Andrew Kang said in RoboStrategy’s August 2026 shareholder letter that venture capital allocated to pure software should shrink materially, while companies building in the atomic world can produce extreme power-law returns. His central argument was that robotics investing is reaching an inflection point even as private-market valuations still lag what he sees as fair value.
The letter, dated Aug. 20, 2026, was addressed to shareholders of RoboStrategy, Inc. Kang wrote that a little more than three months had passed since RoboStrategy listed on May 11. From April 30 to July 31, NAV increased 85%, while NAV per share rose 54%. He said most of that growth came from accretive financings with institutional investors. Kang also said he personally invested an additional $10 million into the fund at an approximately $36.7 per-share market premium.
Even with investor demand rising and portfolio companies growing, Kang said RoboStrategy’s NAV is still anchored to the last private financing round for each company. In his view, those financing rounds may not fully reflect the fair value of private companies, and the disconnect between private-market marks and fair value may persist.
Unitree as a recent valuation datapoint
Kang pointed to Unitree’s IPO as a recent example. He said the company opened with a $66 billion market capitalization, roughly 39 times its last venture valuation of $1.7 billion and 7 times its IPO valuation. He added that the spread may not represent the broader market as a whole. Unitree is not currently in the RoboStrategy portfolio, but Kang said it fits the firm’s investment focus and could be added in the future.
$124 million deployed after listing, with more lead investments
Since going public, RoboStrategy has invested another $124 million across six companies, according to the letter. The firm increased positions in Standard Bots and Dexmate, while making new investments in Nox Metals, Eccentric Machines, Prometheus, and an unnamed robotics company that is expected to emerge from stealth.
Kang said the firm has become more inclined to lead rounds in the companies where it has the highest conviction as the fund has grown. So far, RoboStrategy has led rounds in Dyna Robotics, Standard Bots, Eccentric Machines, REK, and one stealth robotics company. He said lead-investor status can give the fund tighter working relationships with founders and terms that may benefit shareholders, including board seats, information rights, pro rata rights, warrants, rights of first refusal, and added investor protections.
Industry changes have started to favor the portfolio
Kang wrote that the industry has already gone through meaningful changes, including an FCC move in the U.S. to ban foreign mobile robots, the rise of open-source models, and an accelerated timeline for robotics foundation-model development. He said RoboStrategy had anticipated those shifts in advance, leaving the portfolio positioned to benefit from them rather than absorb them as headwinds.
He added that many portfolio companies have moved from pilot work into a growth phase, with real progress in deployments and shipments.
Standard Bots: $200 million Series C at a $1 billion valuation
Kang described Standard Bots as the largest AI-native industrial robotic arm manufacturer in the U.S. RoboStrategy led the company’s $200 million Series C round at a $1 billion valuation.
He said the company expects to deliver 10% of new industrial robot deployments in the U.S. next year. Its robotic arms are already in use at hundreds of companies across almost every state. To meet demand, the company’s Glen Cove factory is set to expand to 70,000 square feet.
Kang also highlighted a June White House roundtable on America-first industrial policy attended by Standard Bots CEO Evan Beard. The meeting was hosted by Commerce Secretary Howard Lutnick and White House trade adviser Peter Navarro. Kang said Beard was the only founder of an American robotics company in the room. He also cited a recent Politico report saying the Office of Strategic Capital at the Department of War is underwriting a financing transaction for Standard Bots.
Figure AI: more than 1,000 humanoid robots produced in July
Figure AI has started ramping output of Figure 03, Kang wrote. In July, the company’s humanoid robot production topped 1,000 units. Its existing BotQ factory has annual capacity for 12,000 robots, and CEO Brett Adcock has said the goal is to reach 1 million robots per year within a decade.
Figure has expanded beyond its initial BMW pilot into more complex sorting applications in logistics at BMW’s Spartanburg facility, according to the letter. The company has also signed a commercial agreement with Catalyst Brands for large-scale humanoid robot deployment. Kang noted that Catalyst operates brands including JCPenney, Aéropostale, and Brooks Brothers.
Apptronik: a 90,000-square-foot data facility in Austin
Kang said Apptronik has opened Robot Park in Austin, Texas, a new 90,000-square-foot data collection facility meant to train Apollo 2 together with Google DeepMind.
According to the letter, Apollo 2 is continuously generating training data used to train and optimize the Gemini Robotics model in preparation for real-world commercial fleet deployment. Apollo 3 is expected to be the company’s first fully commercialized, mass-produced product, with a 2027 target. Apptronik has also hired former Waymo product lead Daniel Chu as chief product officer and brought in executives from Boston Dynamics and Amazon.
Dyna Robotics: DYNA-2 trained on more than 1 million hours of human video
Dyna Robotics released DYNA-2, its flagship world-action model. Kang said the system was trained on more than 1 million hours of human video, which he estimated was two orders of magnitude above the amount of human video previously used in robotics foundation models.
He said DYNA-2 showed scaling laws from humans to robots, with the key finding that robot performance improves smoothly and predictably as more human data is added, without hitting a plateau. In the letter, Kang argued that the model breakthrough improves trainability and physical agency, and that combined with stronger robot hardware it leaves the company well placed for broader deployment markets.
Dexmate: hundreds of humanoids shipped
Kang said Dexmate has become a leading humanoid robot seller in the U.S. and has already shipped hundreds of units. The company appears across many major physical AI teams, with customers that include Nvidia, Amazon, Google, LG, Skild AI, Generalist, and lab groups at UC Berkeley, Carnegie Mellon, Harvard, Columbia, and NYU.
RoboStrategy led Dexmate’s seed and seed-plus rounds at post-money valuations of $123 million and $216 million, respectively. Kang said the investment thesis was that a U.S. company had a chance to replicate Unitree’s business model by offering high-quality humanoid robot platforms to developers and researchers, and that Dexmate is executing on that opportunity.
Path Robotics: performance-based shipbuilding agreement worth up to $600 million
Path Robotics signed a performance-based production agreement with HII covering up to $600 million of shipbuilding work aimed at deploying advanced robots in U.S. Navy shipbuilding programs. Kang called it one of the largest contracts ever awarded to a physical AI company.
The company also launched Rove, a quadruped welding robot that moves to the workpiece instead of requiring the workpiece to be brought to the robot. Kang wrote that this is the only way to automate welding on structures too large to fit inside a work cell, making it especially suited for shipbuilding. Saronic, which he said has raised more than $2.5 billion to build autonomous ships, is set to receive the first Rove units in the first quarter of 2027.
Eccentric Machines, REK, GMI and Nox Metals
Eccentric Machines is building Sentor, which Kang described as an AI-native actuator. The design replaces the standard “single-geared motor” architecture with a new motion system intended to respond directly at the joint to unexpected loads rather than waiting for a robot’s central brain to process them. He framed the effort as a way to push the subconscious layer of physical AI down to the joints while balancing manufacturability and the major trade-offs around actuator performance, efficiency, and mechanics.
REK held what Kang described as the first six-foot humanoid robot fight in the U.S. this month at its San Francisco venue, drawing athletes and creators. He said earlier events had sold out a 2,500-seat venue and that the company is now looking for a facility several times larger.
GMI is building a $500 million AI factory in Taoyuan, with about 7,000 Nvidia GB300 chips spread across 96 racks, drawing 16 MW of power and capable of processing nearly 2 million tokens per second. The company has also launched a $12 billion sovereign AI infrastructure initiative in Japan and is building a global sovereign AI factory network with Magna AI.
Nox Metals closed an $11.5 million seed round about seven months after it was founded, with Palmer Luckey, Y Combinator, and RoboStrategy among the participants. Kang said the company has already cut custom aluminum for about 100 customers across the country, from small machine shops to Anduril and SpaceX. It is moving into a 75,000-square-foot factory in Detroit to meet demand.
Team expansion, media buildout and policy work
Kang said RoboStrategy is not trying to build only a passive investment vehicle. The firm wants to operate as a high-standard investment institution and create excess value for the RoboStrategy, Inc. fund. Alongside investing and research, he said the firm intends to use its media expertise to help portfolio companies reach customers, talent, and investors.
Since inception, RoboStrategy has built the core of its marketing engine across short-form video on X, highly produced YouTube video, and long-form research and investment theses, according to the letter. Kang said the firm’s own content has generated more than 20 million impressions on X, while team media appearances have produced more than 1.5 million views across channels. Coverage has been picked up by Bloomberg, Reuters, CNBC, TechCrunch, BBC, and Yahoo Finance.
Over the next few months, RoboStrategy plans to expand across channels with in-depth reporting, industry shorts, broader relationships with creators and influencers, growth in its newsletter audience, and more live and podcast programming focused on robotics and physical AI. Kang said the firm wants to move beyond its core English-speaking audience and build global reach, starting with Korea, Japan, Taiwan, and China, with the goal of becoming a primary source of information on robotics.
RoboStrategy Advisors has grown to 14 people and is expected to reach 17 by September. Functions now span investing, research, marketing, compliance, finance and operations, and legal and policy. Kang wrote that the U.S. is developing a national robotics strategy and noted that the FCC this month issued a notice banning imports of new foreign mobile robot models. He said the government can and will take additional steps to support domestic companies, potentially including long-duration debt facilities, direct investment, and tax incentives.
To support that effort, Bill Hughes has joined as vice president of legal and policy. Kang said Hughes previously served as a deputy associate attorney general at the U.S. Department of Justice and as a special assistant to the president at the White House.
Kang also said RoboStrategy is building an institutional capital formation and distribution team. He argued that robotics will draw growing interest from a wide range of investors and that the firm’s fund offers a distinctive value proposition for RIAs, private-wealth platforms, and institutions. He cited data showing that roughly $200 billion flows each year from private-wealth channels into semi-liquid asset funds. A larger fund, he wrote, would allow RoboStrategy to lead more financings, benefiting both portfolio companies and shareholders while strengthening distribution.
Research framework and supply-chain bottlenecks
In the research section, Kang laid out how RoboStrategy underwrites robotics investments. He wrote that accurate risk-reward assessment requires clear views on many connected technical trends.
For a robotics foundation-model company, growth prospects depend not only on the strength of its researchers and models, he said, but also on the path of open-source models, the feasibility of deploying cloud-based models across environments, and many variables that affect manufacturing capacity. Kang specifically pointed to the viability of new actuator innovation and the supply elasticity of gear grinders. Those machines, he wrote, are produced by only a small number of companies and are a bottleneck for harmonic drives, which are critical components in many robot form factors.
RoboStrategy plans to publish some of that research to share knowledge and extend its brand and reach. Kang added that the firm’s high standards in investment diligence apply across the rest of its work as well.
Industry outlook: a financing turning point, but sharper separation ahead
Kang said the first half of the year was a wake-up call for venture capital. In his telling, investors are starting to recognize that pure software venture investing should be scaled back while companies building the atomic world may deliver extreme power-law outcomes. He said many investment firms are hearing internal and external pressure from LPs to shift allocation strategies.
Sometimes investors need undeniable proof, Kang wrote, and companies such as SpaceX and Anduril have now provided it as they enter later growth stages. He added that SpaceX alone has created roughly $2 trillion in liquidity events for investors and employees, which could send substantial downstream capital into other deep-tech companies.
Compared with earlier quarters, venture investment in robotics rose sharply in the first quarter and reached the highest level shown in RoboStrategy’s chart. Even so, Kang said robotics venture funding is still only a small fraction of AI venture funding before ChatGPT appeared in 2021. He described the current period as a financing inflection point for the industry and said he expects robotics to catch up with AI over the next several years.
Kang said this is a good time for funds to invest in private robotics companies, while also warning that financing around some names is already getting out of control. Selection matters, he wrote, and so does differentiation. In his view, the rising tide is lifting many companies in the field for now, but businesses that are building toward the wrong target or do not have the talent they need may still run aground. He said both of the following can be true at once: valuations for winners keep climbing, and some overhyped players hit the wall. He said the AI market has already shown that pattern over the past four years.
On technology and deployment, Kang argued that industrial capability is entering an early scaling phase. He said many of the key architectural questions in robotics foundation models are largely solved, and that this year’s main focus is scaling data collection and processing. In his words, physical-world versions of emergent intelligence are already showing up and should accelerate next year toward GPT-5-level performance, including in-context learning, memory, and natural-language instruction following.
He also wrote that researchers now have the broad blueprint for general physical intelligence. The next major unlock, in his view, will be a big jump in sample efficiency during post-training, cutting deployment time for robots in specific jobs from weeks to hours or even minutes. Robotics will not have a single ChatGPT-style moment, he said. The adoption curve will look more like self-driving cars: not recognized by the whole world in a single day, but spreading quickly through new autonomous machines in many shapes and sizes. By next year, Kang wrote, intelligence will no longer be the bottleneck. The bottleneck will be the robots themselves.
The letter ended with fund risk disclosures, including that investors should review RoboStrategy, Inc.’s prospectus at RoboStrategy.co or by phone, that past performance does not guarantee future results, and that net asset value and market price can fluctuate, leaving current performance above or below the figures cited in the letter.
