# Recogni Financial Model

Recogni builds a custom ASIC (VPU-1000) delivering 1,000 TOPS at 5W - the first chip to simultaneously optimise for both AI inference performance and power efficiency for Level 3+ autonomous vehicles.

- Canonical: https://finamodel.com/startups/recogni
- Excel download: https://finamodel.com/startup-models/recogni.xlsx
- Category: AI/ML
- Model type: Unit-economics / DTC
- Funding round: Series A
- Funding: $25M
- Founded: 2019
- Geography: Not in deck (HQ not mentioned; team backgrounds suggest US/Europe).
- Customer: B2C

## About the company

Recogni designs the VPU-1000, a custom AI-inference ASIC targeting Level 3-plus autonomous vehicles. Its Vision Computing Module combines the chip with sensor fusion, passive cooling, night vision, and a claimed 1,000 TOPS at five watts.

The fabless company sells into long automotive OEM and Tier-1 qualification cycles. It reports ongoing technical engagements and commitments to invest, plus seven provisional patents, but no contracts, shipments, revenue, or unit prices; production revenue follows design wins after a multi-year lag.

The model follows OEM pipeline, design wins, qualification milestones, vehicle production ramp, module ASP, and units shipped. It separates non-recurring engineering and R&D from manufacturing COGS, then tests yield, gross margin, working capital, cash burn, funding, and runway.

## What's included

- 5-year monthly revenue build with stage-appropriate growth assumptions
- Full P&L, headcount plan, and operating-expense schedule
- Cash-flow statement, runway, and burn-rate tracking
- Valuation via exit multiple with a DCF cross-check
- Returns analysis with MOIC and IRR
- Unit economics including CAC, LTV, payback, and cohort retention

## Product & value proposition

- Product: Recogni VCM (Vision Computing Module) containing 1x VPU-1000 ASIC plus sensor fusion (infrared, monochromatic, colour, external depth cameras).
- Key specs: 1,000 TOPS at 5 Watts, passive cooling, 120 dB dynamic range, night vision.
- Competitive claim: <100W to process 95% of all AI workload for an entire vehicle; competition requires >26KW to achieve equivalent L3+ performance (7,000+ TOPS current SoA).
- Efficiency claim: >200× more power-efficient than other accelerators.
- Patent protection: 7+ provisional patents filed.
- Technology readiness: Ongoing technical engagements and "commitment to invest" from several automotive OEMs and a Tier-1 parts supplier - no signed contracts stated.

## Market

- Level 2+ ADAS market (addressable by existing solutions): $2B (2022) → $4B (2024).
- Level 3+ AV AI processing market (Recogni's primary TAM): $16B (2025) → $45B (2030).
  - Sub-segments at $16B 2025 level: Self-driving Cars $6B, Ride-hailing Taxis $10B.
- Claim: Recogni technology could also address Level 2+ market ($4B by 2024).
- Market acceleration claim: Recogni will accelerate the AV market timeline by at least 2 years.

## Revenue model

- Inferred go-to-market: direct sales to automotive OEMs and Tier-1 parts suppliers (based on mention of "ongoing technical engagements").
- Hardware unit (VCM module) appears to be the initial product; potential for software/IP licensing not mentioned.

## Traction & metrics

- No revenue, customer count, contract values, or unit shipment figures disclosed.
- Qualitative traction: "Ongoing technical engagements and commitment to invest from several automotive OEMs and Tier-1 parts supplier."
- Patent progress: 7+ provisional patents filed.
- No ARR, MRR, growth rates, or retention metrics in deck.

## Competition / moat

- Competitors shown on performance vs. efficiency scatter plot:
| Company | Performance | Power |
| -------- | ----------- | ------ |
| Hailo | 40 TOPS | 15W |
| (unnamed, likely Qualcomm/NXP) | 30 TOPS | 13W |
| (unnamed) | 24 TOPS | 10W |
| (unnamed) | 15 TOPS | 5W |
| (unnamed) | 15 TOPS | 3W |
| Syntiant | 5 TOPS | 1W |
| (unnamed) | 4 TOPS | 2W |
| **Recogni VPU-1000** | **1,000 TOPS** | **5W** |
- All competitors plotted as "Inference Only"; Recogni claims full-stack ASIC.
- Moat: custom silicon (ASIC vs. GPU/FPGA), patent portfolio, claimed 2-year market timing advantage.

## Team & funding ask / use of funds

- Founding team:
  - RK Anand (CEO) - prior: OttoQ (Founder/CEO), Kumu (President/CEO), Xsigo (Founder/VP Eng), Juniper (EVP/GM)
  - Eugene Feinberg (Technology/CTO) - prior: mPerpetuo (Founder/CTO), EyeFi (Founder/Architect), Cisco, Growth Networks
  - Ashwini Choudhary (Products, Marketing, BD) - prior: mPerpetuo (Founder/COO), Ventiva (CEO), NetFortis (Founder/CEO), Aarohi Comm
  - Gilles Backhus (AI) - prior: Lilium (Sensor Systems), Konux (AI Lead), Kumu, Gauss (ML Scientist)
  - Valerie Chan (Operations) - prior: Forq (Founder/CEO), EyeFi (VP Ops), Warpia (VP Biz Dev), Faircom (VP Sales)

---

## Recommended financial model

- **Archetype + why:** Hardware / semiconductor revenue model - specifically a **chip unit-economics + design-win pipeline model**. Recogni is a pre-revenue fabless ASIC company selling into automotive OEM qualification cycles. The right model tracks design-win pipeline → production ramp → per-unit ASP × volume × gross margin. Not a SaaS ARR or 3-statement operating model at this stage - the business is capex-light (fabless) but heavily R&D-weighted with a 3–5 year design-win-to-revenue lag typical of automotive silicon.

- **Forecast horizon & granularity:** 7 years (2019–2026), annual. Years 1–2 = R&D/pre-production (burn only); Year 3 = first design-win revenue; Years 4–7 = production ramp. Monthly granularity for cash runway in Year 1–2.

- **Key drivers & assumptions:**
  - *Market size* - Level 3+ AI processing TAM: $16B in 2025, $45B in 2030; CAGR ~23% implied
  - *Recogni market share* - 1–3% of L3+ TAM by 2030; automotive design cycles are long and OEM qualification is a major gating factor
  - *ASP per VCM unit* - $500–$1,500 per module (automotive AI compute modules typically $200–$2,000 depending on performance tier; no deck data)
  - *Units per vehicle* - 1–2 VCM modules per L3+ vehicle; L3+ minimal config = 8 cameras, 5 radar, 4 LiDAR
  - *Gross margin* - 40–55% at scale (fabless semiconductor typical range); lower in early ramp due to NRE recovery and yield
  - *NRE (Non-Recurring Engineering) revenue* - $1–5M per major OEM engagement in design phase (common in automotive silicon)
  - *R&D opex* - largest cost line pre-revenue; tape-out costs for automotive-grade ASIC typically $5–30M
  - *Design-win timeline* - 18–36 months from engagement to production sign-off (standard automotive)
  - *Number of OEM design wins* - 2–3 initial OEM engagements based on "several OEMs"; convert 1 in base case by Year 4
  - *Revenue recognition* - upon module delivery (hardware), or milestone-based for NRE contracts
  - *Headcount ramp* - 20–40 engineers by end of Year 2; Silicon Valley/automotive comp rates
  - *Burn rate* - $3–6M/year pre-production; highly R&D intensive
  - *Tape-out / foundry costs* - capitalized or expensed depending on accounting treatment; major cash event

- **Scenarios (Base / Bull / Bear - which variables flex):**
  - *Base:* 1 OEM design win by 2023, production ramp to ~50K units/year by 2026, 2% L3+ market share
  - *Bull:* 2–3 OEM wins, faster Level 3+ market adoption, ASP at high end, NRE revenue from multiple Tier-1s
  - *Bear:* No production win before 2025 (automotive qualification delays), L3+ market slower than projected, competitor catches up on efficiency, cash runway exhausted requiring bridge

- **Required sheets / outputs:**
  1. Assumptions dashboard (all drivers toggled by scenario)
  2. Design-win pipeline tracker (OEM engagement → NRE → production → volume ramp per customer)
  3. Revenue build: NRE revenue + unit volume × ASP by OEM
  4. P&L: Revenue, COGS (per-unit wafer + assembly + test), Gross Profit, R&D, S&M, G&A, EBITDA
  5. Cash burn & runway (monthly for Years 1–2; quarterly thereafter)
  6. Headcount plan
  7. Market context tab: TAM bridge from $16B (2025) to $45B (2030) with Recogni share overlay
  8. Sensitivity table: ASP × Volume, Margin × Design-win timing

## Frequently asked questions

### Is the Recogni financial model free?

Yes. The Recogni model is a free Excel download with live formulas.

### Can I change the assumptions?

Yes. The workbook is editable and its live formulas recalculate when assumptions change.