미국 휴머노이드 기업 Figure AI 초기 투자자가 말하는 이 산업의 진실 | RoboStrategy, Andrew Kang

Andrew Kang highlights that Figure AI's live stream, which ran for 8 to 10 hours and extended to 8 days, demonstrated real capability rather than a curated highlight reel. The human intern won by a small margin, but his hands were blistered and exhausted, showing the physical toll. Kang estimates that humanoid robot intelligence will be good enough for most daily tasks within two to three years, marking a technological revolution that turns physical labor into an accessible product.

Andrew Kang introduces himself as CEO of RoboStrategy, a publicly listed venture fund on NASDAQ exclusively focused on robotics and physical AI. Their portfolio includes Figure AI, Uptronic, Dino Robotics, Centerbots, and Path Robotics. The Figure AI investment was not consensus; other venture investors doubted humanoid robotics due to past failures and perceived risks. Kang's conviction came from recognizing accelerating technology and the team's unique expertise in hardware, robot learning, hand engineering, and fleet management.

Kang explains that RoboStrategy tracks development across all robotics companies, not just their investments, to assess how players stack up in scaling fleets, robot learning research, and hardware development. From this perspective, Figure AI and Tesla Optimus are at the top. He emphasizes excitement about vertically integrated robotics companies that build their own intelligence, hardware, deployments, and manufacturing, because co-optimizing hardware and software improves training efficiency and robot performance.

Kang discusses that robot learning requires actual robot data, either from robots performing tasks or teleoperation. He uses an analogy: if you were transformed into a seven-foot-tall body, you'd be awkward, just as a robot needs data from its own embodiment to build effective models. Collecting this embodiment-specific data requires many robots, which is why scaling fleets is essential. This data advantage is a key reason for investing in vertically integrated companies.

Andrew Kang argues that the US and China will both achieve advanced robotics independently, but also through collaboration, given the open-source research shared between the countries. He dismisses the notion of a zero-sum race, stating that in 5 to 10 years both nations will reach similar capabilities. This perspective suggests a more cooperative future for the industry, contrary to popular narratives of a tech cold war.

Kang compares the robotics industry to the smartphone ecosystem, where hardware platforms like the iPhone enable a vast developer community to build specialized applications. He envisions a future where individuals and startups create robot applications for cooking, elder care, farming, or any physical labor task. This white space, he argues, is enormous and largely untapped, offering opportunities beyond the core hardware manufacturers.

Kang highlights Unitree and other Chinese firms that release robots as research or entertainment platforms, rather than waiting for perfect, ready-to-sell products like many US companies. This approach fosters a developer ecosystem, enabling data collection and model training specific to their hardware, which can create a competitive moat. He notes that his portfolio company Dexmate employs a similar strategy, and this model could lead to downstream innovations from external developers.

Kang asserts that the robotics industry is still in its infancy, unlike LLMs like ChatGPT that are widely deployed, so no one has missed the opportunity. He cautions that starting a robotics company is underappreciated in difficulty, requiring deep expertise in mechanical design, electrical engineering, high-rate manufacturing, and real-world deployment. He advises aspiring entrepreneurs and investors to do thorough research and understand the complexity before diving in.