Key Takeaways
- The semiconductor sector, once a pariah for venture capitalists 15-20 years ago, has exploded back into favor, fueled almost entirely by the insatiable demands of AI. Lip Bu Tan, a legendary investor, recalls a time when “no VC want to join me” in these deals.
- The core of Tan's investment thesis boils down to solving actual problems: “I always look at where is the bottleneck what are you trying to solve” and specifically, if a “customer crying for it.”
- Forget generic pilots. Your first customer must be hyper-scale. These are the giants “willing to pay million of dollars next few years and even giving some warrant” because their buy-in alone provides immense validation and scale.
- Opportunities are emerging in unexpected places beyond just compute chips, including interconnects, optical solutions, AI-driven design (EDA), new materials like gallium nitride, and the frontier of "physical AI."
- Lip Bu Tan's success in this high-stakes arena is distilled into a concise checklist for identifying and capitalizing on the next wave of deep tech innovation.
The Lip Bu Tan's Semiconductor Investment Thesis Checklist
Here’s how venture legend Lip Bu Tan approaches semiconductor investments in the new AI era, focusing on the core problems, market dynamics, and technological frontiers:
- 1. Identify Bottlenecks & Problems: Look at where is the bottleneck, what are you trying to solve. What is the problem we try to solve, is it real, is customer crying for it? (Examples: interconnect, optical solutions, power/thermal management.)
- 2. Explore AI/ML for Design: Can you find some using AI machine learning to drive better design and better solution? (Opportunities in EDA related areas.)
- 3. Invest in New Materials: Look at the new material... like gallium nitride and then silicon carbide and then some of the company starting to being acquired... (Investment in companies like Infi for indium phosphide, and those working on power management ICs like Empower.)
- 4. Target Hyper-Scale First Customers: It's very important from day one you'll have to target the first customer. Usually I like the customer is hyper the skill. If they like what you have, they're willing to pay million of dollars next few years and even giving some warrant is worth it because you have a big one customer you can scale.
- 5. Secure Top Talent: Find the talent... in US and then Silicon Valley and then some Austin and then the other part is Israel a lot of talent so I back quite a few quite a significant amount my investment in Israel.
- 6. Pursue Full Stack Solutions & New Frontiers: Look at the full stack... physical AI next a mix big frontier.
When This Works (and When It Doesn't)
This thesis works best when focusing on real problems that customers are actively seeking solutions for, partnering with large, hyper-scale clients for early traction, and leveraging cutting-edge advancements in AI and material science. It’s tailor-made for founders building capital-intensive, deep tech solutions where incremental gains can unlock massive value for tech giants.
However, this approach isn't a silver bullet. Securing a hyper-scale first customer, by definition, is incredibly hard for an unproven startup. It requires a long sales cycle, deep trust, and often significant upfront investment without guaranteed returns. This thesis also demands specialized technical talent and an understanding of highly complex engineering challenges, making it less suitable for purely software-focused startups or those targeting smaller, fragmented markets. If your solution isn't addressing a multi-million dollar problem for a tech behemoth, Tan's checklist might be too ambitious.
What to Do With This
If you're building a semiconductor or deep tech company in the AI era, pull up Tan's checklist. This week, pick one key component of your product or service and apply it:
Scenario: You’re developing a novel optical interconnect solution for data centers.
1. Identify Bottlenecks: Can you articulate, in one sentence, the exact problem your interconnect solves for a hyperscaler like Google or Microsoft? Is it bandwidth, power consumption, latency, or cost? How do you know they're actively "crying for it"?
2. Target Hyper-Scale First Customers: Draft a list of your top 3 dream customers. Who are the specific individuals at those companies (e.g., Head of Data Center Infrastructure at AWS) that you need to engage? What's your unique value proposition that makes them willing to pay "millions of dollars" and offer warrants?
3. Explore New Materials: Are you leveraging beyond standard silicon photonics? Have you explored options like indium phosphide or other emerging materials that offer a step-function improvement, not just marginal gains? If not, investigate how new materials could differentiate your offering.