Skip to content
MINGYANG FUNDIndependent investment research
MINGYANG / AI INVESTMENT RESEARCH

Explore the AI stack.

Follow a request or trace an investment. Every layer has a different business model, constraint and source of value.

AN EXPLORABLE INVESTMENT LANDSCAPE

One request.
An entire industry.

Move through the layers. Find the bottleneck.
Understand who earns the next dollar.

01WORKFLOWS02INTELLIGENCE03ORCHESTRATION04COMPUTE05DATA MOVEMENT06PHYSICAL CAPACITY07THE SUPPLY BASECLICK A LAYER TO EXPLORE
Software → intelligence → compute → infrastructureEconomic dependencies · not measured transaction flows

From technical progress to investment consequences

01 / WORKFLOWS

Applications & agents

Cheaper models can improve margins or be competed away through lower software prices.

MSFTCRMNOWADBEPLTR
Related research ↗
02 / INTELLIGENCE

Models & reasoning

Model efficiency lowers unit cost; the net effect on hardware depends on demand growth and competitive pricing.

GOOGLMETAMSFTBABABIDU
Related research ↗
03 / ORCHESTRATION

Cloud & inference runtime

Cloud revenue can grow while depreciation and investment absorb the cash generated.

AMZNMSFTGOOGLORCLCRWVNBIS
Related research ↗
04 / COMPUTE

GPUs & custom accelerators

Our demand thesis allows both growing compute consumption and shifts in supplier share.

NVDAAMDAVGOMRVL
Related research ↗
05 / DATA MOVEMENT

Memory, optics & networking

Richer content helps only if qualification, yield, pricing and customer demand support returns.

MUSK HynixSamsungANETAVGOGLWCOHRLITEAAOI
Related research ↗
06 / PHYSICAL CAPACITY

Power, cooling & construction

The strongest exposure can sit at a physical constraint, but the entry valuation still matters.

GEVETNVRTHUBBEMEPWRCEG
Related research ↗
07 / THE SUPPLY BASE

Foundries, packaging & equipment

The original equipment thesis expects a lag from memory economics into investment; that lag must be confirmed by orders.

TSMASMLAMATLRCXKLACASX
Related research ↗