Xingji Shidai Chip Unit \China

Step 2 - Tape-Out and Automotive Qualification (Validation, 18 months, $30M): Complete the ASIC design using Cadence or Synopsys EDA tools, integrating RISC-V CPU cores, custom TPU, and Imagination GPU cores via chiplet architecture. Submit tape-out to TSMC 7nm process (or SMIC 14nm for China-only version) with $15M NRE cost. Spend 12 months on ISO 26262 ASIL-D functional safety certification and AEC-Q100 automotive qualification testing (temperature cycling, vibration, humidity). Deliver 10K engineering samples to pilot customers at $500 per chip, targeting deployment in 5K-10K vehicles by late 2026. Raise $50M Series A from strategic investors (Geely, BYD, or Bosch) by demonstrating production-ready silicon and $5M in pre-orders. The key metric is proving the chip works in safety-critical applications without field failures, which requires exhaustive testing but is the only path to automotive OEM trust.

Step 3 - Volume Manufacturing and Software Moat (Growth, 24 months, $100M): Scale manufacturing to 500K chips annually via TSMC or SMIC, reducing per-unit cost to $150 through volume discounts and yield improvements. Expand the software ecosystem by open-sourcing the compiler toolchain on GitHub, hosting quarterly developer conferences, and building a model zoo with 50+ pre-optimized AI models for autonomous driving (object detection, semantic segmentation, path planning, sensor fusion). Sign 10+ automotive OEM customers (Chinese EV makers, Japanese Tier 1 suppliers, European startups) with multi-year supply agreements totaling $200M in revenue by 2028. The moat is software lock-in: once a customer has deployed your chip and trained their engineering team on your toolchain, switching costs are prohibitive (6-12 months to re-validate a new chip, $5M+ in engineering effort). Raise $150M Series B from growth equity firms (Tiger Global, Coatue) at $1B+ valuation based on $50M ARR and 70% gross margins.

Step 4 - Licensing Model and Platform Expansion (Moat, 36+ months): Transition from pure chip sales to a hybrid model: continue selling chips to automotive customers at $150-300 per unit, but also license the chip design and software stack to adjacent markets (drones, AR glasses, smart cameras, industrial robotics) for $5M upfront plus $10-20 per-unit royalties. This ARM-style licensing model reduces capital intensity (customers fund their own manufacturing) while expanding TAM to $50B+ across all edge AI applications. Invest $50M in next-generation chip design using 3nm process and integrating on-chip DRAM for 50% faster inference. The endgame is becoming the de facto standard for edge AI inference, with 100+ customers shipping 10M+ chips annually by 2030, generating $500M in chip revenue plus $200M in licensing royalties, and achieving a $5B+ valuation at IPO.