About ChipStack
Research for the physical buildout behind AI.
ChipStack is built around a simple view: artificial intelligence is not only a software cycle. It is an industrial buildout touching semiconductors, memory, networking, power, cooling, and the companies supplying the stack. Our work is designed to help investors understand where bottlenecks form, how value accrues, and which public companies sit closest to the constraint.
Coverage
Semiconductors, memory, networking, cooling, power, and AI capex beneficiaries.
Style
Institutional-style synthesis focused on evidence, bottlenecks, and public-market relevance.
Output
Weekly notes, thesis updates, and event-driven alerts when supply-chain conditions change.
Research principles
We analyse the stack from constraint outward.
Instead of starting with headlines, we start with what is scarce, expensive, delayed, or operationally difficult. In industrial cycles, the bottleneck often tells you more than the narrative.
01
Follow the bottleneck
We start with the constraint that is hardest to replicate: advanced packaging, HBM, grid access, networking throughput, switchgear, liquid cooling, or generation capacity. That is usually where pricing power and duration emerge first.
02
Map the full stack
We treat AI as an industrial system, not just a software trend. Every note is anchored to dependencies across silicon, fabrication, memory, power, networking, thermal management, and hyperscaler capex.
03
Prioritise public-market relevance
Research is organised around listed companies, supply-chain leverage, incremental demand signals, and what management commentary implies for forward revenue visibility.
04
Separate narrative from evidence
We distinguish between storytelling and underwriting. Market excitement matters, but we rank evidence higher: bookings, lead times, fab capacity, utility queues, capital budgets, margin profile, and customer concentration.
Workflow
How a ChipStack note gets built.
Every piece begins with a real-world trigger — capex guidance, a product launch, a supply shortage, a utility constraint, a management comment, or a change in deployment economics.
Track hyperscaler capex, guidance changes, facility announcements, and supply agreements.
Map those signals into the physical stack: chips, memory, networking, power, cooling, construction, and generation.
Score listed companies by exposure, bottleneck position, durability, and downside risk.
Publish synthesis through weekly notes, thesis updates, and event-driven alerts when the narrative changes.
Coverage map
The parts of the AI economy we think matter most.
We focus on parts of the stack where industrial constraints, capital intensity, and supply-chain leverage can translate into durable economic advantage.
Semiconductors and advanced packaging
Active coverageHBM and memory supply constraints
Active coverageNetworking, optics, and cluster utilisation
Active coveragePower generation, grid equipment, and electrification
Active coverageCooling, thermal management, and data-centre infrastructure
Active coverageHyperscaler capex strategy and second-order beneficiaries
Active coverageDisclosures
Research, not recommendation.
ChipStack is built for investors who want sharper context around AI infrastructure. We aim to improve understanding of the stack, not provide personalised financial advice.
We care about evidence, underwriting discipline, and industrial logic. We do not assume every company exposed to AI will win, and we explicitly look for bottlenecks, margin pressure, concentration risk, and places where the market narrative may be ahead of fundamentals.
Start here