A single manufacturing step can now do something that normally takes two separate chemicals. It makes a material bright white and water-repellent at the same time, without using either titanium dioxide or PFAS. Researchers at Kyoto University say the technique could remove two of the materials industry’s most persistent chemical dependencies in one move.

Professor Easan Sivaniah, at Kyoto University’s Institute for Integrated Cell-Material Sciences, led the project with collaborators from Tokyo Metropolitan University and Donghua University. The team calls the resulting process Deep Foam Photolithography.

Two chemicals under pressure

Titanium dioxide has long served as the default white pigment across packaging, films, and coatings. The European Union recently banned its use as a food additive, though, over safety concerns. PFAS compounds face a parallel reckoning. These fluorinated chemicals give materials their water- and oil-repellent properties, yet mounting evidence of their environmental persistence has manufacturers worldwide hunting for substitutes.

Sivaniah’s team set out to solve both problems using physics rather than chemistry. Rather than finding safer replacement chemicals, they asked whether a material’s structure alone could deliver the same effects.

A trick borrowed from nature

The underlying phenomenon is called structural whiteness, and it appears throughout nature without any pigment involved at all. Snow, sea spray, and clouds all owe their brightness to how light scatters off foam-like structures made mostly of air. That same principle is at work in plant tissues and the protective nests certain frogs build.

That principle even shows up in centuries-old art. The blinding white of the waves in Hokusai’s The Great Wave off Kanagawa comes entirely from light scattering off bare washi paper fibers. The print itself contains no white ink at all.

Researchers engineered porous structures that replicate this same light-scattering effect, drawing additional inspiration from how light interacts with leaves and flower petals in nature.

A two-step process

Making the material requires just two steps. Light exposure first breaks a polymer down into smaller molecular fragments. A mild solvent treatment then causes those fragments to react and swell, forming an open, porous structure throughout the material.

That single process yields two properties at once. Internally, the porous structure scatters light to produce bright whiteness with no added pigment whatsoever. On the surface, that same structure becomes rough enough to repel water and oil, mimicking the effect seen on a lotus leaf.

Working with textile researchers at Donghua University, the team confirmed the process works on fabrics in addition to printable polymer films. Crucially, it required no newly invented specialty chemicals, succeeding instead with polymers already available commercially.

The resulting platform supports printing resolutions up to 20,000 DPI while delivering structural whiteness and water repellency together, all without titanium dioxide or PFAS. Rather than relying on added pigments or persistent chemical coatings, the approach builds both color and function directly into a material’s physical structure. That points toward a lighter, more sustainable way to manufacture everyday materials.

A Seattle-based company is set to boost autonomous operations in the U.S. Army. Overland AI, the leader in ground-based autonomous capabilities, has been selected by the U.S. Army’s Capability Program Executive (CPE) Mission Autonomy for the Engineer Autonomous Breaching Capability (EABC) initiative.

This $20 million contract signifies a broader shift from proving that autonomy works to delivering an operational breaching package that soldiers can employ.

Breaching mission can be executed most effectively with truly autonomous vehicles

“A breaching mission can be executed most effectively with truly autonomous vehicles,” said Byron Boots, co-founder and chief executive officer of Overland AI.

“Our proprietary software stack is the best solution for ensuring the safety of soldiers by modernizing the battlefield.”

The company revealed that its Overland AI’s software stack, OverDrive, can be integrated onto any ground vehicles to make them autonomous and enable multi-robot coordination. This technology is being leveraged for complex military operations such as breaching to overcome obstacles and facilitate the safe passage of follow-on forces.

With one operator to many robotic platforms, the company’s autonomous systems are designed to reduce the number of combat engineers needed to complete a breach. In doing so, we can reduce unnecessary risk to soldiers during these exceptionally hazardous missions. For operations that have historically cost many American lives, this shift has meaningful implications, according to a press release.

Autonomous mission capabilities, spanning software stacks, hardware platforms

CPE Mission Autonomy was established to accelerate Army acquisition of autonomous mission capabilities, spanning software stacks, hardware platforms, and modular payloads. Along with its partners and subcontractors, Overland AI is delivering the full capability, from operator to battlefield effect. 

Overland AI was selected for its innovative approach to autonomous breaching, honed through extensive work with warfighters across multiple combat engineer units and Department of War research labs, and more than 25 government demonstrations for DARPA, DIU, and the U.S. Army. The EABC program is designed to deliver a cohesive autonomous breaching package, and Overland AI is honored to bring this next generation capability to warfighters, as per the release.

Breaching is a critical combat-engineering function. Forces may need to create passages through obstacles so that troops, vehicles and other equipment can continue advancing. Traditionally, combat engineers must operate in highly exposed environments while identifying and overcoming these obstacles.

The EABC initiative seeks to change that equation by introducing autonomous systems into the process.

The objective is not simply to put a vehicle under remote control. Instead, the Army is pursuing a broader autonomous capability that combines vehicles, software, sensors, control systems and mission-specific equipment into a coordinated package.

That distinction is important as military organizations increasingly look at autonomy as a way to reduce human exposure to battlefield hazards.

The company says its approach has been developed through work with combat engineers, military research organizations and government programs. Overland AI also says it has conducted more than 25 government demonstrations involving organizations including DARPA, the Defense Innovation Unit and the U.S. Army.

This operational experience appears to have been an important factor in the company’s selection.

The Overland AI contract comes as the U.S. military is accelerating the integration of artificial intelligence and autonomous systems across multiple domains.

The Army has increasingly emphasized systems capable of operating with greater independence, while recent programs have explored autonomous intelligence, sensing, targeting, logistics and maneuver.

Oak Ridge National Laboratory (ORNL) and industry partners are deploying advanced manufacturing technologies, such as heavy 3D printing and AI-driven automation, to repair broken US industrial supply chains.

For instance, inside an advanced Tennessee facility, a giant three-armed robot called MedUSA crafted a three-by-five-foot nuclear pressure vessel by guiding molten steel into place. Nearby was a large 12-foot model of an impact limiter, made from strong plastic instead of the usual wood. These were battle tests for the future of American industry.

America’s industrial sector is facing severe operational hurdles, driven by irregular supply chains and a shortage of skilled workers. These issues have created extreme delays in infrastructure sectors like hydropower, where acquiring major custom components can now take over ten years.

Over 350 leaders gathered at ORNL for the Materials and Manufacturing Innovation Days. They discussed ways to speed up the adoption of advanced manufacturing technologies.

The goal of M2IND was simple yet ambitious: remove the administrative and technical obstacles slowing advanced technology, and turn laboratory breakthroughs into factory-floor reality before time runs out.

“We want to make sure we have input from industry to help guide our research and ensure that the work at Oak Ridge and the Manufacturing Demonstration Facility (MDF) is valuable to U.S. manufacturers and takes into account the supply chains needed to commercialize new technologies,” said MDF Director Ryan Dehoff.

Fast-track advanced manufacturing

The event featured two large 3D-printed nuclear energy structures. Researchers worked with the University of Maine to print a glass-fiber polymer model of an impact limiter. This massive component protects spent nuclear fuel shipping containers during transport and replaces traditional wood designs. 

Scientists also used a three-arm robotic system to print a steel pressure vessel with complex domed geometry. This project shows how additive manufacturing can produce large metal parts that usually face production issues. 

Standard reactor vessels use large-scale forging and welding, which face severe domestic capacity limits. The 3D-printed vessel served as a test case to evaluate additive manufacturing as a viable alternative. 

Supply chain issues

Participants highlighted urgent needs for resilient supply chains, alternative materials, and faster part qualification. They also pushed to combine manufacturing with AI, robotics, and digital engineering. New partnerships were formed to quickly turn laboratory research into practical tools for manufacturers.

Representatives from four Department of Energy offices and defense officials joined leaders from energy, manufacturing, and technology sectors at the event. Together, several key obstacles facing American industry were identified. These challenges include limited local capacity for large metal parts and long component lead times. 

ORNL formed key industrial partnerships to bring research into practical factory use. Yet, as forum participants repeatedly emphasized, printing a part is only half the battle. Getting it certified for real-world use is where momentum often dies.

To bridge that gap, ORNL announced a joint research effort with Idaho National Laboratory to standardize digital qualification methods across different facilities

Researchers hope to give regulators the data needed to certify 3D-printed nuclear and energy components in months rather than years. They plan to do this using real-time sensors, machine learning, and defect-tracking algorithms during the build process.

With new partnerships signed with energy titans like SLB and Kairos Power, the push is clear. Urgency has replaced routine. America is racing to rebuild its domestic manufacturing core, starting right on the laboratory floor.

💾

Airline operations across the UK have reportedly resumed following an air traffic control failure that cancelled over 2,000 flights and left hundreds of thousands of travelers stranded at airports throughout the country.

British airports restarted operations on Wednesday, September 9, after the hours-long shutdown a day earlier. London-based aviation data and analytics company, Cirium, reported that around 1,750 flights were cancelled Tuesday, September 8.

The National Air Traffic Services (NATS), UK’s air traffic control body, reported that the departure disruption was caused by a “technical issue” in its flight processing system. Heathrow, Gatwick, Manchester, and Birmingham experienced the most severe delays.

Reports said more than 290 further flights were cancelled on Wednesday. The first day saw 30 percent of departures grounded. Cirium revealed that the figure fell to five percent the following day.

Martin Rolfe, NATS CEO, apologized to the travelers affected by the reported “IT failure”. However, Yvonne Moynihan, Wizz Air UK managing director, and Neal McMahon, Ryanair COO, have both called on Rolfe to resign over the recurring technical issues.

UK flights come to a halt

The issue began Tuesday following a technical problem at NATS, the company in charge of Britain’s air traffic control system. The failure led to the cancellation of more than 2,000 flights and stranded hundreds of thousands of passengers.

According to the media, Ryanair was among the hardest-hit carriers. The airline said approximately 150,000 of its passengers were affected and that it canceled more than 200 flights. British Airways, meanwhile, canceled or diverted around 100 flights, affecting tens of thousands of passengers.

Rolfe disclosed that the organization was investigating the incident and believed the problem was different from those behind previous failures. However, while he apologized for the disruption and said that a cyberattack had been ruled out, he was summoned to meet Transport Minister Heidi Alexander on Wednesday.

Alexander urged all passengers to “continue to check with airlines” to confirm if their flights were operating as scheduled. “I am seeking assurances that lessons will be learned and systems that support aviation are up to the job,” she said on the social media platform X.

Earlier issues spark concern

The disruption has nevertheless raised fresh questions over the resilience of the UK’s air traffic infrastructure. Tuesday’s outage was not the first major technical problem to hit Britain’s air traffic network in recent years.

A system failure in August 2023 caused widespread cancellations during one of the busiest travel periods of the year. It apparently cost airlines around USD 135 million (GBP 100 million). Another technical problem emerged in July 2025 when a radar-related fault disrupted flights at major airports, including Heathrow.

In an interview with Reuters, Michael O’Leary, Ryanair CEO, said the outage was a repeat of the 2023 shutdown caused by a rogue flight plan. “Fire him. He’s been there since 2015. He is useless,” O’Leary said.

The firm also issued a statement saying that over 65,000 of its passengers were affected by the outage. “How many more failures do passengers have to suffer before overpaid failure NATS CEO, Martin Rolfe, accepts responsibility for his repeated mismanagement and resigns,” McMahon noted. “Families traveling on holiday, people traveling for work and thousands of visitors to the UK have once again paid the price for NATS’ abhorrent failure,” he added.

Moynihan also joined Ryanair in calling on Rolfe to resign. “I think in any industry, when you’ve had three failures, you need to consider that three strikes and you’re out. So I think it’s certainly something that he needs to look at,” she said. “A third time and a third excuse is not enough anymore.”

Rolfe took full responsibility for the incident. He has been given one week to investigate what happened and report his findings to Alexander.

💾

Eclipse Space has unveiled three satellite platforms designed to let nations and companies build and operate their own space-based constellations, rather than leasing capacity from someone else’s network. The Redmond, Washington-based company was founded by engineers who previously built and scaled SpaceX’s Starlink system. It announced the platforms alongside a new power systems partnership with France’s Gama, ahead of World Space Business Week in Paris.

Eclipse designs and integrates everything a customer needs to run an independent constellation, including satellites, payloads, ground infrastructure, launch, and ongoing operations. CEO Derek Huerta said the company built its business around a different model than typical satellite operators. It gives customers complete infrastructure they can operate themselves, rather than infrastructure Eclipse continues to control.

Three platforms, one architecture

CitraSat, the first of the three platforms, handles direct-to-device communications, connecting standard mobile phones without requiring a dedicated dish or terminal. Its flat-panel design integrates a 3.5-meter S-band phased array, multi-gigabit E-band backhaul for ground connectivity, and V-band links for communicating between satellites in the same constellation.

The flat-stack design maximizes how many satellites fit on a single launch, and each unit carries an operational life of more than five years.

SliceSat targets high-throughput broadband instead, delivering high-speed connectivity directly from orbit without relying on terrestrial fiber networks. It packs four Ku-band phased arrays alongside multi-gigabit V/E-band feeder links and the same V-band inter-satellite connectivity found on CitraSat. The design aims at building resilient, low-latency broadband networks at regional or global scale.

SurgeSat extends the platform into an entirely new mission category: high-performance computing and data storage conducted directly in orbit. Designed for 100-kilowatt-class power output, SurgeSat is built to process and store data for large-scale AI and computing workloads. It does so without routing that data through third-party systems, keeping both the data and the infrastructure processing it fully under customer control. It shares its core architecture, inter-satellite networking, and manufacturing approach with the two communications platforms.

Power built to scale

Powering SurgeSat’s computing ambitions, along with Eclipse’s broader plans for high-capacity constellations, drove the company’s new partnership. That partnership is with Gama, a Paris-based developer of large deployable space structures and power systems. The two companies plan to combine Gama’s deployable solar array expertise with Eclipse’s spacecraft power systems. They’re targeting scalability up to more than 100 kilowatts and a first in-space demonstration in 2028.

Andrew Nutter, Gama’s CEO, said both companies share an engineering culture built around high-volume production, since many of the same engineers previously worked together scaling hardware for Starlink. Eclipse described the arrangement as an example of its broader fabless approach. The company owns system architecture, design, and software, while relying on specialized partners for individual subsystems rather than manufacturing every component in-house.

From stealth to platforms

Eclipse emerged from stealth in June 2026. The company was founded roughly a year earlier by CEO Derek Huerta, a former SpaceX satellite payload engineering manager, alongside other SpaceX alumni. The company has said it targets a minimum production rate of five satellites per day, with an integrated demonstration spacecraft planned for 2027. This week’s platform announcements mark the company’s first detailed public look at the spacecraft hardware underpinning that broader ambition.

Most rocket launches are watched on the way up. Engines ignite, the vehicle climbs through the atmosphere, and its stages separate as it heads toward space.

But for reusable rockets and spacecraft, reaching space is only one part of the mission. They must also find a way back to Earth.

The return is far more complicated than simply falling from the sky. The vehicle has to reduce its speed, survive the heat of re-entry, and remain stable as it descends. 

Depending on its design, it may then use engines, parachutes, or wings to complete the landing. That journey begins by establishing a controlled path back toward the planet.

Slowing down to return to Earth

A spacecraft remains in orbit not because it has escaped Earth’s gravity, but because it is moving sideways fast enough to keep falling around the planet. To come home, it must first slow down.

The spacecraft performs a deorbit burn by firing its engines against its direction of travel. This slows it enough to lower its orbit and enter the upper atmosphere, where drag further reduces its speed.

Reusable rocket boosters follow a different path because they usually separate from the vehicle before reaching orbit. 

After separation, a booster may perform a boost-back burn to reverse part of its trajectory and head toward land. Alternatively, it can continue downrange toward a landing platform at sea.

The booster then fires its engines again during its descent. These burns help control its speed and position it for the final approach.

Every maneuver must be precisely timed. Even a small error in the vehicle’s angle, speed, or orientation could cause it to miss its landing zone.

Surviving atmospheric re-entry

Entering Earth’s atmosphere at high speed creates extreme heat. As the spacecraft strikes the air, the gas in front of it compresses rapidly and heats up. This can produce a layer of intensely hot plasma around the vehicle.

Spacecraft use heat shields to survive these conditions. Capsules commonly carry ablative shields made from materials designed to char and gradually wear away. 

As the outer material is removed, it carries heat away and prevents the extreme temperatures from reaching the cabin.

The angle of entry is also critical. A spacecraft entering too steeply could experience excessive heating and deceleration. 

One approaching at an angle that is too shallow could travel beyond its planned landing area or briefly skip back out of the atmosphere.

The vehicle must also remain stable during this phase. Capsules adjust their orientation to produce limited lift and control their path. Reusable boosters use grid fins, which interact with the surrounding airflow to steer the descending rocket toward its target.

Completing the landing

Once atmospheric drag has removed most of the spacecraft’s speed, its landing system takes over. The method used depends on the vehicle’s design and mission.

Crew and cargo capsules generally deploy parachutes during the final stage of descent. The parachutes open in a sequence, gradually reducing the capsule’s speed. 

Some capsules splash down in the ocean and are collected by recovery teams, while others return to designated landing areas on land.

Reusable boosters rely on powered landings. Grid fins guide the rocket during descent before one or more engines restart for the final landing burn. This burn rapidly reduces the booster’s speed as its landing legs deploy.

Onboard computers continuously calculate the booster’s altitude, orientation, and velocity. They make rapid adjustments to engine thrust and steering, allowing the rocket to land vertically on a pad or a platform at sea.

Spaceplanes use another method. After re-entry, they glide through the atmosphere using wings and aerodynamic control surfaces before landing horizontally on a runway.

Returning to Earth is therefore not one action but a carefully coordinated sequence. Engines place the vehicle on the correct path, heat shields protect it during re-entry, and guidance systems control the descent. 

Parachutes, landing engines, or wings then complete the journey, bringing the spacecraft back to Earth intact and, in some cases, ready to fly again.

Launching a rocket requires enough power to break free from the ground, but returning one demands precise control at every stage. It is this combination of speed, heat management, and split-second navigation that turns a fall from space into a safe landing.

The Trump administration is putting $1.9 billion toward restarting a nuclear plant in Iowa as the U.S. works to keep up with rising electricity needs from artificial intelligence. On Tuesday, the Department of Energy said it had finalized a loan of up to $1.9 billion to help NextEra Energy bring the Duane Arnold Energy Center back online.

The 615-megawatt plant is set to restart in early 2029. It will be able to power almost 500,000 homes and support the expanding cloud and AI infrastructure in Iowa.

Federal cash backs nuclear comeback

Duane Arnold ran for over 45 years before closing in 2020. It was supposed to shut down later that year, but a derecho damaged the facility and forced it offline several months ahead of schedule.

NextEra now plans to bring the plant back online. The federal loan will help pay for the work needed to restore the facility and add its power back to the regional grid.

The Energy Department expects the project to boost electricity supplies and lower power costs. When it restarts, Duane Arnold will be Iowa’s only working nuclear power plant.

Deputy Energy Secretary James Danly said the financing “marks another step in advancing America’s nuclear renaissance.”

Google locks in nuclear power

Google is a key partner in the restart. In October 2025, Google and NextEra announced a deal where Google agreed to buy electricity from the plant through a multi-year contract.

This agreement gives NextEra a steady customer for the electricity the plant will produce. For Google, it means a reliable power source for its growing cloud and AI operations in Iowa.

“We expect Duane Arnold to be back online in early 2029, and the plant will provide more than 600 MW of clean, safe, “always-on” nuclear energy to the regional grid,” Google said in a blog post last year.

Google also pledged an additional $7 billion investment in Iowa over two years. The spending includes new and expanded infrastructure supporting its cloud and AI businesses.

AI pushes tech toward nuclear

The deal shows how tech companies are changing as AI increases electricity demand. Running advanced AI models needs big data centers, which use a lot of power.

Google is far from alone in seeking nuclear generation. A Microsoft-backed project to restart Pennsylvania’s Crane Clean Energy Center, formerly known as Three Mile Island Unit 1, received a $1 billion Energy Department loan in 2025.

Meta has also reached agreements supporting up to 6.6 gigawatts of new and existing nuclear capacity by 2035. Amazon, Microsoft, Google and Meta are all expanding data center infrastructure as AI workloads continue to grow.

Washington accelerates atomic power

The Duane Arnold financing also fits into the Trump administration’s broader nuclear strategy. President Donald Trump signed several executive orders in 2025 aimed at accelerating nuclear development in the U.S.

One order directed the Energy Department to halt most previous plans for diluting and disposing of surplus plutonium. It instead called for a program to process the material and make it available to private companies.

The Iowa project now adds another major nuclear investment to that push. If the restart stays on schedule, Duane Arnold will return 615 MW of generating capacity to the grid in 2029, giving Iowa a major new source of dependable electricity as demand from AI infrastructure continues climbing.

The Ministry of Science and Information and Communications Technology (MSIT) and the Korea Institute of Science and Technology Information (KISTI) have announced that the next-generation national supercomputer will be launched by December this year.

Dubbed Supercomputer No.6, the machine will be equipped with 8,496 graphics processing units (GPUs) and boast a 614 petaflops performance, likely pushing it into the Top 10 supercomputers in the world. 

Supercomputers are extremely fast and large computers that can carry out extremely complex mathematics to solve problems of engineering and science that regular computers cannot handle. From energy exploration to weather forecasting, space research to drug discovery, supercomputers are tasked with carrying out simulations that help us better understand these fields and help scientists make discoveries. 

Countries typically built these large-scale computers at their national institutes to make scientific breakthroughs and attain economic leadership, while also deploying them for national security measures. 

The world’s top supercomputers are listed in the TOP500, with the latest edition listing China’s LineShine and ending US dominance at the top of the list. South Korea has 15 supercomputers on the list, with KISTI’s Supercomputer No. 5 occupying the 109th spot. KISTI and MSIT aim to break into the top 10, with Supercomputer No. 6 coming this December. 

Supercomputer No.6

KISTI’s Supercomputer No. 6 is built around NVIDIA’s GraceHopper 200 chips that integrate the central processing units (CPUs) and GPU units. The upcoming unit will feature 8,496 GPUs and 9,936 CPUs, 205 petabytes (PB) of onboard storage, and networking speeds that exceed 400 Gbps. 

Theoretically, Supercomputer No. 6’s computing performance is 614 petaflops, which is a huge leap when compared to its predecessor, Supercomputer No. 5, “Nurion, unveiled in 2017 with computing performance of 25.7 petaflops. Apart from a 24x increase in theoretical computing performance, the successor supercomputer will also get a 10x increase in storage and 4x faster networking speed. 

MSIT and KISTI signed an agreement with Hewlett-Packard Enterprise for system installation last year. Post-installation inspections began last month, and performance and stability testing is underway as South Korea looks for a December launch. 

Once operational, the supercomputer will enable access to massive computing resources for researchers engaged in computational science, data simulation, and AI-based scientific research. Plans include redistributing the resources across three programs: National Strategy (30 percent), R&D Innovation (55 percent), and Industrial Innovation (15 percent). 

Not smooth sailing

The announcement of the new supercomputer and its likely availability in the near term is great news for researchers. However, the acquisition was not this smooth sailing. Soon after the Supercomputer No. 5 was introduced in 2018, South Korean researchers began looking for its successor as early as 2021. 

However, the boom in generative AI, which has matured with ChatGPT and Claude, has now caused a huge surge in demand for GPUs. As prices for GPUs skyrocketed, South Korea couldn’t secure the chips at budgeted prices and had to spend 448.3 billion won (~US $334.9 million), a 53 percent jump over its initial budget of 292.9 billion won (~ US$218.8 million) to make the supercomputer a reality. 

“Supercomputer No. 6 is core infrastructure that will drive a major leap in scientific and technological research productivity in the AI era,” said Kim Sung-soo, Director General of the R&D Policy Bureau at the Ministry of Science and ICT in a press release. 

“We expect it to generate tangible outcomes that contribute to securing national strategic technologies, including K-Moonshot and the 7 Major SEEDs, as well as research innovation across industry, academia, and research institutes.”

Bluecore Energy, a startup building small modular nuclear reactors (SMRs) on floating barges, announced an oversubscribed $50 million seed round led by Silverton Partners.

Within just nine months of launch, Bluecore Energy has rapidly accelerated its path toward floating nuclear power by securing a port terminal, receiving its first barge, and starting construction on a test reactor.

Designed specifically for commercial applications, the reactor incorporates multiple layers of protective shielding and continuous water-based cooling to ensure safe, resilient operation out at sea.

The company’s founder, Kofi Asante, announced in the LinkedIn post: “Our focus is simple. Create and deliver zero-emission energy as safely and quickly as possible. Over 3 billion people live within an hour of water. We want to power them all.” 

Powering roughly 15,000 homes

Bluecore Energy emerged from stealth on 21 July with $10 million in pre-seed funding to deploy floating SMRs directly on mobile barges. 

The startup is the first SMR developer based inside a major U.S. shipping hub, operating directly out of its headquarters at the Port of Long Beach, California. 

The goal is to build a 10 MWe water-cooled reactor system using established maritime nuclear physics enclosed in steel pressure vessels and concrete shielding. The reactor will be capable of powering roughly 15,000 homes and can operate for years on a single fueling.

Designed to be modular, multiple units can be chained together to scale up and deliver reliable, long-term power to major ports, utilities, AI data centers, offshore facilities, and energy-constrained coastal communities.

Bluecore’s initial target is its hometown Port of Long Beach — the first American port to publicly back floating nuclear energy. 

“Nuclear energy at sea is not a new idea. Our opportunity is to take lessons from decades of maritime nuclear experience and combine them with modern engineering to build systems for a new generation of commercial applications,” stated Asante, who is a former Uber Freight executive.

The startup has already secured and delivered its first physical barge alongside an electric test reactor to its port site. This will allow the engineering team to actively construct, test, and refine system components in real-world conditions. 

Alongside these hands-on development milestones, Bluecore is rapidly advancing its formal regulatory, design classification, and commercialization pathways by deepening its technical engagement with both the US Nuclear Regulatory Commission and the US Coast Guard.

Accelerating development

Its platform leverages offshore mobility to overcome land-based energy constraints. 

The company’s small modular reactors will be mounted on mobile barges that can be moored or anchored like offshore oil rigs. These will be connected directly to land infrastructure using high-capacity subsea power cables. Positioned out in open waters, the units naturally draw on the ocean for continuous cooling. 

According to TechCrunch, this seed round drew strong support from returning pre-seed backers, including Slauson & Co., Harlem Capital, and Ripple co-founder Chris Larsen. And high-profile new investors, such as Collab Capital, Kevin Hart’s HartBeat Ventures, and angel investors from Tesla, Uber, Amazon, and Google.

With the new capital, Bluecore will place its first order for nuclear fuel. The primary operational challenge facing the company is navigating hardware supply chain issues. 

The startup, staffed by veterans from SpaceX, Rivian, and Navy submarine units, is already undergoing joint design review with the US Nuclear Regulatory Commission and the Coast Guard.

💾

Europe has added another powerful machine to its growing supercomputing network with the inauguration of Arrhenius, a new system in Sweden capable of more than 60 petaflops, or over 60 million billion operations per second.

The supercomputer was inaugurated at Linköping University and will be available to researchers, companies and public-sector organizations across Europe. It is designed to tackle computationally intensive work ranging from climate simulations and materials research to life sciences, engineering and artificial intelligence.

A 60-petaflop machine for European research

Arrhenius is classified as a mid-range supercomputer, but its performance places it among the world’s most powerful machines. It ranked 42nd on the TOP500 list published in June 2026.

Its computing capacity is intended for workloads that would be impractical or take considerably longer on conventional computing systems. Researchers can use it to run complex simulations, analyze large datasets, and develop compute-intensive AI applications.

Potential applications include climate modeling, materials science, drug discovery, life sciences, and engineering. The system is also designed with security and data integrity in mind, allowing organizations to use it for proprietary product development and research involving sensitive personal data, according to the EuroHPC Joint Undertaking.

Arrhenius is already being used by European researchers. Users gained early access through the EuroHPC AI for Science access call before the official inauguration, while broader access is now available through regular EuroHPC access calls as well as benchmark and development modes.

Designed to deliver more computing for less energy

Raw performance is only one measure of a modern supercomputer. The amount of electricity required to achieve that performance has become increasingly important as computing systems grow larger.

According to EuroHPC, Arrhenius delivers more than 61 billion operations per watt. That placed it 24th on the Green500 list, which ranks supercomputers according to energy efficiency. The system was procured through a contract with HPE signed in July 2025, with an estimated total value of €68.5 million. 

Up to 35% of the cost is co-funded by the EuroHPC Joint Undertaking through the Digital Europe Program, with additional contributions from the Swedish Research Council to Sweden’s National Academic Infrastructure for Supercomputing (NAISS). EuroHPC and NAISS, hosted by Linköping University, jointly manage access to the machine.

Part of Europe’s push for computing sovereignty

Arrhenius is more than a standalone Swedish research computer. It joins 12 supercomputers already procured by the EuroHPC Joint Undertaking as Europe attempts to build computing infrastructure that it can control and access across the continent.

The initiative is also expanding beyond traditional high-performance computing. EuroHPC is overseeing the implementation of 19 AI Factories and 13 AI Factory Antennas across Europe, while also developing a European quantum-computing infrastructure that will integrate quantum technologies with conventional supercomputers.

The Swedish system is named after Carl Axel Arrhenius, the Swedish geologist and chemist who discovered the mineral gadolinite in 1787. Gadolinite was later found to contain several previously unknown elements belonging to the rare-earth group.

For European researchers and businesses, however, the machine’s more immediate significance is its computing capacity. By making tens of petaflops of processing power accessible across borders, Arrhenius gives scientists and industry another resource for workloads that conventional computing simply cannot keep pace with.