Republic opens private markets as Positron takes on GPUs: A TWiST500 doubleheader! | E2176
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
This episode features two interviews showcasing the future of democratized finance and energy-efficient AI. First, Kendrick Nguyen, co-CEO of Republic, discusses how the company has evolved from equity crowdfunding to building comprehensive financial infrastructure for tokenization and alternative investments. (03:23) Republic now offers everything from retail equity crowdfunding to institutional venture capital, secondary trading, and real-world asset tokenization. Second, Mitesh Agarwal, CEO of Positron, explains how his company is developing specialized chips for AI inference that are far more energy-efficient than traditional GPUs. (26:23) While NVIDIA GPUs excel at training AI models, Positron's chips can achieve 93% memory bandwidth utilization compared to NVIDIA's 25-40% for inference workloads.
- Main Theme: Both companies represent the democratization of traditionally exclusive markets - Republic making private investments accessible to retail investors, and Positron making AI compute more efficient and potentially more affordable through specialized hardware designed specifically for inference rather than training.
Speakers
Kendrick Nguyen
Co-CEO of Republic, Kendrick previously worked at AngelList where he gained extensive knowledge of the venture capital ecosystem. He has been instrumental in building Republic's comprehensive financial infrastructure platform since its founding in 2016. Under his leadership, Republic has raised over $200 million, including a notable $150 million round in 2021 led by Valor Equity Partners.
Mitesh Agarwal
CEO of Positron, Mitesh is leading the development of specialized AI inference chips that promise significant improvements in energy efficiency over traditional GPUs. His company has raised $75 million and has already shipped thousands of their Atlas FPGA-based systems to customers including Cloudflare and Paracel, with their ASIC-based Asimov system planned for 2026.
Key Takeaways
Financial Infrastructure is the Future of Investing
Republic is building what Kendrick calls "the AWS for investing" - a comprehensive financial infrastructure that handles everything from primary issuance to secondary trading, tokenization, and cross-border transactions. (24:17) Rather than competing with traditional venture firms, Republic is creating the underlying technology that enables fractionalized ownership, retail participation, and liquidity in previously inaccessible markets. This approach allows them to serve multiple customer segments through a single integrated platform while generating revenue from various business lines.
Specialized AI Hardware Will Challenge NVIDIA's Dominance in Inference
While NVIDIA GPUs are excellent for AI training, they're inefficient for inference workloads, utilizing only 25-40% of their theoretical memory bandwidth. (33:04) Positron's specialized chips achieve 93% bandwidth utilization by optimizing for memory capacity and bandwidth rather than raw compute power. This represents a fundamental shift where different AI workloads will require different hardware solutions, creating opportunities for specialized chip companies to capture significant market share in the growing inference market.
Proof-of-Concept Products Accelerate Market Validation
Positron's strategy of launching Atlas (FPGA-based) before Asimov (ASIC-based) demonstrates the value of getting real-world feedback quickly. (37:54) By shipping functional systems within 18 months and gathering customer feedback from production deployments, they can iterate on their architecture while generating revenue. This approach reduces risk for both the company and investors by proving market demand before committing to expensive custom silicon development.
Regulatory Evolution Enables Market Democratization
The increase in equity crowdfunding caps from $1 million to $5 million in 2020 brought later-stage companies into the retail investment space. (07:11) Kendrick expects these caps to continue rising, potentially to $100 million, which would further democratize access to private markets. This regulatory evolution reflects a broader shift toward allowing retail investors to participate in opportunities previously reserved for accredited investors and institutions.
Capital Efficiency Drives Competitive Advantage
Positron achieved remarkable capital efficiency by taping out their first generation chip with only $75-80 million raised, compared to traditional silicon companies that typically require $300-400 million. (47:05) This efficiency comes from their strategic approach of proving concepts with FPGAs first, focusing on specific use cases where they can outperform incumbents, and maintaining a clear path to profitability through direct hardware sales rather than building their own cloud infrastructure.
Statistics & Facts
- Republic has raised more than $200 million total, including a $150 million round in late 2021 led by Valor Equity Partners. (02:58)
- AI inference compute spending is projected to grow from $105 billion in 2024 to $300-400 billion by 2028, while training spend is expected to reach $500-600 billion by 2028. (43:13)
- Positron's Asimov chip will have 2 terabytes of memory capacity compared to NVIDIA's next-generation Rubin chip at 384 gigabytes - a 5x difference in memory capacity. (42:02)
