Etched: How Three Harvard Dropouts Are Challenging NVIDIA with Physics-Informed AI Chips
From Harvard Classrooms to Silicon Valley's Boldest Bet
In March 2023, three Harvard University students made a decision that would reshape the future of artificial intelligence hardware. Gavin Uberti, Chris Zhu, and Robert Wachen walked away from their mathematics and computer science degrees to pursue an audacious vision: building specialized chips that could outperform NVIDIA's dominance in AI computing.
Today, their company [Etched](https://www.etched.com/) stands at a $5 billion valuation with nearly $620 million in funding, representing one of the most remarkable university-to-unicorn journeys in deep tech history.
The Physics Behind the Innovation
While traditional GPU manufacturers design chips for general-purpose computing, Etched took a radically different approach rooted in physics-informed design principles. The team recognized that transformer models—the architecture powering ChatGPT, Claude, and virtually every major AI system—have specific computational patterns that can be hardwired into silicon.
Gavin Uberti, now CEO, explains the core insight: "We're building hardware for superintelligence by creating ASICs (Application-Specific Integrated Circuits) designed exclusively for transformer inference." This specialization delivers significantly higher throughput and lower cost compared to general-purpose GPUs, fundamentally changing the economics of AI deployment.
The technical breakthrough lies in optimizing for the matrix multiplication and attention mechanisms that define transformer architectures. By eliminating unnecessary computational flexibility, Etched's chips achieve performance gains that would be impossible with traditional GPU designs.
From Research to Reality: The Harvard Connection
All three founders were immersed in Harvard's rigorous computer science and mathematics programs when they identified the opportunity. [Gavin Uberti](https://www.linkedin.com/in/guberti), who leads the company as CEO, brought a unique combination of theoretical understanding and entrepreneurial drive. [Chris Zhu](https://www.linkedin.com/in/czhu1729), serving as CTO, contributed deep technical expertise in chip architecture and machine learning systems.
Their decision to drop out wasn't impulsive—it emerged from months of research demonstrating that specialized AI chips could achieve order-of-magnitude improvements over existing solutions. The Harvard ecosystem, with its proximity to cutting-edge AI research at institutions like MIT and access to Boston's deep tech investor network, provided the perfect launchpad.
Breaking the NVIDIA Monopoly
NVIDIA's H100 and A100 GPUs have become the de facto standard for AI training and inference, commanding premium prices and creating supply constraints across the industry. Etched's approach threatens this monopoly by offering purpose-built alternatives that excel specifically at transformer inference—the workload that matters most for deploying AI applications at scale.
The company's flagship product promises to deliver transformer inference at a fraction of the cost per token compared to GPU-based solutions. For companies running large language models in production, this cost reduction could mean the difference between profitable AI products and unsustainable burn rates.
Industry analysts note that Etched's success would validate a broader trend toward specialized AI accelerators. "The era of one-size-fits-all GPUs for AI is ending," observes one venture capital report. "Companies like Etched represent the next wave of hardware innovation."
Funding and Validation
The startup's trajectory has been nothing short of extraordinary. After raising $120 million in a Series A round in June 2024, Etched secured an additional $500 million in January 2026 at a $5 billion valuation. Investors include prominent names like [Primary Ventures](https://www.primaryventures.com/), Positive Sum Ventures, and legendary entrepreneur Peter Thiel.
This level of investment in a company less than four years old reflects both the massive market opportunity and the technical credibility of the founding team. The funding enables Etched to scale manufacturing, expand its engineering team, and accelerate product development to meet growing demand from AI companies.
Implications for University Entrepreneurship
Etched's story offers crucial lessons for university entrepreneurship centers and technology transfer offices. The founders didn't wait to complete their degrees—they recognized that timing in the AI hardware market was critical and moved decisively when the opportunity crystallized.
Their success demonstrates several key principles:
**Deep Technical Foundation**: The founders' Harvard education in mathematics and computer science provided the theoretical grounding necessary to identify and solve complex chip design challenges.
**Market Timing**: Understanding when to leave academia requires recognizing inflection points in technology markets. The explosion of transformer-based AI created a narrow window for specialized hardware companies.
**Risk-Taking Culture**: Harvard's entrepreneurial ecosystem, while perhaps less celebrated than Stanford's, nonetheless fostered the confidence to pursue ambitious technical ventures.
**Access to Capital**: Proximity to Boston and San Francisco venture capital networks enabled rapid fundraising once the technical vision was validated.
The Road Ahead: Physics AI and Beyond
As Etched scales production and begins delivering chips to customers, the company represents a broader trend toward physics-informed AI hardware. The next generation of AI systems will increasingly rely on specialized accelerators designed around specific model architectures and computational patterns.
For universities hosting the next generation of deep tech founders, Etched's trajectory offers a compelling case study in how fundamental research, technical education, and entrepreneurial ambition can converge to create transformative companies. The question isn't whether students should drop out—it's how universities can better support those who identify genuine market opportunities that require immediate action.
At GENF's London forum in July 2026, conversations about Physics AI and simulation-driven ventures will undoubtedly reference Etched as a prime example of how university research translates into world-changing companies. The startup's success validates the thesis that specialized, physics-informed approaches to AI hardware represent the future of the industry.
Conclusion
Etched's journey from Harvard dorm rooms to challenging NVIDIA's dominance exemplifies the power of university-based deep tech entrepreneurship. By combining rigorous technical education with bold vision and precise market timing, three young founders have positioned themselves at the forefront of AI hardware innovation.
For entrepreneurship center directors, faculty champions, and university administrators attending GENF, Etched offers a roadmap for nurturing the next generation of deep tech unicorns. The key ingredients—world-class technical education, access to cutting-edge research, entrepreneurial support systems, and connections to risk capital—remain as relevant as ever.
As AI continues its exponential growth trajectory, the demand for specialized hardware will only intensify. Companies like Etched, born from university research environments and built by technically sophisticated founders, will define the infrastructure powering the next era of artificial intelligence.
---
**About the Founders:**
- - **Gavin Uberti** (CEO): [LinkedIn](https://www.linkedin.com/in/guberti) | Harvard University
- **Chris Zhu** (CTO): [LinkedIn](https://www.linkedin.com/in/czhu1729) | Harvard University
- **Robert Wachen**: Harvard University
**Company**: [Etched](https://www.etched.com/) | Founded 2022 | Cupertino, California
**Funding**: $620M total | $5B valuation (January 2026)
---
*This article is part of GENF's University Spinoff Series, showcasing how academic research translates into transformative deep tech companies.*