Author: Rinat Mirzaitov

Seven in ten are not building a humanoid Ask what a humanoid robotics company does and the obvious answer is that it builds humanoid robots. The tracker says otherwise. Of 1,112 organizations in more than 80 countries and territories, 310 have a complete humanoid as their primary focus. That is 28%. The other 802 make the parts, write the software, deploy other people’s robots, or build a different kind of robot and watch this one closely. The largest group is not the builders. It is the 356 organizations that supply the body, from motors and reducers to sensors, hands and…

SoftBank’s $5.4 billion plan to buy ABB Robotics has reached a question that matters across physical AI: how much of a global robotics business can a buyer actually control? Reuters reported on November 5 that China had blocked the acquisition. SoftBank’s November 6 disclosure clarified the scope: Chinese authorities prohibited the purchase of ABB’s robotics business in China, with notification received on November 4. SoftBank said both companies were still working on the transaction. The immediate investor question is what assets and rights an acceptable deal would contain. A revised agreement and regulatory clearance remain missing. By November 7, Reuters,…

Agility Robotics is bringing one of humanoid robotics’ strongest documented deployment stories to public markets. Nasdaq confirms that its merger with Churchill Capital Corp XI closed on November 4, with the switch to ticker AGLT effective November 5. For investors, the next question is whether useful robot work can become a sustainable business. The transaction assigned $2.5 billion of equity value to Agility’s existing holders, rather than establishing a live stock-market valuation. A November 2 filing estimated approximately $465.1 million in gross capital from the transaction and related financings, including $100 million already received. That figure is before transaction expenses.…

Military trials are opening a new market, but the strongest early case rests on dangerous support work, human oversight and measurable results. The first humanoid to earn a lasting place on the battlefield may spend its day moving supplies, inspecting damaged infrastructure or handling equipment where sending a person carries serious risk. That is the most plausible near-term opportunity suggested by the evidence available today. It could also become a meaningful business before humanoids are capable of replacing soldiers across complex combat roles. Military interest is becoming concrete. The U.S. Army selected five xTechHumanoid winners on September 11, 2026, following…

Anthropic’s October 26, 2026 IPO filing brings physical AI into its formal growth strategy. The Claude developer identifies robotics, advanced manufacturing, and physical work as markets where frontier AI could create new business opportunities. The development matters beyond Anthropic’s proposed stock-market listing. It signals that one of the largest frontier AI developers sees an opportunity to extend its models from digital workflows into systems that interact with real equipment. But the distinction between strategic ambition and commercial evidence remains important. Anthropic has demonstrated early capabilities in robotics and laboratory automation. Its filing does not establish a commercial humanoid-robot business, disclose…

Tesla says its first production-quality Optimus robots are coming off manufacturing lines in Fremont. For investors, the next test is what happens after they leave the factory: can they do useful, reliable work at a cost customers will pay? In its Q3 2026 update, released October 21, Tesla says the current manufacturing process is suitable for a commercial product, while acknowledging that hardware and production methods will continue to evolve. The disclosure provides no Optimus output totals or customer operating results. A detail on the same page keeps the announcement in perspective. Tesla lists California Optimus as “Early Ramp” and…

The largest AI software companies no longer treat humanoid robots as someone else’s business. In 2026 alone, OpenAI said it will build its own humanoid, Google DeepMind shipped a model that drives a humanoid’s whole body, and Microsoft released its first robotics model. Anthropic published a hardware interface standard, and DeepSeek bought into Unitree’s IPO with a pact to co-develop robot models. China’s platform giants moved earlier and wider. Tencent, Alibaba, ByteDance and Ant Group each now run embodied AI models, robot programs or large investment books in the sector. Yet “entering the humanoid race” means three different things. A…

In September 2026, Figure sent its humanoid robot, running its new Helix 2.5 model, into 30 homes it had never seen. It tidied living rooms, folded towels and made beds. By Figure’s own count, it finished the whole job 56% of the time. In April, the startup Generalist showed its GEN-1 model folding 86 shirts in a row. Generalist reports 99% average success, but only on tasks the model had been trained to do. Those two results sum up physical AI in 2026. One system copes with places it has never seen, and the other does familiar work almost perfectly.…

Teaching a robot your job becomes a harder proposition when you suspect it is learning to replace you. Some Tesla factory workers reportedly raised that concern when asked to record their movements for Optimus training. Their complaints point to a potential obstacle for humanoid developers: securing cooperation from the people whose work they hope to automate. The scale of the objections and their effect on training remain unknown. According to The Next Web’s account of The Information’s September 25 report, Tesla used movement-recording suits on factory workers in Texas and California. Some complained because they understood the robots were intended…

Useful physical data remains the better-evidenced bottleneck in robot learning. Figure’s Helix 2.5 results strengthen that assessment: the company reports that pretraining on its Index dataset raised full-task success in unfamiliar homes from 9% to 56%. [1] For investors, the infrastructure thesis is becoming more concrete. Developers are paying to acquire experience and committing to larger computing systems. The evidence supports a market for collecting, processing and learning from physical data. It does not yet establish that general-purpose robotics requires more training compute than frontier language models. The distinction matters. Demand for infrastructure can become commercially significant before compute becomes…