Unlock the SemiAnalysis’ end-to-end AI accelerator market model – quantifying historical and forward shipments, ASPs, and revenue by SKU, company, and customer.
From Nvidia, AMD, Google TPU, Meta MTIA, AWS, Microsoft, OpenAI, Tesla and more, we track merchant and semi-custom roadmaps alongside supply and demand to forecast revenue for leaders like Nvidia, AMD, Broadcom, Marvell, Intel, Alchip, and Mediatek.
Tie silicon to supply chains with wafer starts, 2.5D packaging volumes and yields (TSMC, Samsung, Intel, Amkor, ASE SPIL), die-attach steps, and HBM details (type, capacity, layers, stacks, total bits, manufacturer) to anticipate bottlenecks and opportunities.
See who’s buying: shipments and installed base by over 60 major customers – from U.S. and Chinese hyperscalers to enterprises, neoclouds, startups, and sovereign AI – plus peak and effective FLOPS using training MFU by chip.

Build, price, invest in, and scale your AI cloud with confidence using SemiAnalysis’ AI Cloud Total Cost of Ownership Model – an end-to-end toolkit that quantifies GPU server $/hr, training $/PFLOP, and inference $/M tokens across Nvidia, AMD, Intel, and custom accelerators.
See the road ahead with granular install-base projections through 2028, shipment estimates to 2034, and scenario-driven rental price forecasts rooted in supply-demand and the evolving accelerator cost curve.
Optimize performance and margins with analysis of pipeline/tensor/expert/data parallelism, plus market-wide throughput benchmarks to compare “most advanced” vs. average costs.
Make informed investment decisions with a full three-statement financial model – NPV, IRR, ROA/ROIC/ROE, residual value, cash flows, depreciation, power/colo, financing mix, and support for various contract structures.

Decode the AI network stack with SemiAnalysis’ AI Networking Model, completing our infrastructure suite by mapping switches, optics, cables, and AEC/DACs across scale-up, scale-out, front-end, and OOB networks.
Dive into 80+ hyperscaler configuration panels (Microsoft, Google, Meta, Amazon, Oracle, X.AI, and neoclouds) that tie every in-use accelerator SKU to precise attachment ratios, pricing, and power for switches, transceivers, fibers, AECs/ACCs/DACs.
Trace spend with by-customer as well as market-wide units and spend for switches, transceivers and cables. Includes forecasts on market-wide AI volumes for 200G/400G/800G/1.6T transceivers from leaders like Broadcom, Arista, and Nvidia.
Gain detailed breakdowns of vendor wallet share at each hyperscaler for optical modules (1.6T/ 800G/ 400G), switches (BlackBox/ WhiteBox), and AECs (800G/400G).

Map the world’s AI build-out with SemiAnalysis’ AI Datacenter Model – an end-to-end view of critical IT power, capacity, and constraints across >5,000 facilities, from hyperscale to colo, grounded in permits, property records, FOIA, power data, and satellite imagery accelerated by computer vision.
Quantify the demand side with accelerator shipment and deployment forecasts through 2027, translating SKU-level power needs into regional critical IT demand, PUE, utility power, and annual consumption across the U.S., EMEA, APAC, China, and more.
Gain insight who is short on megawatts with historical, current, and forecast capacity by hyperscaler, site, cluster, and region – plus capex outlooks by power, cooling, and facilities, and self-build vs. leased MW for Microsoft, AWS, Google, Meta, Oracle, Apple, CoreWeave, Tesla, x.AI, and others.
Tie infrastructure to economics with AI/GPU cloud TCO, electricity tariff surveys, solar economics, power-cost scenarios, and carbon footprints for training and inference – so investors and operators can price, plan, and prioritize with confidence.

Illuminate the foundry landscape with SemiAnalysis’ Foundry Metrics Model—a decision-ready reference that quantifies wafer shipments, ASPs, capacity, utilization, node mix, margins, capex, and regional/application revenue with quarterly forecasts through 2030.
Benchmark every major foundry in minutes, with a TSMC spotlight covering top customers, wafer starts by customer, Apple exposure, node cost modeling, and segment mix.
Utilizes 20+ years of history for top players, 10+ for others, and a triangulated methodology blending wafers-per-tool models, filings, customs data, and primary research.
Expand your edge with optional add-ons for end-market demand, vendor/node roadmaps and forecasts, top-customer datasets, and P&L modeling for leading-edge nodes and fabs.
Delivered quarterly as an Excel workbook and PDF snapshot—with event-driven updates—this institutional data book becomes your single source of truth for capacity shifts, pricing trends, and competitive dynamics.

Follow the money of AI with SemiAnalysis’ AI Tokenomics Model – our demand-side framework that ties hardware inputs (Nvidia/AMD GPUs, Google TPU, Trainium, and more) to software outputs and translates usage and revenue growth into future accelerator demand.
Benchmark ROI and profitability across hyperscalers, frontier labs, and neoclouds with bottoms-up token-throughput forecasts that factor in system configs (GB200 NVL72, VR144, TPU v7, Trainium 3), model architectures (GPT-5, Sonnet 4, DeepSeek V3, Kimi K2), and real workloads (coding, chat, document analysis, agentic).
See the full market map – compute supply vs. demand by Microsoft, Google, Amazon, Meta, Oracle, OpenAI, Anthropic, DeepSeek, CoreWeave, Nebius, Crusoe – grounded in our ClusterMAX ratings and addressable-market sizing across apps, API endpoints, and token-consuming software companies.
Quantify the business shift from seat-based SaaS to token consumption with unit-economics benchmarks, ROIC calculations, and revenue mix projections spanning rental, model, and software income streams.
Built for investors, operators, and policymakers, this model delivers a usage-driven hardware forecast plus inference and training demand curves – so you can price growth, plan capacity, and evaluate AI business models with confidence.

Turn wafer capacity into a defensible equipment outlook with SemiAnalysis’ Wafer Fab Model – a bottoms-up engine that links node roadmaps and layer-by-layer process flows to sales forecasts through 2027.
Go deep on advanced logic with a granular lithography stack (EUV, ArFi, KrF) from first EUV use to 1.4-nm-class with GAA and backside power delivery, plus a next-gen spec database covering high-NA EUV.
See model-, unit-, and revenue-level lithography forecasts by chipmaker and node, alongside capacity and capex views for leading-edge logic, 10-nm-class DRAM, and NAND – segmented across logic, specialty, memory, and China.
Benchmark every major vendor – ASML, Applied Materials, Lam, KLA, Tokyo Electron, ASM, Screen, Kokusai, and more – with sales projections for litho, deposition, etch, track, and metrology/inspection, plus an easy-reference tool database (throughput, uptime, rework, ASP).
Anticipate inflections with scenarios for pattern shaping (e.g., Sculpta) and DSA, and get quarterly updates and onboarding to align forecasts with real fab build-outs.

Turn noise into conviction with SemiAnalysis Core Research – a weekly and ad-hoc intelligence stream distilling fast-moving semiconductor and AI shifts into market-moving signals.
Built on deep supply-chain forensics, proprietary datasets, and architectural teardowns, our work connects silicon physics to stock performance, capex cycles, and hyperscaler deployment economics.
We equip enterprises, investors, and governments with pre-consensus insights on GPU/TPU/ASIC roadmaps, foundry capacity, export controls, and subsidy-led manufacturing realignments.
From wafer-level cost models to cluster deployment tracking and regulatory scenario analysis, our methodology turns complexity into clear, actionable calls.
If you value our retail newsletter, Core Research is the institutional upgrade – a signal-rich dive into how Public Equity markets are affected by new developments.

SemiAnalysis ChipBook is a monthly AI and semiconductor tracker that condenses thousands of curated data points into clear, actionable views of the industry and its fastest-moving sectors.
Used by leading investment funds and semiconductor companies, it helps investors generate ideas and validate theses while guiding executives on R&D prioritization and capacity expansion timing.
Each edition delivers three parts: Core Industry Trackers, Timely Datasets pulled from 125+ actively monitored series, and distilled insights from our proprietary models (including the AI Accelerator and Datacenter models).
Coverage spans smartphone/PC imports, WFE trade flows, auto exports, bare wafer and PCB trade—plus timely breakouts like China WFE by tool type, Taiwan CoWoS supplier revenue index, Nvidia server component supplier index, and memory IC exports.
Your subscription includes monthly analyst calls for tailored context, with consistent geographic depth across the U.S., Taiwan, China, Korea, and Japan -giving you a reliable compass for cycle positioning and inflection spotting.