Compute
AI accelerator
A processor optimized for AI workloads such as training or inference, including GPUs and custom ASICs.
Investor lens: architecture leadership, availability and performance per watt can shape pricing power.
Compute
ASIC
Application-specific integrated circuit: a chip designed for a narrow workload rather than general-purpose computing.
Investor lens: custom silicon can lower cost but may reduce demand for merchant accelerators.
Semiconductors
Advanced packaging
Packaging technologies that connect multiple chips and memory at very high bandwidth.
Investor lens: packaging can become a bottleneck even when wafer capacity is available.
Financials
Backlog
Contracted or ordered business not yet recognized as revenue.
Investor lens: backlog quality matters more than headline size; cancellation terms and timing are critical.
Financials
Bookings
Value of orders signed during a period, whether or not recognized as revenue.
Investor lens: bookings can lead revenue, but definitions vary by company.
Financials
Capex
Capital expenditure used to build long-lived assets such as data centers, fabs, networking and power infrastructure.
Investor lens: AI growth can be real while free cash flow falls if capital intensity rises faster.
Semiconductors
Foundry
A company that manufactures chips designed by other companies.
Investor lens: leading-edge capacity, yields and customer concentration determine economics.
Financials
Free cash flow
Cash generated after operating costs and capital expenditures.
Investor lens: separates accounting earnings from the cash ultimately available to shareholders.
Financials
Gross margin
Revenue minus direct cost of goods or services, expressed as a percentage of revenue.
Investor lens: margin direction often reveals pricing power, mix and competitive pressure.
Memory
HBM
High Bandwidth Memory, stacked memory used close to accelerators to feed AI workloads with data quickly.
Investor lens: HBM supply and yields can constrain system shipments and shift profit pools.
Cloud
Hyperscaler
A cloud provider operating computing infrastructure at massive global scale.
Investor lens: hyperscalers are both major AI customers and increasingly silicon competitors.
AI economics
Inference
Running a trained model to generate outputs for users or applications.
Investor lens: inference economics determine whether AI usage can scale profitably.
Power
Interconnection queue
The process and waiting list for connecting new generation or large loads to the electrical grid.
Investor lens: long queues can delay data-center deployment and increase scarcity value for powered sites.
AI economics
Model training
The compute-intensive process of adjusting model parameters using large datasets.
Investor lens: training creates bursty, high-end compute demand and pushes architecture requirements.
Networking
Optical interconnect
High-speed data links using light to connect servers, switches and clusters.
Investor lens: bandwidth growth can pull demand through optics even when GPU growth moderates.
Macro
Real yield
A bond yield adjusted for expected inflation.
Investor lens: higher real yields increase discount rates and can compress valuations for long-duration growth assets.
Financials
SBC
Stock-based compensation paid to employees using equity rather than cash.
Investor lens: high SBC can dilute shareholders even when reported free cash flow looks strong.
AI economics
Tokens
Units of text or data processed by AI models during training or inference.
Investor lens: token growth only creates value when price, cost per token and usage economics work together.
Infrastructure
Utilization
The percentage of available compute or infrastructure capacity actually in use.
Investor lens: high utilization can improve returns; low utilization turns expensive capacity into a balance-sheet problem.
No matching term yet.