

The AI Trade Split in Two as Meta Rallied and Semis Slid
On July 1, Meta rallied more than 10% intraday while semiconductor stocks sold off and SOXX fell 6.48% on the day. The market was not dumping the whole tape; it was repricing the AI stack into winners that can monetize and suppliers that are already priced for perfection.

AppLovin's Open Access Bet Turns Adtech Into a Distribution Story
AppLovin's stock rose after Raymond James initiated coverage and the company opened AppLovin Ads to all advertisers. The deeper thesis is that self-serve access turns AppLovin from a managed performance platform into a broader distribution business.

Getty and Shutterstock's Breakup Shows AI Content Is a Rights Market, Not Just a Scale Market
Getty Images and Shutterstock are scrapping their merger after UK conditions made the deal uneconomic, and Shutterstock shares dropped roughly 30%. The deeper read is that AI-generated content has made rights, regulation, and distribution more important than simple scale.

Meta's Cloud Pivot Turns Excess AI Compute Into a Monetization Test
Reuters reported that Meta is building a cloud business to sell excess AI computing capacity, and the stock jumped more than 10% intraday. The deeper story is that AI infrastructure is starting to behave like an asset with resale value instead of just a cost center.
2026-07-01
The AI Capex Boom Is Turning Into a Depreciation and Free-Cash-Flow Test
Alphabet, Amazon, Meta, and Microsoft are all spending at a pace that looked absurd two years ago. The real question is no longer who can spend the most, but who can turn that spending into durable free cash flow before depreciation and lease commitments eat the story.
Anthropic's Sonnet 5 Is a Pricing Reset for Enterprise AI
Anthropic's June 30 launch of Claude Sonnet 5, paired with Claude Science and temporary launch pricing of $2 per million input tokens and $10 per million output tokens, signals that enterprise AI is moving from benchmark theater to workflow economics.
Data Center Power Is Becoming the Real AI Scarcity Asset
The IEA and EIA now describe data centers as one of the biggest drivers of future electricity demand. The scarce input is no longer only GPUs: it is grid capacity, interconnection speed, substation gear, and politically permitted electrons.

DeepSeek's DSpark Shows Open-Source Inference Is Becoming Part of the Moat
DeepSpec says DSpark is the same checkpoint with an added speculative-decoding module, while DeepSeek-V4-Pro ships 1.6T parameters, 49B activated, 1M-token context, 27% FLOPs, and 10% KV cache versus V3.2. The read-through is that inference efficiency is now a core product feature, not a postscript.

Gemini Omni Makes Video the New Google AI Funnel
Google says Gemini Omni can create and edit video from multimodal inputs and is rolling into the Gemini app, Google Flow, and YouTube Shorts. That pushes the AI fight from chat UI into creator tools, media production, and cloud distribution.
HBM Squeeze Battle Is No Longer Just About HBM
TrendForce says suppliers are prioritizing HBM and advanced-layer NAND, squeezing mature-node supply and pushing prices up across the memory stack. Samsung's HBM4E and Micron's latest cash generation show the market has moved from a shortage trade into a capacity-allocation regime.

Liquid Cooling Is Becoming the Default Thermal Stack for AI Data Centers
NVIDIA, Supermicro, Vertiv, and Dell are all describing 100kW+ racks, 120-132kW liquid-cooled designs, and large efficiency gains. The read-through is that liquid cooling has moved from an optional upgrade to the default thermal stack for AI factories.
Qualcomm's AI Pivot Is Now a Stack Bet, Not a Press Release
The Meta CPU agreement, Modular acquisition, and Dragonfly roadmap suggest Qualcomm is trying to become an edge-to-cloud AI platform rather than just a handset chip vendor.
Silicom's First AI Inference Order Is a Small Revenue Line With a Big Stack Signal
A first production order for an inference-specific solution suggests the next phase of AI capex is moving deeper into networking, data infrastructure, and edge-aware hardware. GPUs still matter, but the value chain is widening.
Travelers' Insurance-Specific LLM Shows Where Enterprise AI Actually Pays
TravelersLLM is a useful case study in regulated AI: proprietary data, narrow scope, measurable underwriting lift, and no need to chase consumer hype.

Gemini Omni Turns Video Into Google's New AI Battleground
Google says Gemini Omni can create and edit video from multimodal inputs and is rolling into the Gemini app, Google Flow, and YouTube Shorts. That pushes the AI fight from chat UI into creator tools, media production, and cloud distribution.
2026-06-30
Agility Robotics Turns Humanoid Labor Into a Public-Market Story
Agility Robotics is taking the first U.S. humanoid robotics pure-play public at a $2.5 billion valuation, and the real story is not the ticker. It is the proof that robotics is moving from demo culture toward industrial deployment, with safety, sourcing, and customer orders now deciding who gets capital.
Google's Gemini Capacity Wall Shows the Real Bottleneck in AI Is Compute Allocation
Google limiting Meta's use of Gemini is a clean signal that frontier AI is now constrained by capacity, not just model quality. That shifts pricing power toward the owners of cloud, custom silicon, and power.
2026-06-29
The BIS Just Turned the AI Boom Into a Macro Warning
The BIS is warning that the AI boom now looks like a financing story, with valuation, credit, and capex risk all moving together.
South Korea Is Turning the AI Boom Into Industrial Policy
South Korea's new chip and AI investment plan turns memory, packaging, data centers, and robotics into a national growth strategy.
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.

