Dispatch #9: Mid-Year Reckoning — Record Humanoid Funding, China's Robotaxi Lead, and the Deployment Gap

Welcome back to The Robotics Dispatch, our every-fourth-week briefing that ties real robotics headlines back to the concepts you have been building. Over eight editions we have followed the humanoid funding wave (#1), robot vision and AI models (#2), CES floor demos (#3), the sensor wars in self-driving (#4), factory robots and reshoring (#5), first real-world deployments (#6), the race for dexterous hands (#7), and — most recently — the safety standards still trying to catch up with all of the above (#8). This week, at the midpoint of 2026, we take stock: what has actually happened in the past fortnight, and where does the humanoid-and-autonomy arc stand?

The lead: Europe’s biggest humanoid fundraise

The clearest in-window story landed on June 10, when NEURA Robotics — a German startup building what it calls “cognitive robots” capable of perceiving, reasoning, and acting in unstructured human environments — announced a Series C of up to $1.4 billion. The round is backed by Amazon, Nvidia, Qualcomm, Tether (the stablecoin issuer, which has been quietly building a large robotics portfolio), Bosch, Schaeffler, and the European Investment Bank, among others. NEURA’s new valuation sits at $7 billion, making it the most-funded humanoid robotics company in Europe by a significant margin, and the round is the largest ever for a full-stack humanoid maker (The Next Web).

The company says the capital will drive serial production to multi-million units by 2030 and fund the global rollout of “NEURA Gyms” — proprietary real-world training environments where robots practise physical tasks to generate the data needed to improve their AI (The Next Web). The phrase you will see everywhere alongside this is Physical AI: the combination of large AI models with bodies that can touch, lift, and navigate the real world. Nvidia’s presence here is not coincidental — it has become one of the most consistent investors across the robotics AI stack.

NEURA’s round did not happen in isolation. The first half of 2026 has seen a run of enormous funding events across robotics and robot AI:

Skild AI (which builds a general-purpose “robot brain” — an AI model that can control any robot body without being retrained from scratch) closed a $1.4 billion Series C led by SoftBank on January 14, valuing the company at over $14 billion.
Apptronik (maker of the Apollo humanoid, in active pilots with Mercedes-Benz and GXO Logistics) closed a $520 million extension of its Series A on February 11, backed by Google and Mercedes-Benz, bringing the total round to $935 million at a $5 billion valuation.

Across all robotics companies, Dealroom reports $55.8 billion raised so far in 2026 — roughly double the full-year 2025 record.

Three of the largest robotics funding events in the first half of 2026, plotted by round size. Skild AI makes AI software for robots; Apptronik and Neura Robotics both build humanoid hardware. Note that the equal bar length for Skild and NEURA reflects the same dollar amount but very different businesses.

The AV arc: a scorecard says China leads on global reach

On the self-driving side, the fortnight produced something quietly significant: the first widely-reported quantitative scorecard ranking global robotaxi operators by operational metrics rather than valuation or hype. The Road to Autonomy Index, developed by Autnmy AI and published on June 17 (Axios) and June 21 (TechCrunch), draws from public databases — federal reports, SEC filings, state vehicle registrations — and ranks companies on operations, scale, revenue, commercial partnerships, manufacturing capacity, and safety record. The index updates every 12 hours.

The headline finding: Baidu’s Apollo Go edges Waymo to claim the top spot globally, with Pony.ai, WeRide, and Tesla rounding out the top five (TechCrunch).

Company	Weekly paid trips	Fleet size	Cities
Baidu Apollo Go	~350,000	1,000+	27 globally
Waymo	~500,000	3,500	11 (U.S.)
Pony.ai	Not disclosed	—	Multiple

Waymo’s weekly trip count is higher, and its fleet is larger. Baidu’s city footprint spans 27 cities worldwide and the company has logged over 22 million cumulative rides globally, a scale that weights heavily in the index’s cross-country methodology (Axios). The counter-view: Waymo is operating in one of the world’s most complex urban environments (San Francisco and Los Angeles) in weather and traffic conditions that are not typical of many Chinese cities. Neither reading invalidates the other; they reflect that the “who is winning” question in autonomy depends heavily on what you choose to measure.

What both sets of numbers agree on is that the robotaxi market has left demo phase behind. Half a million paid Waymo rides per week across 11 U.S. cities is not a pilot. Neither is 22 million Baidu cumulative trips across 27 cities. The frame from Dispatch #4’s sensor debate — when would autonomy actually start being used at scale? — has been answered.

Also this fortnight

Tesla’s Optimus Gen 3 is now in its first 24/7 industrial shift tests at the Fremont factory, following the conversion of the old Model S production line to humanoid assembly that began in January 2026 (TechTimes, June 9). The robots are not for sale — Elon Musk has put consumer availability no earlier than 2027 — but round-the-clock durability testing marks a genuine step beyond demo. Endurance at industrial duty cycles, not peak capability in a rehearsed scenario, is what separates a prototype from a product.

Meanwhile, the SoftBank–ABB Robotics acquisition ($5.37 billion, announced October 2025) is expected to close mid-to-late 2026 (CNBC). The deal folds one of the world’s largest industrial robotic arm makers — serving clients including BMW and the electronics manufacturing sector — into SoftBank’s growing portfolio alongside Skild AI, Berkshire Grey, and AutoStore. The consolidation signals that the era of isolated robot companies may be ending; the next wave looks more like platforms with shared AI brains.

Connecting the dots

The two tutorials you read in the run-up to this Dispatch are more directly relevant to these headlines than they might first appear.

ROS 2 launch files and parameters may look like a developer convenience, but the same orchestration principle is running inside every Waymo vehicle in the fleet. Each of those 3,500 cars runs dozens of software nodes simultaneously: LiDAR drivers, camera pipelines, sensor fusion, planner, control arbiter, safety supervisor. They are brought up by orchestration systems built on exactly the same idea you saw last week — declare your nodes, wire your parameters, start the whole stack in one command. The difference from your first Python launch file is scale and fault-tolerance engineering, not the underlying concept.

Debugging hardware is what running robotics at production scale actually requires on a daily basis. Half a million Waymo rides per week means constantly diagnosing sensor anomalies, software exceptions, and edge-case behaviours in real time across a distributed fleet. The systematic fault-isolation discipline from Week 25 — divide the system, observe at boundaries, swap the known-good — is the same discipline those engineering teams practise, just with more automated tooling around it.

And if you read Six Months In two weeks ago, you will recall it closed Year One by noting that the skills we had built in tutorials were the same fundamentals industry is wrestling with at scale. This Dispatch is a real-time illustration: the billion-dollar questions being asked of NEURA’s robots — can they perceive reliably, reason in novel situations, and act safely? — are the perception, state-estimation, and control problems you have been building up to across 26 weeks.

Next week we go deeper into how ROS 2 nodes communicate: ROS 2 Services and Actions (Week 28) — the request/response and long-running-task patterns that sit just beneath the surface of every robot software stack described above.

The takeaway

The mid-2026 picture is one of genuine momentum meeting genuine friction. The money is real — $55.8 billion into robotics in roughly six months reflects investors’ conviction that AI’s next phase is physical. The trip counts are real. The industrial tests are real. Funding rounds at $1.4 billion for a German humanoid maker and for a robot AI software company signal that the competitive field has gone global in ways that would have seemed speculative two years ago.

But “multi-million robots by 2030” is a forecast, not a fact. Baidu’s 22 million cumulative trips represent years of operation across a very different regulatory environment than the one Waymo navigates. And 24/7 shift testing is not the same as reliable production deployment.

Our read: the most useful frame for the second half of 2026 is not “will humanoids make it” — they have, in a narrow sense — but “at what cost, under what rules, and in which environments first.” The safety standards from Dispatch #8 are still being written. The deployment data from these pilots is only now arriving in volume. Six months from now we will have a clearer answer to both questions — and considerably more of it.