Walter Goodwin on Why Fractile Bypasses Broadcom
Fractile CEO Walter Goodwin explains why outsourcing physical chip design to Broadcom traps AI hardware startups in identical supply queues.
The $7B Startup Building AI Chips to Challenge Nvidia | Fractile Founder & CEO, Walter Goodwin
With Harry Stebbings, Walter Goodwin · Saturday, November 14, 2026
Walter Goodwin, founder and CEO of AI chipmaker Fractile, breaks down how memory bandwidth bottlenecks limit AI model inference and why 3D stacking DRAM atop logic dies provides a superior architectural alternative to SRAM or standard HBM chips. Goodwin also discusses Fractile's decision to bypass Broadcom, the geopolitical energy crunch facing Western data centers compared to China, and why chip heterogeneity risks creating stranded capacity.
Key takeaways
- Hyperscalers and startups alike, including Google for TPU, Meta for MTIA (Jalapeno), MatX, and Etched, rely on Broadcom to handle backend physical design and foundry execution. Read more →
- The primary ceiling on AI expansion has shifted from GPU allocation and wafer supply to grid capacity, physical land, and substation hookups. Read more →
- Neocloud GPU hosting is commoditizing rapidly because competent hosting providers offer almost zero technical differentiation once clusters are online. Read more →
- Data centers that pair GPUs like Nvidia Vera Rubin with SRAM chips like Groq face severe utilization mismatches when AI workloads shift. Read more →
- Model training is shifting into a fixed upfront cost, making inference and test-time compute the primary operational expense of frontier intelligence. Read more →
- Compute power scaled 1,000,000-fold over the last 25 years, while memory bandwidth improved only 40-fold, creating a 25,000-fold bottleneck. Read more →
6 articles from this episode
Walter Goodwin on Why Power Caps Will Clear Out Old AI Chips
Fractile CEO Walter Goodwin explains why data center power constraints will force operators to dump current AI silicon within six or seven years.
Where Value Settles in the AI Stack
Fractile CEO Walter Goodwin explains why neoclouds are commoditizing while custom silicon and model inference co-design capture the profits.
Why Mixed AI Chip Fleets Will Strand Capital
Fractile CEO Walter Goodwin explains why pairing GPUs with SRAM chips creates stranded data center capacity as context lengths grow.
Why AI Silicon Breaks on Long-Running Swarms
Fractile CEO Walter Goodwin explains why test-time compute scaling and multi-day agent swarms break both SRAM and HBM chips.
Walter Goodwin Says Memory Bandwidth Lags Compute by 25,000x
Fractile CEO Walter Goodwin explains why memory bandwidth bottlenecks AI inference and how vertical 3D DRAM stacking solves it.
More 20VC with Harry Stebbings episodes
- Mistral CEO: Why OpenAI & Anthropic Will Come for Your Business | Are Chinese Models Truly Safe?Nov 15 · 5 articles
- Sequoia Leaves London, Close Scout Program & End Evergreen Fund | NASDAQ Loses $2.5T in 24 hoursNov 15 · 6 articles
- Modal CEO: Has Open-Source LLM Inference Peaked? | What Happens to the Cost of ComputeNov 8 · 4 articles
- Google Loses $500B in 12 Hours | DoorDash Buys Instacart for $17B | Kalshi Raises $1BNov 8 · 6 articles