EngineAI just opened a 129,000 square foot facility in Honghualing that produces a humanoid robot every 15 minutes. The first batch of its T800 robots is already coming off the assembly line, targeting deployment in industrial production environments. EngineAI closed a 200 million dollar Series B in April at a valuation over 1.4 billion dollars.
This is the third major Chinese humanoid robot factory to come online in 60 days. Leju Robotics and Dongfang Precision opened a Foshan facility on March 29 producing 10,000 units annually. Unitree is targeting 20,000 units in 2026 with a 580 million dollar funding push for a 75,000 unit facility. Agibot shipped over 5,000 units in 2025. UBTech is targeting 10,000 units this year. China shipped roughly 5,000 humanoid robots in 2025 and is on pace to produce tens of thousands in 2026.
US competitors are not at this production scale. Figure AI delivered maybe a few hundred robots in 2025. Tesla Optimus production targets keep slipping. Boston Dynamics Atlas remains research-grade. The conversation in Western tech press about humanoid robots is still mostly about whether they work. The conversation in China has already moved to how many you can build per minute and where they get deployed first.
Virgin Galactic flew its VSS Unity spaceplane yesterday for the first time since retiring it from commercial service in June 2024. The mothership Eve carried Unity to altitude over Spaceport America in New Mexico and released it, and Unity glided back to a runway landing. The flight was not for tourists. It was pilot training for the next generation Delta class spaceplane currently in final assembly near Phoenix.
Virgin Galactic is one of the few companies still pursuing pure space tourism rather than orbital infrastructure or government contracts. The technology works. Not sure whether the economics work. Cash reserves dropped from 982 million to 338 million in a single year and no revenue comes in until Delta enters service. The entire company is essentially betting that wealthy people will pay three quarters of a million dollars for four minutes in space, frequently enough to sustain a fleet flying twice a week. The next 12 months will determine whether that bet was the right one.
China is now the world's largest manufacturer of tunnel boring machines, the giant machines that dig highway, rail, and underwater tunnels. Seven out of every ten tunnel boring machines used worldwide now come from China, and 95 percent of the machines operating inside China are domestically made. A decade ago China had almost no capability in this field.
Same pattern showing up across category after category. Build the capability through enormous domestic demand, gain experience on the hardest projects, then export the machinery worldwide. It happened with high speed rail, solar panels, EVs, and batteries. Heavy tunneling equipment is the same story playing out in a sector almost nobody is watching.
A wave of well-funded startups is racing to apply AI to advanced mathematics. Axiom Math raised 200 million dollars at a 1.6 billion valuation in March. Harmonic, backed by Robinhood founder Vlad Tenev, is in the same race, alongside Logical Intelligence and Math Inc, plus the big labs at OpenAI and Google.
The near-term impact will be AI quietly handling the tedious intermediate steps that fill a working mathematician's day. The interesting signal is that mathematicians are leaving academia for these companies, because the combination of machine learning pattern recognition and mathematical precision is starting to look like a path toward more general intelligence.
A small Utah company called OxEon Energy built the only piece of hardware ever to extract oxygen from the atmosphere of another planet. Their solid oxide electrolyzer flew on NASA's Perseverance rover as part of the MOXIE experiment, ran 16 times over 18 months, and produced propellant-grade oxygen by electrolyzing Martian carbon dioxide. It was the first experimental extraction of a natural resource from another planet for human use. OxEon was founded in 2017, employs around 40 people, and operates out of North Salt Lake.
The Department of Energy awarded the company 36 million dollars under the Bipartisan Infrastructure Law to expand solid oxide electrolysis manufacturing to 25 megawatt annual capacity. NASA awarded an additional 3.97 million for next-generation cathode materials. The same core technology that made oxygen on Mars now produces hydrogen and synthesis gas for industrial customers on Earth, with eVTOL aircraft power systems and Navy fuel projects in development.
A Seattle startup called Interlune is preparing to mine the Moon. Founded in 2020 by former Blue Origin president Rob Meyerson, Blue Origin chief architect Gary Lai, and Apollo 17 astronaut Harrison Schmitt, the company is building autonomous excavation equipment to extract helium-3 from lunar regolith (moon debris). The first prospecting mission, a multispectral camera deployed via Astrolab, is scheduled for this summer. A sample-gathering mission is targeted for 2027.
Interlune has roughly 500 million dollars in signed agreements with the US Department of Energy, the US Air Force, Bluefors, and Maybell Quantum. Bluefors alone has committed to buying up to 10,000 liters of helium-3 annually from Interlune between 2028 and 2037. NASA awarded Interlune a 6.9 million dollar contract in May for a helium-3 extraction payload that will be flight ready by fall 2027. The market rate for helium-3 is roughly 20 million dollars per kilogram.
Helium-3 demand from the quantum computing industry already exceeds Earth supply. Quantum computers need helium-3 in dilution refrigerators to reach near absolute zero. As quantum systems scale through this coming decade, demand for helium-3 grows from dozens of liters to thousands of liters annually. Interlune is positioning to be the supplier when Earth runs out.
Researchers have developed a way for earbuds to identify their wearer through heartbeat patterns picked up by the earbuds' microphones.
The system uses the unique acoustic signature of a person's heartbeat, captured inside the ear canal, as a biometric identifier similar to a fingerprint or face scan.
Early testing showed strong accuracy in distinguishing between users. Potential applications include automatic device unlocking, secure authentication for payments, and continuous identity verification during use. The technology is still in research phase with no commercial product announced yet.
Researchers at the Chinese Academy of Sciences published a solid-state lithium-metal battery achieving 451.5 Wh/kg energy density with three minute full charge cycles, more than double the energy density of current commercial EV cells.
In the past four months Chinese players have announced commercial solid-state battery milestones at a steady pace and the Chinese government plans to release national solid-state battery standards this year.
Western timelines for solid-state battery production are mostly 2028 to 2030. Chinese timelines for the same technology are 2026 to 2027.
NASA researchers at Glenn Research Center just discovered a new material that can hold molten regolith (martian dust) at over 3,000 degrees Fahrenheit without being corroded. The research was on Moon dirt but the technology applies directly to Mars, where soil is similarly rich in oxidized iron, silicon, and aluminum that can be extracted through melting and electrolysis.
Shipping cargo to Mars costs roughly 1 to 4 million dollars per kilogram with current rockets. A single ton of building material costs more than most houses on Earth. Permanent settlement is economically impossible if humans are launching every beam, every wire, every oxygen tank from Earth. The only path that works is making the heavy stuff on Mars from what is already there. Soil becomes oxygen for breathing, it provides iron for structures and silicon for solar panels.
The new material solves one of the bigger industrial bottlenecks. 🚀
Tesla is building a 100 gigawatt per year solar panel factory in Brookshire, Texas near Houston, co-located with its Megapack battery factory. SpaceX is planning a similar 100 GW facility independently. If both succeed, those two companies alone would produce more solar panels annually than the entire current US solar industry. Equipment deliveries fall 2026. Full target 2028.
Tesla is building the full supply chain at one site, from raw silicon through finished panels, and pairing it with grid-scale battery storage in the same complex.
The era of building data centers wherever the grid happens to have capacity is ending. The era of building generation, storage, and compute together at one site is starting. Tesla is positioning to be the supplier for that buildout.
The AI industry is going public. SpaceX filed confidentially April 1 and is targeting a June 12 IPO at roughly $1.75 trillion. OpenAI is preparing a confidential filing in the coming weeks for a September listing that could exceed $1 trillion. Anthropic is targeting October at a $900 billion valuation that may actually top OpenAI. Cerebras already IPO'd this week, popping 68% on day one before giving some back.
Until now the entire AI infrastructure thesis has been priced through private market signals, leaked funding rounds, and inference about valuation chains. Once these three companies are public, the AI capex cycle becomes auditable in real time through quarterly filings. Analysts will model GPU spend the way they model semiconductor demand. Index funds and pension money will buy AI exposure passively.
Retail investors are about to get direct access to companies that until now were locked behind eight figure private placements. Whether that ends well depends on what the actual financials look like once they have to be disclosed.
Autonomous vehicles have a perception problem most people do not realize. Cameras fail in fog and darkness. Lidar degrades in heavy precipitation. Radar works in any weather but cannot tell what it is seeing. Every L3 to L5 autonomy system today is a tradeoff between resolution and reliability.
A US startup called Teradar unveiled what it calls the industry's first long-range high-resolution terahertz vision sensor at CES 2026. Terahertz waves sit between radar and lidar on the electromagnetic spectrum. The promise is high resolution that works in any weather, fog, rain, snow, or darkness. Teradar has eight development partnerships across the US and Germany and plans to start production in 2028.
The US Department of Energy is funding a program called Rapid RUNNERS at Oak Ridge National Laboratory to 3D print large hydropower turbine components, the rotating runners that convert water flow into electricity. The first three Francis runners are being produced now using robotic welders that deposit metal layer by layer.
Large hydropower turbine components, those over 10 tons, currently cannot be manufactured in the United States. They are sourced from international steel foundries with 18 month lead times. When a turbine fails, that is 18 months of lost clean energy while parts ship across the world. Hydropower provides nearly a third of US renewable energy generation, but its critical supply chain is entirely offshore.
Additive manufacturing is being deployed across power generation, not just for hydropower but also for steam turbine blades at Siemens and gas turbine components at GE. The infrastructure to make America's clean energy infrastructure is finally being rebuilt.
China just became the first country to require digital ID cards for humanoid robots. Hubei Province launched the registration system this week, building on the national standards framework released in March 2026. Every robot gets a unique code with full traceability covering joint wear, battery status, operational accuracy, and incident liability.
China had 140 plus manufacturers release 330 plus humanoid models in 2025, which authorities now describe as the country's first year of mass production. The ID system is essentially the infrastructure response to that scale. When something goes wrong, regulators want to trace the cause to a specific machine and specific manufacturer immediately.
Meanwhile China is already two regulatory layers ahead of the western countries with National standards in March and traceability registration in May.
Mind Robotics just raised $400 million at a $3.4 billion valuation to scale AI powered industrial robots in manufacturing. The round was led by Kleiner Perkins with participation from the venture arms of Volkswagen and Salesforce. Total funding now exceeds $1 billion in less than a year.
Mind Robotics was founded in 2025 by Rivian CEO RJ Scaringe, originally as an internal Rivian $RIVN project called Project Synapse before being spun out in November. Rivian's factories serve as the live training ground for the AI systems and Rivian remains a shareholder.
The moat in industrial AI robotics is not just the models or the hardware, it's access to real manufacturing environments to train them. Mind Robotics has that through Rivian. Manufacturing companies next on the list to get a bump from AI?
Western media doesn't seem to cover much of Chinese space activity but their space station Tiangong is in routine crewed rotation with two crewed missions planned for 2026. Chang'e 6 returned samples from the far side of the moon in 2024. Chang'e 7 launches this year targeting the lunar south pole. Tianwen 2 is already en route to an asteroid for sample return.
When the ISS retires around 2030, Tiangong may be the only continuously occupied national space station unless commercial US stations meet their schedules, which is currently uncertain. China's lunar timeline targets a crewed landing in 2030, matching the current US Artemis schedule.
Scientists just proposed a new way to look for alien life by instead of asking whether specific molecules are present, they look at how the molecules are organized.
The approach was published in Nature Astronomy this week. The team analyzed about 100 datasets from asteroids, fossils, meteorites, microbes, and lab samples. They found that amino acids and fatty acids show distinct organizational patterns when produced by life versus by abiotic chemistry. Life leaves a signature in how molecules are distributed, not just which ones are present.
The interesting part about this is that NASA's Europa Clipper spacecraft, currently on its way to Jupiter, already carries an instrument that can measure exactly this kind of data when it arrives in 2031. A method that did not exist a week ago can now help interpret findings from a spacecraft already in flight.
A Danish research project called MIRAQLE is developing quantum enhanced MRI technology that promises to make scans up to 100,000 times more sensitive than current machines. Led by Aarhus University with German company NVision Imaging Technologies, the system uses quantum physics to amplify contrast agent signals and visualize cancer metabolism during routine scans.
The clinical impact would be significant because a standard MRI shows where a tumor is but Quantum enhanced MRI would show how the tumor is behaving and whether treatment is working, compressing the diagnostic feedback loop from months to days. Initial focus is on liver cancer, one of the hardest cancers to detect early. The technology integrates with existing MRI scanners rather than requiring new equipment.
Real example of quantum technology arriving in clinical medicine. The same platform is being explored for treatment monitoring across multiple cancers, plus Alzheimer's disease and multiple sclerosis applications.
Samsung is reportedly launching its first AI smart glasses on July 22 in London. The product, possibly called Galaxy Glasses, will run Android XR with Google's Gemini AI built in, partner with Warby Parker and Gentle Monster on frames, and start around $400. No display in the first version, just camera, microphones, speakers, and AI.
Meta has been shipping Ray-Ban smart glasses since 2024. Samsung enters this July with Google. Apple is rumored to follow in 2027. OpenAI is reportedly working on hardware for 2027 as well.
Is the bet is that the next consumer AI device will be glasses, not phones?
Astronomers just used machine learning to find more than 10,000 new exoplanet candidates in old NASA data that humans had already looked at without finding them. If verified, this would more than double the total number of known planets outside our solar system in a single step.
The research team applied AI to data from NASA's TESS satellite, but instead of analyzing bright stars where the signal is strong, they aimed it at the 83 million fainter stars TESS had also observed but mostly ignored. The noisier data was too hard for humans to comb through.
The findings are preliminary and still need peer review and individual follow-up confirmations. AI is starting to extract new science from datasets we already have, not by replacing telescopes but by finding what humans missed in the data we already collected.