⚡API

Silicon Analysts Data API

Structured semiconductor data for AI agents, applications, and researchers. 36 REST endpoints, free to use — no API key required for read endpoints.

Your first call in 60 seconds

Four ways in. All of them work with no key and no signup — a key raises limits and unlocks history later.

curl — no key, no signup

Every read endpoint answers anonymously with current data.

# Get all AI accelerator cost breakdowns
curl https://siliconanalysts.com/api/v1/accelerators

# Filter by vendor
curl "https://siliconanalysts.com/api/v1/accelerators?vendor=NVIDIA"

# Calculate chip cost (POST)
curl -X POST https://siliconanalysts.com/api/v1/calculate/chip-cost \
  -H "Content-Type: application/json" \
  -d '{"dieWidth": 26, "dieHeight": 33, "processNode": "5nm"}'

# Get HBM market data
curl https://siliconanalysts.com/api/v1/hbm

# Search analysis articles
curl "https://siliconanalysts.com/api/v1/articles?q=tsmc&limit=5"

Connect an AI assistant (MCP)

Point any MCP client at https://siliconanalysts.com/api/mcp and ask questions in plain language. Full setup detail in Connect to AI Agents.

claude mcp add --transport http silicon-analysts https://siliconanalysts.com/api/mcp

Claude.ai — web · desktop · mobile (no config file)

Settings → Connectors → Add custom connector
Paste https://siliconanalysts.com/api/mcp
Click Allow — OAuth is handled automatically

Google Sheets or Excel — no add-on

Paste into a cell; it refreshes on its own. Excel, Power Query and the keyed paths are in Pull into Excel or Google Sheets.

=IMPORTDATA("https://siliconanalysts.com/api/v1/market-data/hbm-pricing?format=csv")

OpenAPI 3.1 spec

The whole surface as one machine-readable document, generated from the same registry this page renders.

/api/v1/openapi.jsonImport into ChatGPT Actions or any OpenAPI toolkit.

Connect to AI Agents (MCP)

New

MCP (Model Context Protocol) lets AI assistants like Claude directly query Silicon Analysts data during conversations — no manual copy-paste needed. Ask about chip costs, wafer pricing, or supply chain trends and get live data in real time.

Using Claude.ai? Connect in three clicks

Easiest

No API key, no config file, no terminal. Claude completes the sign-in for you and your usage is attributed to your account automatically.

Settings → Connectors → Add custom connector
Paste https://siliconanalysts.com/api/mcp
Click Allow — OAuth is handled automatically

Then just ask — “what does an H100 die cost at TSMC 4N?” — and Claude queries the data directly.

Client configs (Claude Code · Claude Desktop · Cursor)

Claude Code

claude mcp add --transport http silicon-analysts https://siliconanalysts.com/api/mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "silicon-analysts": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://siliconanalysts.com/api/mcp"]
    }
  }
}

Cursor.cursor/mcp.json

{
  "mcpServers": {
    "silicon-analysts": {
      "url": "https://siliconanalysts.com/api/mcp"
    }
  }
}

Every config above works anonymously — current data at full fidelity, ~20 req/24h per IP. Add an Authorization: Bearer sa_live_… header (or generate a key below) to reach 100 req/24h, historical series, and per-account usage attribution. Windsurf and other MCP-capable IDEs take the same shape in mcp_config.json; see /for-agents for the no-signup path. LangChain / Vercel AI SDK users can point any MCP or plain-HTTP tool adapter at the same endpoint; every response carries meta.freshness so agents know when re-fetching pays off. On the market-data series it also reports last_sourced and days_since_source — how old the underlying evidence actually is, so a cadence can never imply movement that did not happen.

Use from ChatGPT / OpenAI (no submission, works today)

ChatGPT (Team/Business/Enterprise) can add this server as a connector in developer mode — paste https://siliconanalysts.com/api/mcp as the MCP server URL. For the API, pass it to the Responses API MCP tool:

{
  "model": "gpt-5",
  "tools": [{
    "type": "mcp",
    "server_label": "silicon-analysts",
    "server_url": "https://siliconanalysts.com/api/mcp",
    "headers": { "Authorization": "Bearer sa_live_YOUR_API_KEY_HERE" },
    "require_approval": "never"
  }],
  "input": "What is the current CoWoS packaging cost per H100-class chip?"
}

Available MCP Tools

Tool	What it answers
get_accelerator_costs	AI chip specs, costs, and margins (18 chips)
calculate_chip_cost	Die dimensions → manufacturing cost estimate
estimate_lead_time	Mask layers + utilization → manufacturing cycle time (weeks)
get_hbm_market_data	HBM pricing, market share, and forecasts
get_market_pulse	Current supply chain signals and headlines
get_wafer_pricing	Wafer price ranges by process node
get_packaging_costs	Packaging costs and capabilities + HBM specs
get_hbm_qualification	HBM qualification matrix + status timeline per platform
get_market_dataset	Data points for one market dataset, with source and confidence per point
get_foundry_allocation	Foundry capacity allocation snapshot by node
get_foundry_economics	Quarterly per-node wafer ASP + utilization from public IR
get_recent_changes	What changed in the data ledger (7d/30d window)
get_market_intelligence	Sourced semiconductor intelligence briefs
get_market_advantage	Vetted time-advantage evidence (coverage-race wins)
get_benchmark_history	Historical benchmark observations, as-of any date
get_fab_capacity	Fab-level capacity snapshots over time
get_forecasts	Third-party forecast vintages by original publication date
get_wfe_signals	WFE-maker order-book (ASML/AMAT/Lam/KLA/TEL) from IR
get_fab_events	Fab construction/ramp milestone log, delays included
get_policy_events	Trade-policy timeline anchored to government documents

Example Prompts

After connecting, try asking Claude:

“What's the current wafer price range for TSMC N3?”

“Calculate chip cost for a 10x10mm die on N3 with CoWoS packaging”

“How long would a 3nm chip take to manufacture with CoWoS packaging at current utilization?”

“Compare H100 vs B200 manufacturing cost breakdown”

“What are the latest supply chain signals in advanced packaging?”

📘

Integration Guides

Building agents on the Anthropic SDK, LangChain, or the Vercel AI SDK? Step-by-step guides with working code show how to wire up these MCP tools in minutes.

Anthropic SDK →LangChain →Vercel AI SDK →All integrations →

Endpoints

All 36 endpoints, grouped. Cards expand for parameters and an example response; the same surface is machine-readable at /api/v1/openapi.json.

Start here5

GET/api/v1FreeAPI Manifest

This manifest; lists available endpoints.

Example Response

{
  "name": "Silicon Analysts Data API",
  "version": "1",
  "endpoints": [ ... ],
  "usage": { "tracking": "All requests logged" }
}

Cache: Default · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/usageFreeSelf-Serve Usage Export (Audit Trail)

Your own call history across REST + MCP (self-serve audit trail): timestamp, endpoint/tool, redacted params, status, latency, payload size, the request_id matching the X-Request-Id header your client received, and trace_id when your client sent a W3C traceparent header (we echo traceparent/tracestate on every response and store the trace-id — join our audit trail against your own distributed traces). Requires an API key — anonymous callers have no identity to export.

Query Parameters

window24h | 7d | 30d (default 7d)

limitMax rows after merging both surfaces (1–2000, default 500)

beforeISO timestamp cursor: rows strictly older than this

Example Response

{
  "success": true,
  "data": { "usage": [{
    "ts": "2026-07-05T10:12:03.412Z",
    "surface": "mcp",
    "target": "get_wafer_pricing",
    "status": "success",
    "latency_ms": 41, "response_bytes": 18234,
    "request_id": "7f3c…", "trace_id": "4bf92f3577b34da6a3ce929d0e0e4736"
  }] },
  "meta": { "count": 1, "window": "7d", "note": "request_id ↔ X-Request-Id; trace_id ↔ your traceparent" }
}

Cache: none (private, no-store) · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/catalogFreeDataset Catalog + Data Contracts

The dataset catalog: all 48 published datasets with access tier, surfaces (page/API/MCP), and a link to each DATA CONTRACT — definition, sources, cadence, revision policy, a coverage window COMPUTED from the backing series (never authored), backfill-vs-live capture honesty, and known gaps.

Example Response

{
  "success": true,
  "data": [{
    "id": "collect-wfe-signals",
    "name": "WFE Order-Book Signals",
    "kind": "collecting", "tier": "pro",
    "surfaces": { "page": "https://siliconanalysts.com/data/wfe-order-book", "api": "https://siliconanalysts.com/api/v1/wfe-signals", "mcp": "get_wfe_signals" },
    "contract": "https://siliconanalysts.com/api/v1/catalog/collect-wfe-signals/contract"
  }],
  "meta": { "counts": { "total": 48 }, ... }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/catalog/{id}/contractFreeData Contract (per dataset)

One dataset's data contract. coverage_window is computed live from the backing series, so a depth claim that disagrees with the data cannot exist; backfill_vs_live states archive depth vs continuous-capture start; known_gaps is owner-authored or honestly marked pending.

Example Response

{
  "success": true,
  "data": {
    "id": "collect-external-forecasts",
    "coverage_window": { "first": "2018-01-15", "last": "2026-08-05", "render": "date", "rows": 215, "includes_projections": false },
    "backfill_vs_live": { "kind": "backfilled", "subject_first": "2018-01-15", "capture_since": "2026-07-13", "backfill_rows": 187, "live_rows": 28 },
    "revision_policy": "Rows are immutable observations...",
    "known_gaps": { "status": "pending_author" },
    ...
  }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/schemaFreeData Dictionary (Point-in-Time Semantics)

Machine-readable data dictionary: per-dataset point-in-time semantics (which timestamp is first-seen vs data vintage, what mutates vs what is frozen, how corrections work), field documentation, refresh cadences, and the endpoints/MCP tools serving each dataset. The DDQ answer sheet for data-engineering reviewers.

Example Response

{
  "success": true,
  "data": {
    "version": "data-dictionary-v1",
    "pit_classes": { "append_only_event": "Rows never change; the table is its own complete history.", ... },
    "sourcing_policy": "Public sources only — no insider data. ...",
    "datasets": [{
      "id": "hbm_qualification_events",
      "pit_class": "append_only_event",
      "first_seen_field": "detected_at",
      "vintage_field": "as_of",
      "correction_policy": "Append-only; a wrong event is superseded by a later event...",
      "fields": [ ... ]
    }]
  }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

Core datasets13

GET/api/v1/market-dataFreeMarket Data Datasets

List all 16 semiconductor market data datasets with metadata (name, category, unit, description, cadence). Covers wafer pricing, HBM/DRAM pricing, CoWoS capacity, fab utilization, NRE costs, AI chip BOM, capex, lead times, NAND pricing, cloud GPU pricing.

Example Response

{
  "success": true,
  "data": [{
    "id": "wafer-price-tsmc",
    "name": "Silicon Wafer Price by Node (TSMC 300mm)",
    "category": "Wafer Pricing",
    "unit": "USD/wafer",
    "description": "Estimated 300mm wafer prices...",
    "proOnly": false
  }, ...],
  "meta": { "count": 14 }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/market-data/{datasetId}FreeMarket Dataset — Data Points

Data points for a specific market dataset. Free tier returns recent data points with source citations and confidence ratings. Pro returns full history from 2004.

Query Parameters

datasetIdDataset identifier (e.g., wafer-price-tsmc, hbm-pricing, cowos-capacity)

Example Response

{
  "success": true,
  "data": {
    "dataset": {
      "id": "hbm-pricing",
      "name": "HBM Price per GB by Generation",
      "category": "memory",
      "unit": "$/GB"
    },
    "dataPoints": [
      { "series_key": "hbm3e", "period_label": "Q2 2026", "value_mid": 15.3,
        "data_type": "Estimate", "confidence": "Medium",
        "source_name": "TrendForce", "source_url": "https://..." }
    ]
  },
  "meta": { "count": 3, "gated": true, "note": "Free preview — Pro returns full history" }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/acceleratorsFreeAI Accelerator Cost Breakdowns

AI accelerator manufacturing cost breakdowns (18 chips: NVIDIA H100/H200/B100/B200/GB200/GB300/Rubin, AMD MI300X/MI355X/MI455X, Intel Gaudi 3, Google TPU v5p/v6e, AWS Trainium 2/3, Microsoft Maia 100, Meta MTIA v2). Includes logic die, HBM, packaging, test costs, sell price, and gross margin.

Query Parameters

vendorFilter by vendor (case-insensitive)

chipFilter by chip name (partial match, case-insensitive)

fieldsReturn only specified fields (comma-separated)

Example Response

{
  "success": true,
  "data": [{
    "chip": "NVIDIA H100 SXM5",
    "vendor": "NVIDIA",
    "processNode": "TSMC 4N",
    "estMfgCostUsd": 3320,
    "estSellPriceUsd": 32500,
    "chipGrossMarginPct": 87.5,
    "costBreakdown": {
      "logicDieCostUsd": 300,
      "hbmCostUsd": 1350,
      "packagingCostUsd": 750,
      "testAssemblyCostUsd": 920
    },
    "provenance": {
      "last_updated": "2026-06-20T22:38:53Z",
      "source_type": "derived",
      "confidence_tier": "medium",
      "dataset_version": "chipSpecs-v1.0",
      "data_as_of": "2026-04-07T00:00:00.000Z"
    },
    ...
  }],
  "meta": {
    "count": 18,
    "source": "Silicon Analysts",
    "freshness": {
      "cadence": "Weekly — cost models recompute when inputs move (Monday refresh fleet)",
      "next_refresh": "2026-07-06T12:30:00.000Z",
      "delta_check": "https://siliconanalysts.com/api/v1/changes?window=7d — empty = nothing moved"
    }
  }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/hbmFreeHBM Market Analysis Dataset

Full HBM market analysis dataset (accelerators, specs, market share, spot prices, leading indicators, qualification feed, revenue forecast, supplier revenue, validation checks, bit demand). bitDemand is a derived monthly series of HBM bits by SKU class + customer type (EB ranges) — NOT a training-vs-inference split (dual-use SKUs make that attribution dishonest; no public source publishes one).

Query Parameters

Example Response

{
  "success": true,
  "data": {
    "accelerators": [ ... ],
    "specs": [ ... ],
    "marketShare": [ ... ],
    "spotPrices": [ ... ],
    "leadingIndicators": [ ... ],
    "qualificationFeed": [ ... ],
    "revenueForecast": [ ... ],
    "supplierRevenue": [ ... ],
    "validationChecks": [ ... ],
    "bitDemand": [{
      "metric": "hbm_bits_sku_class",
      "entity": "Blackwell",
      "period": "2026",
      "value_text": "≈1.15–1.92 EB (6–10M units × 192–384 GB/unit)",
      "value_num": 1.54, "value_min": 1.15, "value_max": 1.92, "unit": "EB",
      "observed_week": "2026-08-03", "as_of": "2026-01-01",
      "confidence": "medium", "agreement": "strong", "independent_sources": 2,
      "source_url": "https://...", "evidence": "Derived: public shipment-mix estimates × per-SKU HBM content...",
      "provenance": { "last_updated": "...", "source_type": "derived", "confidence_tier": "medium", "dataset_version": "hbmData-v1.0" }
    }]
  },
  "meta": { "source": "Silicon Analysts", "bitDemandNote": "bitDemand is NOT a workload split — ..." }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/hbm-qualificationFreeHBM Qualification Tracker

Sourced HBM qualification tracker: which memory vendor (SK Hynix, Samsung, Micron) passed which AI-accelerator customer qualification (NVIDIA, AMD, Broadcom), by generation (HBM3/HBM3E/HBM4) and stack height. Current-state matrix + per-relationship status-change timeline back to 2022; every datapoint dated, sourced (URL), and confidence-scored. Refreshed daily, human-reviewed.

Query Parameters

vendorFilter: SK Hynix | Samsung | Micron

customerFilter: substring match (e.g. "Vera Rubin", "MI350")

generationFilter: HBM3 | HBM3E | HBM4

include_timelinesSet false to return the matrix only

Example Response

{
  "success": true,
  "data": {
    "matrix": [{
      "relationship": "sk hynix|nvidia h100|hbm3|",
      "vendor": "SK Hynix", "customer": "NVIDIA H100",
      "generation": "HBM3", "stack_height": null,
      "status": "volume_shipping", "as_of": "2022-06-01",
      "confidence": "high", "source_type": "press_release",
      "source_url": "https://news.skhynix.com/sk-hynix-to-supply-industrys-first-hbm3-dram-to-nvidia/"
    }],
    "timelines": { "sk hynix|nvidia h100|hbm3|": [ ... ] }
  },
  "meta": { "methodology": "...", "vendors": [ ... ] }
}

Cache: 10 minutes (anonymous) · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

Access: Matrix: open. Timelines: full depth with a free API key or signed-in session; anonymous callers may receive the latest event per relationship only.

GET/api/v1/articlesFreeAnalysis Article Index

Analysis article index with metadata for 30+ semiconductor industry articles. Includes title, date, description, executive insight, key takeaways, cited sources, and full URL.

Query Parameters

limitReturn only the N most recent articles

qSearch in title and description (case-insensitive)

Example Response

{
  "success": true,
  "data": [{
    "slug": "nvidia-gpu-prices-double-...",
    "title": "Nvidia GPU Prices Double as AI Demand...",
    "date": "2026-01-23",
    "description": "Analysis of Jensen Huang's comments...",
    "insight": "The spillover of AI-driven demand...",
    "keyTakeaways": ["...", "..."],
    "url": "https://siliconanalysts.com/analysis/...",
    "sources": [{ "publisher": "...", "title": "..." }]
  }],
  "meta": { "count": 30, "source": "Silicon Analysts" }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/market-pulseFreeSupply Chain Market Intelligence

Semiconductor market pulse data. Supply chain intelligence across 6 sectors (Logic, Memory, Packaging, Connectivity, Power, Geopolitics) with trend indicators and impact analysis.

Query Parameters

categoryFilter by category (case-insensitive)

trendFilter by trend direction (up, down, neutral)

limitReturn only the N most recent items

Example Response

{
  "success": true,
  "data": [{
    "category": "Memory",
    "headline": "HBM3e prices expected to increase 15-20%...",
    "trend": "up",
    "date": "Jan 2026",
    "source": "Silicon Analysts Research",
    "impact_analysis": "HBM demand continues to grow..."
  }],
  "meta": { "count": 12, "categories": ["Logic", "Memory", ...] }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/foundry/wafer-pricingFreeWafer Pricing by Process Node

Wafer price ranges by process node (TSMC, Samsung, Intel). Includes defect density, NRE costs, and node maturity status.

Query Parameters

nodeFilter to a specific node (e.g. tsmc-n3, tsmc-n5, tsmc-n16, samsung-14nm, intel-16, gf-12lp, umc-22-28, smic-28)

Example Response

{
  "success": true,
  "data": [{
    "node": "tsmc-n3",
    "label": "TSMC 3nm (N3)",
    "waferCost": { "avg": 19500, "min": 17000, "max": 22000, "unit": "USD/wafer" },
    "defectDensity": { "mature": 0.09, "early": 0.14, "unit": "defects/cm²" },
    "nreCost": { "maskSetCost": 15000000, "totalNRE": 22000000, "unit": "USD" },
    "maturity": { "status": "Ramping", "yieldRisk": "High" },
    "source": "Analyst Reports, TrendForce, Morgan Stanley",
    "citation": "TrendForce / Morgan Stanley, Nov 2024"
  }],
  "meta": { "count": 16, "updated": "2026-08-04" }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/llm-pricingFreeLLM API Token Pricing

Current LLM API token list-prices ($/Mtok input, output, cached input) across frontier and open-source models (Claude, GPT, Gemini, Grok, DeepSeek, Llama). Every row cites the provider's own pricing page with as_of vintage and confidence. Fully open — no key required.

Query Parameters

providerFilter by provider (Anthropic, OpenAI, Google, xAI, DeepSeek, Together)

modelFilter by model id or display name (case-insensitive substring)

Example Response

{
  "success": true,
  "data": [{
    "provider": "Anthropic",
    "modelId": "claude-opus-4-8",
    "displayName": "Claude Opus 4.8",
    "inputPerMtok": 5,
    "outputPerMtok": 25,
    "cachedInputPerMtok": 0.5,
    "unit": "USD/Mtok",
    "source": "https://platform.claude.com/docs/en/about-claude/pricing",
    "provenance": { "source_type": "research", "confidence_tier": "high", "data_as_of": "2026-07-11T00:00:00.000Z" }
  }],
  "meta": { "count": 12, "updated": "2026-07-11" }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/foundry/economicsFreeFoundry IR Economics (Quarterly ASP + Utilization)

Foundry IR economics: per-foundry, per-node, per-fiscal-quarter wafer ASP (min/max/blended, USD per 300mm-equivalent wafer) and fab utilization (%), derived from public IR materials (earnings releases/transcripts/decks) via the documented rev-mix scaling calculation. Six foundries (tsmc/umc/intel/samsung/smic/gf); every row carries source_urls + release_dates + confidence (high/medium/low). Latest period per foundry is free; multi-period history (quarter/from/to) requires a Pro key — free callers are clamped to latest with a meta.note, never 403d. include_facts=true returns the underlying evidence facts.

Query Parameters

nodeCanonical node token (n3, n5, n16-20, 22-28nm, 18a, …)

quarterExact period, e.g. 2026Q1 or 2025FY (Pro)

fromPeriod range start, e.g. 2024Q1 (Pro)

toPeriod range end (Pro)

include_factstrue — include the published IR evidence facts (verbatim quote + source per datum)

Example Response

{
  "success": true,
  "data": {
    "foundries": [{
      "foundry": "tsmc",
      "fiscal_quarter": "2026Q1",
      "fiscal_note": null,
      "utilization": { "pct": 93.8, "basis": "derived", "range": { "min": null, "max": null }, "confidence": "medium" },
      "blended_asp_usd": { "value": 8500, "basis": "derived_mix", "confidence": "high" },
      "source_urls": ["https://investor.tsmc.com/english/quarterly-results/2026/q1"],
      "release_dates": ["2026-04-16"],
      "nodes": [{
        "node": "n3",
        "node_label": "3nm (N3)",
        "node_group": "leading_3nm",
        "wafer_asp_usd": { "min": 18000, "max": 22000, "blended": 20000 },
        "asp_basis": "derived_mix",
        "revenue_share_pct": 25,
        "utilization": null,
        "confidence": "medium",
        "source_urls": ["https://investor.tsmc.com/english/quarterly-results/2026/q1"],
        "release_dates": ["2026-04-16"],
        "derivation": { "method_version": "rev-mix-v1", "derived_at": "2026-07-13T14:31:02Z" },
        "provenance": { "last_updated": "2026-07-13T14:31:02Z", "source_type": "derived", "confidence_tier": "medium", "dataset_version": "foundry-economics-v1" }
      }]
    }],
    "node_margin_estimates": [{
      "node": "N3", "label": "3nm (N3)",
      "est_wafer_price_usd": 20000, "est_wafer_price_range": [18000, 22000],
      "est_wafer_cost_usd": 13000, "est_wafer_cost_range": [12000, 14000],
      "est_gross_margin_pct": 35, "est_gross_margin_range": [30, 40],
      "confidence": "MEDIUM"
    }],
    "node_margin_note": "Single-vintage (2025) Silicon Analysts estimates for TSMC from public wafer-price/cost analysis. TSMC does not disclose per-node margin. Estimates, not observations; confidence per row.",
    "quarters_returned": ["2026Q1"],
    "truncated": false
  },
  "meta": { "methodology": "Scaling calculation over public IR statements...", "history_access": "Latest period free; history requires Pro", "count": 1 }
}

Cache: 5 minutes (latest view); Pro history responses are private · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/foundry/packaging-costsFreePackaging Costs & Capabilities

Packaging cost benchmarks and technology capabilities (CoWoS, EMIB, SoIC, FC-BGA, etc.). Also returns HBM memory specs and per-stack costs.

Query Parameters

typeFilter by packaging type (e.g. cowos, emib, soic, fc-bga)

Example Response

{
  "success": true,
  "data": {
    "packaging": [{
      "id": "cowos",
      "name": "CoWoS-S (2.5D)",
      "description": "Silicon Interposer. Standard for high-end chips.",
      "category": "hpc-ai",
      "costBenchmark": { "cost": 750, "unit": "USD/package", "costRange": { "min": 600, "max": 900 } },
      "capabilities": { "interposerActive": true, "stacking": false, ... }
    }],
    "hbmSpecs": [{
      "type": "HBM3e",
      "bandwidth": "1.2 TB/s",
      "capacity": "36 GB",
      "costPerStack": 300,
      "unit": "USD/stack"
    }]
  },
  "meta": { "packagingCount": 14, "hbmSpecCount": 5, "basis": "modelled_total", "basis_note": "Modeled FULL-PACKAGE cost per package ... NOT a quoted foundry/OSAT price." }
}

Cache: 1 hour · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/allocationFreeFoundry & Packaging Allocation

Foundry & advanced-packaging allocation by node/technology (TSMC, Samsung, Intel, etc. × N2, N3, CoWoS-S, CoWoS-L, SoIC). Returns current allocation status (fully_booked → available), lead-time weeks (min/max), utilization, price trend, geo risk, and publicly-reported customer allocation shares, plus optional time-series history (lead time, %-capacity-locked, per-customer share). Every record carries provenance (source type, confidence tier, last updated).

Query Parameters

foundryFilter by foundry (case-insensitive, e.g. TSMC, Samsung, Intel)

nodeFilter by node/technology (substring, e.g. n2, n3, cowos-l, soic)

categoryFilter by category (leading_edge, advanced, mature, packaging, hbm)

customerFilter to nodes/readings involving a customer (substring, e.g. NVIDIA)

include_historyInclude time-series history (true/false)

history_metricNarrow history to one metric (lead_time, pct_locked, customer_allocation, cowos_capacity, foundry_utilization)

limitMax current-snapshot items to return (1–100)

Example Response

{
  "success": true,
  "data": {
    "current": [{
      "id": "tsmc-n2",
      "foundry": "TSMC",
      "node": "N2 / A16",
      "allocation_status": "fully_booked",
      "lead_time_weeks_min": 78,
      "lead_time_weeks_max": 156,
      "lead_time_trend": "stable",
      "price_trend": "rising",
      "customers": ["Apple", "AMD", "MediaTek", "Qualcomm"],
      "customer_shares": [{ "customer": "Lead mobile customer (public est.)", "share": 50 }],
      "latest_signal": {
        "metric": "pct_locked",
        "value_text": "production schedule largely booked through late 2026",
        "as_of": "2026-06-01", "observed_week": "2026-06-29", "confidence": "medium"
      },
      "provenance": {
        "last_updated": "2026-06-15",
        "source_type": "research",
        "confidence_tier": "medium",
        "dataset_version": "allocation-v1"
      }
    }],
    "history": [{
      "metric": "pct_locked", "entity": "TSMC CoWoS", "node": "CoWoS-L",
      "value_text": ">85%", "value_num": 85, "unit": "%",
      "observed_week": "2026-06-15", "agreement": "strong",
      "independent_sources": 2,
      "provenance": { "confidence_tier": "medium", "dataset_version": "allocation-v1" }
    }]
  },
  "meta": { "snapshot_source": "table", "current_count": 13, "history_count": 11, "signals_current_through": "2026-06-29" }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/fabsFreeFab Capacity — Latest State

Fab-level capacity, latest state: newest published reading per (fab, tech_node_class) at/before as_of, joined with fab registry metadata (65 fabs; Frontend kwspm, Backend k units/month). Includes availability_status and full sourcing metadata (foundry_ir / wfe_vendor_earnings / government_filing taxonomy). The dated series lives at /api/v1/fabs/capacity-series (Pro).

Query Parameters

fab_idOptional: fab id (lowercase slug, e.g. tsmc-18-4)

foundryOptional: foundry name, case-insensitive (e.g. TSMC)

countryOptional: country, case-insensitive (e.g. Taiwan)

tech_node_classOptional: node class (e.g. '3nm Class') or backend packaging tech (e.g. CoWoS-S)

as_ofOptional latest-state cutoff YYYY-MM-DD (default today)

limitMax rows scanned (1–3000, default 1500)

Example Response

{
  "success": true,
  "data": [{
    "fab_id": "tsmc-18-4", "foundry": "TSMC", "fab_name": "Fab 18 P4",
    "country": "Taiwan", "tech_node_class": "3nm Class",
    "capacity": 100, "capacity_unit": "kwspm",
    "effective_date": "2026-06-30", "availability_status": "operating",
    "is_projection": false, "source_type": "foundry_ir", "source_url": "https://..."
  }],
  "meta": { "mode": "latest_state", "as_of": "2026-08-11", "fabs": 24,
            "units": { "Frontend": "kwspm (thousand 300mm wafer starts/month)",
                       "Backend": "k_units_month (advanced-package units)" } }
}

Cache: Default · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

Auth: None required

Feeds & change detection8

GET/api/v1/forecastsFreeThird-Party Forecast Vintages

Third-party forecast vintages: research-firm and company-guidance projections (TrendForce, WSTS, SEMI, TSMC/Micron/hyperscaler capex, …) archived with their ORIGINAL publication date, so the revision history is queryable — "what did TrendForce say about 2026 HBM bits in January vs July?". Every row carries originator, metric, target period, value, unit, as_of (publication date), a source URL, and a verbatim quote. Latest slice free; full history (from/to/since/all) needs a Pro key — free callers get the latest slice with a meta note, never 403d.

Query Parameters

originatorForecaster slug (e.g. trendforce, wsts, tsmc, micron, meta)

metricMetric slug (e.g. capex_usd, hbm_bit_demand_growth_pct)

target_periodPeriod forecast, e.g. CY2026 / 2027H1 / FY2026Q4

entity_idEntity the forecast is about, when scoped

fromHistory range start YYYY-MM-DD (Pro)

toHistory range end YYYY-MM-DD (Pro)

sinceEverything since YYYY-MM-DD (Pro)

alltrue — full series (Pro)

Example Response

{
  "success": true,
  "data": { "rows": [{
    "originator": "trendforce", "originator_type": "research_firm",
    "metric": "csp_capex_growth_pct", "entity_id": null,
    "target_period": "CY2027", "value_num": 50, "unit": "pct",
    "as_of": "2026-08-03", "source_type": "originator_primary",
    "source_url": "https://www.trendforce.com/presscenter/news/20260803-13161.html",
    "quote": "..."
  }] },
  "meta": { "view": "latest", "count": 1 }
}

Cache: 5 minutes (latest view); Pro history responses are private · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/wfe-signalsFreeWafer-Fab-Equipment Order-Book Signals

Wafer-fab-equipment order-book signals from ASML / AMAT / Lam / KLA / TEL quarterly IR: bookings, backlog, segment revenue, and guidance in the STATED currency (never converted). The 12–24-month leading indicator for fab capacity. Every row carries company, fiscal period, metric, value(s), unit, release date, source URL, and a verbatim quote. Latest slice free; full history Pro — never 403d.

Query Parameters

companyasml | amat | lam | kla | tel

segmentSegment/end-market when stated (litho, etch, semi_systems, services, china, memory, …)

fiscal_periodAs the company labels it, e.g. 2026Q2 / FY2026Q3

fromHistory range start YYYY-MM-DD (Pro)

toHistory range end YYYY-MM-DD (Pro)

sinceEverything since YYYY-MM-DD (Pro)

alltrue — full series (Pro)

Example Response

{
  "success": true,
  "data": { "rows": [{
    "company": "asml", "fiscal_period": "2026Q3", "period_basis": "calendar",
    "metric": "guidance_revenue", "segment": "",
    "value_num": null, "value_low": 11000000000, "value_high": 12000000000,
    "unit": "eur", "as_of": "2026-07-15", "source_type": "ir_release",
    "source_url": "https://www.asml.com/en/news/press-releases/2026/q2-2026-financial-results",
    "quote": "..."
  }] },
  "meta": { "view": "latest", "note": "Values in stated currency — never converted." }
}

Cache: 5 minutes (latest view); Pro history responses are private · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/fab-eventsFreeFab Construction & Ramp Milestones

Fab construction & ramp milestone log: announcement → groundbreaking → equipment move-in → risk production → HVM, plus expansions and schedule slips recorded as their own dated events (delays are never overwritten). Back to 2020. Every row carries foundry, fab name, event type, event date, a summary, source URL, and a verbatim quote. Latest slice free; full history Pro — never 403d.

Query Parameters

foundryFoundry/IDM (TSMC, Samsung, Intel, Micron, SK Hynix, …)

fab_idRegistry fab id where the fab is tracked (e.g. tsmc-fab21)

countryCountry slug

fromHistory range start YYYY-MM-DD (Pro)

toHistory range end YYYY-MM-DD (Pro)

sinceEverything since YYYY-MM-DD (Pro)

alltrue — full series (Pro)

Example Response

{
  "success": true,
  "data": { "rows": [{
    "foundry": "TSMC", "fab_name": "Fab 25 Taichung Phase 1", "country": "Taiwan",
    "event_type": "announced", "event_date": "2026-07-29",
    "announced_date": "2026-07-29", "is_projection": false,
    "summary": "TSMC's 1.4nm (A14) Fab 25 in Taichung is running ahead of schedule…",
    "source_url": "https://..."
  }] },
  "meta": { "view": "latest" }
}

Cache: 5 minutes (latest view); Pro history responses are private · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/policy-eventsFreeTrade-Policy Timeline

Semiconductor trade-policy timeline: export controls, entity listings, license policies, tariffs, and subsidies anchored to government primary documents (Federal Register/BIS, MOFCOM, METI, EU, …), back to the Oct 2022 BIS rule. Every row carries jurisdiction, agency, event type, title, published/effective dates, affected entities, a document reference, source URL, and a verbatim quote. Latest slice free; full history Pro — never 403d.

Query Parameters

agencyIssuing agency short code (BIS, USTR, MOFCOM, METI, EU-COM, …)

fromHistory range start YYYY-MM-DD (Pro)

toHistory range end YYYY-MM-DD (Pro)

sinceEverything since YYYY-MM-DD (Pro)

alltrue — full series (Pro)

Example Response

{
  "success": true,
  "data": { "rows": [{
    "jurisdiction": "us", "agency": "BIS",
    "event_type": "export_control", "title": "…",
    "published_date": "2026-05-13", "effective_date": "2026-05-13",
    "affected_entities": ["…"], "document_ref": "…",
    "source_url": "https://www.federalregister.gov/…", "quote": "…"
  }] },
  "meta": { "view": "latest" }
}

Cache: 5 minutes (latest view); Pro history responses are private · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/changesFreeWhat Changed (7d / 30d)

"What Changed" — recent movements in Silicon Analysts data over a 7d/30d window, derived from the daily snapshot ledger. Each moved metric returns direction (up/down), magnitude (percent for value metrics, percentage-points for rate metrics), old/new values, the two snapshot dates compared, window_days_actual (real lookback — clamps to available history), and per-record provenance. Domains: wafer_pricing, chip_cost, gpu_secondary (used-GPU market bands + derived residual value), margin_benchmark, foundry_capacity, defect_density, nre_cost, electricity_price, cloud_pricing, llm_pricing.

Query Parameters

windowComparison window: 7d (default) or 30d

sinceISO date (YYYY-MM-DD): compare the latest snapshot vs the newest snapshot at/before this date — "what changed since my last fetch". Overrides window.

datasetIdFilter to one domain (e.g. wafer_pricing, chip_cost, margin_benchmark)

minDeltaOverride the significance threshold (percent / percentage-points)

limitMax changes to return (1–200, default 50)

Example Response

{
  "success": true,
  "data": { "changes": [{
    "dataset_id": "wafer_pricing", "entity_id": "tsmc-n3",
    "field": "avg", "label": "TSMC N3 wafer price",
    "old_value": 19500, "new_value": 20000,
    "abs_delta": 500, "pct_delta": 2.6, "pp_delta": null,
    "direction": "up", "unit": "usd",
    "source_url": "https://...",
    "provenance": { "last_updated": "2026-08-10", "source_type": "derived",
                    "confidence_tier": "medium", "dataset_version": "changes-v1" }
  }] },
  "meta": { "window": "7d", "window_days_actual": 7, "count": 1 }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/trend-shapeFreePublic Trend Shape (Indexed)

Public trend SHAPE for any ledger series: the real dated trajectory rebased to an index (start = 100), so the shape and relative move are exact while absolute values stay Pro (/api/v1/snapshot-series). Returns indexed points per ledger day plus per-metric change_pct + direction. The "gate values, not shape" surface — a free viewer can see how a series moved without the sellable numbers. N2/Apple series are redacted.

Query Parameters

domainRequired: one ledger domain (e.g. hbm_spot_price, memory_spot, margin_benchmark)

entityRequired: entity id within the domain (lowercase slug, e.g. hbm3e-12-hi)

metricOptional: narrow to one metric (e.g. spot_price); omit for all metrics

fromInclusive start date YYYY-MM-DD (optional)

toInclusive end date YYYY-MM-DD (optional)

limitMax raw ledger rows scanned (1–2000, default 730)

Example Response

{
  "success": true,
  "data": [
    { "recordedAt": "2026-07-06", "change": 100, "spot_price": 100 },
    { "recordedAt": "2026-07-07", "change": 100, "spot_price": 100 }
  ],
  "meta": { "indexed": true, "base": 100,
            "note": "Shape only — absolute values require Pro (/api/v1/snapshot-series)" }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/intelligenceFreeMarket Intelligence Briefs

Market Intelligence — the freshest sourced semiconductor market briefs, generated daily from a Tavily + Claude scan. Each brief returns title, severity, confidence (0-100), quantitative_impact, executive summary, a short analysis, category (Logic/Memory/Packaging/Connectivity/Power/Geopolitics), and a curated sources list, with per-record provenance. Returns all severities; `published` flags the Critical/High briefs that also have a public page. Public sources only; N2/Apple omitted.

Query Parameters

severityFilter to one severity (Critical, High, Medium, Low)

categoryFilter to one sector (Logic, Memory, Packaging, Connectivity, Power, Geopolitics)

sinceOnly briefs at/after this ISO timestamp

publishedOnlyOnly briefs with a public page (true/false, default false)

limitMax briefs to return (1–100, default 25)

Example Response

{
  "success": true,
  "data": { "briefs": [{
    "slug": "samsung-hbm4-hits-80-golden-yield-2026-08-10",
    "title": "Samsung HBM4 Hits 80% 'Golden Yield' — Four Months Ahead of Schedule",
    "severity": "High", "confidence_score": 82, "category": "Memory",
    "quantitative_impact": "Yield from <60% (Feb 2026) to ~80% (Aug 2026) — approx. +20 pp",
    "summary": "…", "published": true,
    "sources": [{ "title": "…", "url": "https://…" }]
  }] },
  "meta": { "count": 1 }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

GET/api/v1/market-advantageFreeMarket Advantage (Time-Advantage Evidence)

Market Advantage — human-vetted TIME-ADVANTAGE evidence: dated, immutable proof that Silicon Analysts recorded a semiconductor supply-chain event before the first English-language coverage of it. Per win: our immutable first-seen timestamp, the cited source's publication time, the first English coverage's time + public URL/publisher, and the lead in hours. Only wins a human has confirmed for citation are returned; facts (timestamps + public URLs), not claims. Empty until a win is confirmed.

Query Parameters

limitMax wins to return (1–100, default 25)

order_bylead_time (biggest lead first, default) | recent

event_kindFilter: hbm_qual | capacity_signal | market_intelligence

Example Response

{
  "success": true,
  "data": {
    "wins": [{
      "event_kind": "hbm_qual",
      "event_ref": "SK Hynix|NVIDIA|HBM4|12-Hi",
      "our_first_seen_at": "2026-07-01T09:12:00Z",
      "english_first_seen_at": "2026-07-01T14:30:00Z",
      "english_publisher": "Reuters",
      "english_url": "https://...",
      "lead_time_hours_vs_detection": 5.3,
      "provenance": { "source_type": "derived", "confidence_tier": "high", "dataset_version": "market-advantage-v1" }
    }],
    "count": 1
  },
  "meta": { "citeAs": "Silicon Analysts — Market Advantage", "methodology": "https://siliconanalysts.com/data-quality#immutability" }
}

Cache: 5 minutes · Rate limit: 20 req/day (no key) · 100 req/day (free key) · 10k req/hr (Pro)

Calculators2

POST/api/v1/calculate/chip-costFreeChip Cost Calculator

Programmatic chip cost estimation. Accepts die dimensions, process node, and optional parameters. Returns full cost breakdown including yield, GDPW, and per-unit economics.

Request Body (JSON)

dieWidthDie width in mm (required)

dieHeightDie height in mm (required)

processNodeFoundry node ID (optional, default: tsmc-n5)

waferCostOverride wafer cost in USD (optional, uses benchmark if omitted)

packagingTypePackaging architecture (optional, default: flip-chip)

packagingCostOverride packaging cost in USD (optional)

hbmStacksNumber of HBM stacks (optional, default: 0)

hbmCostHBM cost per GB (optional, default: 0)

testCostTest cost per unit in USD (optional, default: 2)

marginTargetTarget gross margin percentage (optional, default: 50)

volumeProduction volume (optional, default: 10000)

energyFacilityOverheadApply the ~1.75× facility-overhead multiplier to the energy adder (optional, default: true; only meaningful with energyRegion)

substratePackaging substrate SCENARIO: wafer-300mm (default, no-op) | panel-310x310 — applies the midpoint of Yole's realistic 20–30% CoPoS panel-savings band to the packaging cost only; silicon GDPW unchanged (SA scenario; TSMC CoPoS pilot ~Jun 2026, MP 2028–29)

kgdTestCoverageKnown-Good-Die pre-bond test coverage % (0–100, optional). Returns a `kgdScenario` block: per-die test cost at this coverage vs. the packaged-defect waste it avoids, effective KGD quality, and the cost-minimizing coverage. Informational — never changes the headline cost. Absent = no-op (SA scenario).

Example Response

{
  "success": true,
  "data": {
    "estimatedChipCost": 1004.57,
    "dieArea": 806,
    "grossDiesPerWafer": 67,
    "frontendYield": 56.32,
    "totalYield": 55.19,
    "costBreakdown": {
      "waferCostPerGoodDie": 500.28,
      "packagingAndTestCost": 2,
      "hbmCost": 0,
      "marginCost": 502.28
    },
    "inputs": {
      "dieWidth": 26,
      "dieHeight": 31,
      "processNode": "tsmc-n5",
      "waferCost": 18500,
      "packagingType": "flip-chip",
      "hbmStacks": 0,
      "marginTarget": 50
    }
  },
  "meta": { "source": "Silicon Analysts", "apiVersion": "1" }
}

Cache: none · Rate limit: 20/24h anon · 20/24h free key · 10,000/hr Pro key (shared calc bucket)

POST/api/v1/calculate/lead-timeFreeLead Time Estimator

Heuristic chip manufacturing lead-time estimation (MANUFACTURING CYCLE TIME — wafer fab + packaging assembly/test, as min/max weeks). Mask layers × utilization-scaled days-per-mask-layer + a packaging-class adder. EXCLUDES allocation/booking queue time — see /api/v1/allocation for booking windows.

Request Body (JSON)

maskLayersTotal mask layers, integer 10–200 (required unless processNode is set)

utilizationFoundry utilization % 0–100 (optional; defaults from live foundry-allocation data for the node, else a documented constant — the fired default is echoed in assumptions)

packagingTypeCoarse class conventional | flip-chip | cowos (default flip-chip), or any platform packaging id (fc-bga, wirebond-bga, cowos-l, copos, qfn, fowpl, 3d-stacking, ...)

Example Response

{
  "success": true,
  "data": {
    "fabDays": { "min": 104, "max": 137 },
    "fabWeeks": { "min": 14.8, "max": 19.5 },
    "packagingWeeks": { "min": 8, "max": 14 },
    "totalWeeks": { "min": 22.8, "max": 33.5 },
    "effectiveDpml": { "min": 1.092, "max": 1.438 },
    "utilizationWeight": 1,
    "inputs": {
      "maskLayers": 95, "utilization": 98, "utilizationSource": "allocation_live",
      "packagingType": "cowos", "packagingClass": "cowos",
      "processNode": "tsmc-n3", "nodeDpmlFactor": 1.15
    },
    "assumptions": ["maskLayers defaulted to 95 (typical for TSMC 3nm (N3) ...)", "..."],
    "methodology": "fabDays = maskLayers × effective DPML (days per mask layer). ...",
    "sources": [{ "name": "Utilization operating curve", "citation": "FabTime Cycle Time Tutorial & Queueing Formulas (INFICON), 2023 ..." }],
    "provenance": { "source_type": "computed", "confidence_tier": "low", "dataset_version": "leadTimeData-v1.0" }
  },
  "meta": { "source": "Silicon Analysts", "allocationCaveat": "Manufacturing cycle time only. Booking/allocation queue time is EXCLUDED ..." }
}

Cache: none · Rate limit: 20/24h anon · 20/24h free key · 10,000/hr Pro key (shared calc bucket)

Full history (Pro)8

GET/api/v1/snapshotProPoint-in-Time Snapshot Replay (Pro)

Point-in-time replay of the daily snapshot ledger: the full frozen state of every published metric on a given day ("what did the feed say on date D?"). Serves the newest ledger day at/before the requested date and reports both dates. Ledger starts 2026-06-06 — append-only, never backfilled.

Query Parameters

dateYYYY-MM-DD (optional, default today). Served day = newest snapshot at/before this date.

domainFilter to one ledger domain (wafer_pricing, chip_cost, gpu_secondary, margin_benchmark, foundry_capacity, defect_density, nre_cost, electricity_price, cloud_pricing, llm_pricing, hbm_spot_price, memory_spot, hbm_leading_indicator, hbm_market_share, hbm_revenue_forecast, hbm_supplier_revenue)

limitMax rows (1–2000, default 2000; truncated flag reports overflow)

Example Response

{
  "success": true,
  "data": {
    "date_requested": "2026-06-20",
    "date_served": "2026-06-20",
    "ledger_start": "2026-06-06",
    "domain": "wafer_pricing",
    "rows": [{
      "domain": "wafer_pricing", "entity_id": "tsmc-n3", "metric": "avg",
      "value_num": 19500, "unit": "usd_per_wafer",
      "captured_at": "2026-06-20T08:00:12.412Z"
    }],
    "count": 45, "truncated": false
  }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/exportProFlat-File History Export (Pro)

Flat-file history export (JSONL default, CSV optional): any daily-ledger domain, the append-only event feeds (HBM qualification timeline, capacity signals, benchmark observations, fab capacity, foundry IR facts), or any curated market-data time series (hbm-pricing, component-lead-times, wafer-price-tsmc, … — enumerate via /api/v1/market-data). One HTTP call, download-ready (Content-Disposition), bounded at 10,000 rows with an explicit truncated flag — narrow with from/to to page longer histories. Where the dataset has published method notes, the X-Methodology response header links them.

Query Parameters

fromInclusive start date YYYY-MM-DD (optional). Market-data series are period-indexed: from/to apply to period_sort_key as YYYYMM bounds.

toInclusive end date YYYY-MM-DD (optional)

formatjsonl (default) | csv

Example Response

{"_meta":{"dataset":"hbm_qualification_events","count":214,"truncated":false,"time_axis":"detected_at","note":"Append-only history export."}}
{"detected_at":"2026-06-14T09:02:11.000Z","relationship":"SK Hynix|NVIDIA|HBM4|12-Hi","new_status":"qualified","as_of":"2026-06-13","source_url":"https://...","confidence":"high"}
{"detected_at":"2026-06-16T09:01:47.000Z","relationship":"Samsung|NVIDIA|HBM4|12-Hi","new_status":"in_qualification"}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/snapshot-seriesProLedger Time Series (Pro)

Full chartable time series for one entity from the daily snapshot ledger: one row per ledger day, pivoted across metrics ({ recordedAt, avg: 19500, ... }) — the single-fetch read path for trend charts. Append-only; series began 2026-06-06 and cannot be backfilled.

Query Parameters

domainRequired: one ledger domain (wafer_pricing, chip_cost, gpu_secondary, margin_benchmark, foundry_capacity, defect_density, nre_cost, packaging_benchmark, chip_archetype, electricity_price, hbm_spot_price, memory_spot, hbm_leading_indicator, hbm_market_share, hbm_revenue_forecast, hbm_supplier_revenue, cloud_pricing, llm_pricing, foundry_economics)

entityRequired: entity id within the domain (lowercase slug, e.g. tsmc-n3, h100-sxm5)

metricOptional: narrow to one metric (e.g. avg); omit for all metrics pivoted per day

fromInclusive start date YYYY-MM-DD (optional; ledger starts 2026-06-06)

toInclusive end date YYYY-MM-DD (optional)

limitMax raw ledger rows scanned (1–2000, default 730); on overflow the newest rows are kept and the truncated flag is set

Example Response

{
  "success": true,
  "data": [
    { "recordedAt": "2026-06-06", "avg": 19500, "min": 18000, "max": 21000 },
    { "recordedAt": "2026-06-07", "avg": 19500, "min": 18000, "max": 21000 }
  ],
  "meta": { "domain": "wafer_pricing", "entity": "tsmc-n3",
            "count": 2, "truncated": false,
            "note": "Append-only ledger — cannot be backfilled." }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/fabs/capacity-seriesProFab Capacity — Full Dated Series (Pro depth)

Full dated fab-level capacity series (per fab · tech_node_class · effective_date, ascending) with availability status, projections flag, and sourcing metadata. Node conversions appear as capacity shifting between node-class rows across effective_dates. Bitemporal: effective_date = what the reading reflects; curated backfill extends history.

Query Parameters

fab_idOptional: fab id (lowercase slug)

foundryOptional: foundry name, case-insensitive

countryOptional: country, case-insensitive

tech_node_classOptional: node class or packaging tech

limitMax rows (1–3000, default 1500); newest kept on overflow, truncated flag set

Example Response

{
  "success": true,
  "data": [
    { "fab_id": "tsmc-18-4", "tech_node_class": "3nm Class",
      "effective_date": "2025-12-31", "capacity": 80, "capacity_unit": "kwspm",
      "is_projection": false, "source_type": "foundry_ir" },
    { "fab_id": "tsmc-18-4", "tech_node_class": "3nm Class",
      "effective_date": "2026-06-30", "capacity": 100, "capacity_unit": "kwspm" }
  ],
  "meta": { "count": 2, "truncated": false }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/benchmarks/historyProCalculator Benchmark History (Pro depth)

Historical calculator benchmarks from the bitemporal observation ledger: wafer cost by node/foundry (deflationary curves), defect-density (D0) learning curves, packaging + CoWoS-interposer costs, test cost, backend yield, HBM $/GB. Per-source dated observations (as_of = date the reading reflects; curated backfill extends history) or month/quarter median/min/max rollups via grain=.

Query Parameters

entity_idOptional: entity slug (e.g. tsmc-n5, cowos, cowos-interposer, gpu-ai)

foundryOptional: foundry slug (tsmc, samsung, intel)

fromInclusive as_of start YYYY-MM-DD (optional)

toInclusive as_of end YYYY-MM-DD (optional)

grainraw (per-source observations, default) | month | quarter (median/min/max rollups)

limitMax observations scanned (1–2000, default 730)

Example Response

{
  "success": true,
  "data": [{
    "benchmark_type": "wafer_cost", "entity_id": "tsmc-n5", "foundry": "tsmc",
    "as_of": "2026-06-30", "value_num": 16000, "unit": "usd_per_wafer",
    "source_name": "SemiAnalysis", "source_url": "https://...",
    "confidence_tier": "high"
  }],
  "meta": { "grain": "raw", "count": 1 }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/chip-historyProChip Cost History (Pro)

Historical manufacturing cost data for a specific AI accelerator. Track cost trends over time.

Query Parameters

chipChip ID (required, e.g. h100-sxm5)

fromStart date YYYY-MM (optional, default: 12 months ago)

toEnd date YYYY-MM (optional, default: now)

Example Response

{
  "success": true,
  "data": { "chip": "h100-sxm5", "history": [
    { "period": "2026-06", "estMfgCostUsd": 3320, "hbmCostUsd": 1450,
      "logicDieCostUsd": 1180, "packagingCostUsd": 480, "testCostUsd": 210 }
  ] },
  "meta": { "count": 1, "from": "2025-08", "to": "2026-08" }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/wafer-trendsProWafer Price Trends (Pro)

Historical wafer pricing trends for a process node. Track foundry pricing over time.

Query Parameters

nodeNode ID (required, e.g. tsmc-n3)

fromStart date YYYY-MM (optional)

toEnd date YYYY-MM (optional)

Example Response

{
  "success": true,
  "data": { "node": "tsmc-n3", "history": [
    { "period": "2026-06", "avg": 20000, "min": 18500, "max": 21500,
      "confidence_tier": "high", "source_url": "https://..." }
  ] },
  "meta": { "count": 1 }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

GET/api/v1/margin-trendsProGross Margin Trends (Pro)

Historical gross margin data for semiconductor categories. Track quarterly margin trends.

Query Parameters

categoryCategory (required: gpu-ai, mobile-soc, mainstream-cpu, automotive-mcu, fpga)

fromStart date YYYY-MM (optional)

toEnd date YYYY-MM (optional)

Example Response

{
  "success": true,
  "data": { "category": "gpu-ai", "history": [
    { "period": "2026Q2", "grossMarginPct": 74.5,
      "source_name": "NVIDIA Q2 FY2027", "source_url": "https://..." }
  ] },
  "meta": { "count": 1 }
}

Cache: 5 minutes (private) · Rate limit: Pro subscription required · 10k req/hr

Auth: Required — Pro subscription

Pull into Excel or Google Sheets

Every market series returns flat CSV with ?format=csv, and every table on our /data pages is real HTML. Nothing to install, no add-on, at any access level.

Google Sheets — no key

Paste into a cell; it refreshes on its own.

=IMPORTDATA("https://siliconanalysts.com/api/v1/market-data/hbm-pricing?format=csv")

Not sure which series? This returns the whole catalog — ids, names, units — as rows:

=IMPORTDATA("https://siliconanalysts.com/api/v1/market-data/list?format=csv")

Foundry allocation and lead times work the same way:

=IMPORTDATA("https://siliconanalysts.com/api/v1/allocation?format=csv")

Sheets caches IMPORTDATA for about an hour, so a cell will not update the instant our data does.

Excel — no key

The formula above is Sheets-only — Excel answers #NAME?. Use Data → From Web and paste the URL itself; no Advanced, no headers.

https://siliconanalysts.com/api/v1/market-data/hbm-pricing?format=csv

Access	How	Depth
No key	Formula in a cell / Data → From Web	Current values · 200 pulls per day
Free key	Power Query or Apps Script (header auth)	+ the 3 most recent historical points per series
Pro key	Same Power Query / Apps Script, Pro key	Full series; /api/v1/export?format=csv for bulk pulls

Spreadsheet pulls run in their own rate-limit bucket — 200 per day, separate from the per-IP API budget. Authentication is header-based, which is why the no-key formulas above return current values only: Google Sheets cannot send a header from a cell. A clamped CSV looks identical to a full one in a spreadsheet, so if you need the complete series, use one of the keyed paths below rather than checking row counts.

Excel with a key (Power Query)

Data → Get Data → From Web → Advanced, add an x-api-key header — or paste this straight into the Power Query editor:

let
  Source = Json.Document(Web.Contents(
    "https://siliconanalysts.com/api/v1/market-data/hbm-pricing",
    [Headers=[#"x-api-key"="sa_live_YOUR_KEY"]])),
  Points = Table.FromRecords(Source[data][dataPoints])
in
  Points

Google Sheets with a key (Apps Script)

=IMPORTDATA cannot send headers, so reaching history from Sheets takes a small Apps Script (Extensions → Apps Script). Its columns match the CSV header exactly, so formulas keep pointing at the same fields if you later switch paths.

function SILICON_ANALYSTS(path) {
  const res = UrlFetchApp.fetch('https://siliconanalysts.com' + path,
    { headers: { 'x-api-key': 'sa_live_YOUR_KEY' } });
  const pts = JSON.parse(res.getContentText()).data.dataPoints;
  const cols = ['dataset_id','period_label','period_sort_key','series_key','series_label','value_low','value_mid','value_high','value_label','unit','confidence','data_type','is_projection','is_recent','source_name','source_date','source_note','dataset_source_note','basis','basis_note'];
  return [cols].concat(pts.map(p => cols.map(c => p[c] ?? '')));
}
// In a cell: =SILICON_ANALYSTS("/api/v1/market-data/hbm-pricing")

Bulk pulls: /api/v1/export?dataset=capacity_signals&format=csv downloads full dated history as RFC-4180 CSV (10k-row pages via from/to). It is a file download rather than a live cell, so use it for archives and the paths above for anything that should refresh. Refresh incrementally with /api/v1/changes?since= instead of re-pulling everything.

Authentication & Rate Limits

All endpoints accept API keys for higher rate limits and authenticated usage tracking. Anonymous IP-based access works for real use of current data (20 GET req/day, 10 POST calculate req/day). Historical / time-series data requires a free API key — a key raises limits, unlocks a history preview, and attributes usage to your account.

Generate a free API key below for higher rate limits (100 req/day) and a history preview. Pro keys unlock 10,000 req/hr and the full time series. Include your key in requests via header:

# API key via X-API-Key header
curl -H "X-API-Key: sa_live_abc123..." https://siliconanalysts.com/api/v1/accelerators

# Or via Authorization header
curl -H "Authorization: Bearer sa_live_abc123..." https://siliconanalysts.com/api/v1/hbm

Tier	GET Endpoints	POST Calculate	History depth	Auth
Anonymous (no key)	20 req/day	10 req/day	Current only	None required
Free API key	100 req/day	20 req/day	Preview (latest ~3 pts / 30d)	X-API-Key header
Pro API key	10,000 req/hr	10,000 req/hr	Full series + export	X-API-Key header

Your API Keys

API Playground

Try any endpoint live. No authentication required for read endpoints.

Try any of 26 endpoints with live data — no API key required for public endpoints.

Endpoint

vendor (optional)

chip (optional)

fields (optional)

cURL

curl "https://siliconanalysts.com/api/v1/accelerators"

Need higher rate limits, full history, or bulk data access?

Pro API access is coming soon — higher rate limits, full history depth on every series, and bulk export/snapshot replay. Join the waitlist and we'll reach out before launch.

AI Agent Discovery

Silicon Analysts is optimized for discovery by AI agents and LLM-powered search.

/llms.txt — Machine-readable site description for LLM agents

/api/v1 — JSON API manifest for programmatic endpoint discovery

/api/v1/openapi.json — OpenAPI 3.1 description of all 36 endpoints

JSON-LD structured data on every page (Dataset, SoftwareApplication, Article)

/sitemap.xml — Full sitemap with all tools and articles

Trust & Reliability

Provenance

Data Quality & Provenance

Per-record provenance taxonomy, source-type and confidence-tier reference, update cadence per dataset.

Status

System Status ↗

Live uptime + 90-day history for /api/v1, /api/mcp, web. Hosted on Better Stack.

Data Sources

Cost estimates are derived from Epoch AI Monte Carlo models, Raymond James semiconductor research, TrendForce quarterly reports, and SemiAnalysis teardown data.

HBM market data is sourced from SK Hynix, Samsung, and Micron investor relations, TrendForce, and proprietary supply chain tracking models.

Estimates are directional and may vary ±15-20% from actual manufacturing costs. Per-record provenance, source-type taxonomy, and confidence tiers are documented at /data-quality. For full terms see Terms & Data Provenance.

Explore Interactive Tools

Use our interactive tools to model chip costs, compare accelerators, and analyze markets

Cost Bridge Chart Chip Price Calculator HBM Market Analysis