# Fungible Financial Model

Fungible is building a Data Processing Unit (DPU) - a new class of programmable silicon - to make data centers data-centric rather than compute-centric.

- Canonical: https://finamodel.com/startups/fungible
- Excel download: https://finamodel.com/startup-models/fungible.xlsx
- Category: Hardware/Deep-tech
- Model type: SaaS ARR / Valuation
- Funding round: Series C
- Funding: $200M
- Founded: 2019
- Geography: Not in deck. Implied global addressable market (cloud providers, hyperscalers).
- Customer: B2B

## About the company

Fungible builds programmable data-processing units intended to make data centres more data-centric. Its DPU silicon addresses infrastructure bottlenecks that cannot be solved efficiently by traditional compute processing alone, targeting customers and OEMs building high-performance storage and networking architectures.

As a semiconductor company, Fungible’s commercial inflection point is a qualified design win followed by an OEM or customer shipment ramp. The model must account for long qualification cycles, foundry and packaging commitments, inventory exposure, and heavy R&D before volume revenue arrives.

Model design wins, qualification timing, OEM partners, chips shipped, ASP, wafer, packaging, test, and inventory costs. Include architecture R&D, applications engineering, sales, foundry commitments, and support. Win rate, time to production, shipment ramp, pricing, gross margin, customer concentration, supply yield, and inventory turns should drive scenarios.

## 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: A fully programmable DPU chip designed to offload and accelerate data-centric workloads (networking, storage I/O, security, disaggregation) from the CPU.
- Core claim: DPUs execute data-centric workloads far more efficiently than CPUs. They enable disaggregation and pooling of compute and storage at scale.
- Value prop dimensions: Economics, Security, Reliability, Performance, Agility.
- Insertion angle: "No changes in application code / no changes to network / no changes to server design" - zero-disruption deployment.
- Architecture shift enabled: Moves data center networking from complex multi-tier topology to a flat, efficient fabric supporting resource pooling at scale.

## Market

- No explicit TAM/SAM/SOM figures in deck.
- Market framing (qualitative): All data centers globally - server, data center building, and global network layers.
- Penetration claims (qualitative):
  - "90% of servers in 5 years" will be data-centric architecture.
  - "90% of buildings in 5 years."
  - "90% of regions in 5 years."
- Technology growth rates shown on chart:
  - Storage performance: 2x/year
  - Network performance: 2x/1.5 years (one line) and 2x/2 years (second line, older era)
  - Compute: 2x/2 years
- No dollar market size stated.

## Competition / moat

- Moat narrative (implied, not explicitly stated): First-mover in a new silicon category (DPU as a distinct microprocessor class). Programmability vs. fixed ASICs. No-disruption insertion (no app/network/server changes required).
- Competitors not named in deck.
- Positioning: DPU sits alongside CPU, GPU, FPGA in the heterogeneous scale-out stack - not replacing CPU, but handling the data-centric layer.

## Team & funding ask / use of funds

---

## Recommended financial model

- **Archetype + why:** Fabless semiconductor revenue model - unit shipment × ASP, by customer segment (hyperscaler / cloud / enterprise DC). This is a hardware IP company; the right model tracks tape-out milestones, design win pipeline, unit ramp, COGS (wafer cost + test + packaging), and gross margin trajectory. If software/firmware licensing emerges later, add a recurring software attach rate layer.

- **Forecast horizon & granularity:**
  - Pre-revenue phase: Monthly, 18–24 months (R&D burn, headcount, tape-out capex)
  - Revenue ramp phase: Quarterly, Years 1–5
  - Total horizon: 5 years

- **Key drivers & assumptions:**

| Driver | Value / Range | Source |
| -- | -- | -- |
| Target market: servers with DPU | 90% of servers in 5 years | - |
| Global server shipments (annual) | ~13–15M units/yr (hyperscaler + enterprise) | industry estimates; not in deck |
| DPU attach rate (% of servers) | 5% Y1 → 40% Y5 | early ramp; analogous to GPU adoption curve |
| ASP per DPU | $500–$2,000 | comparable to SmartNIC/DPU market (Nvidia BlueField, Pensando); not in deck |
| COGS as % of ASP | 40–50% | typical fabless gross margin target of 50–60% |
| R&D headcount burn (pre-revenue) | $5–15M/yr | deep silicon team; no data in deck |
| Tape-out cost (7nm–5nm node) | $20–50M per tape-out | industry norm for advanced nodes; not in deck |
| Time to first silicon | 18–24 months from funding | - |
| Time to volume production | 36–48 months | - |
| Storage performance improvement | 2x/year | - |
| Network performance improvement | 2x/1.5 years | - |

- **Scenarios (Base / Bull / Bear - which variables flex):**
  - **Bear:** Slow hyperscaler adoption; DPU attach rate stays <10% by Year 5; ASP pressure from incumbents (Nvidia/Marvell); model shows cash burn without profitability in 5-year window.
  - **Base:** Attach rate reaches 25–30% by Year 5; ASP holds at $800–$1,200; one or two hyperscaler design wins by Year 3.
  - **Bull:** Category inflection similar to GPU; 50%+ attach rate; premium ASP for programmability; software licensing adds 15–20% revenue on top.
  - **Flex variables:** DPU attach rate, ASP, time-to-first-design-win, tape-out timeline slippage, node cost.

- **Required sheets / outputs:**
  1. **Assumptions** - all drivers, clearly tagged vs
  2. **R&D & Headcount Plan** - engineering ramp, tape-out milestones, opex burn
  3. **Revenue Model** - unit shipments × ASP by segment (hyperscaler / cloud / enterprise), by quarter
  4. **COGS & Gross Margin** - wafer cost, yield, packaging, test; GM% trajectory
  5. **P&L (Income Statement)** - Revenue → Gross Profit → Opex (R&D + S&M + G&A) → EBITDA → Net Income
  6. **Cash Flow & Runway** - capex (tape-out), working capital, funding milestones
  7. **Balance Sheet** - simplified (cash, IP/intangibles, debt/equity)
  8. **Scenario Toggle** - Bear / Base / Bull switcher
  9. **Sensitivity Table** - ASP × attach rate → revenue and GM at Year 3 and Year 5
  10. **Dashboard** - KPI cards: runway, unit shipments, ASP, gross margin %, cumulative cash burn

## Frequently asked questions

### Is the Fungible financial model free?

Yes. The Fungible 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.