The race to secure electricity for AI models has reached new heights: Meta has signed an agreement with the startup Overview Energy that could see a thousand satellites beam infrared light to solar farms that power data centers at night.

In 2024, Meta’s data centers used more than 18,000 gigawatt-hours of electricity—roughly enough to power more than 1.7 million American homes for a year—and its need for compute power is only increasing. The company has committed to building 30 gigawatts of renewable power sources, with a focus on industrial-scale solar power plants.

Typically, data centers turning to solar power must either invest in battery storage or rely on other generation sources to operate at night.

Overview, a four-year-old, Ashburn, Virginia, outfit that emerged from stealth in December, has a different solution: The company is developing spacecraft that collect plentiful solar power in space. It then plans to convert that energy to near-infrared light and beam it at sufficiently large solar farms—on the order of hundreds of megawatts—which can convert that light to electricity.

By using a wide, infrared beam to power existing terrestrial solar infrastructure, Overview thinks it can sidestep the technological challenges and safety and regulatory issues that bedevil plans to transmit power to Earth through high-power lasers or microwave beams. CEO Marc Berte says you’ll be able to stare right into his satellite’s beam with no ill effects.

The technology would increase the return on investment from building solar farms and reduce reliance on fossil fuels — if it can be deployed at scale.

Overview says it has already demonstrated power transmission to the ground from an aircraft, and is planning to launch a satellite to low Earth orbit in January 2028 to perform its first power transmission from space.

In today’s announcement, Meta said it signed the first capacity reservation agreement with Overview to receive up to 1 gigawatt of power from the company’s spacecraft, although it’s not clear if any money changed hands. Overview developed a new metric for this contract, megawatt photons, which is the amount of light required to generate a megawatt of electricity.

Berte expects to begin launching the satellites that would fulfill that commitment in 2030, with a goal of flying 1,000 spacecraft in geosynchronous orbit, a high orbit in which each satellite remains fixed above the same point on Earth. He expects each of the company’s spacecraft to provide power from space for more than 10 years.

Once in space, Berte says the fleet of spacecraft will be able to cover about a third of the planet, with an initial deployment that will reach from the West Coast of the United States across to Western Europe. As the Earth rotates below and customer solar farms enter evening and night, Overview’s spacecraft should boost their electrical generation with additional light from space.

Berte sees opportunity in combining both generation and transmission, with the flexibility to deliver power to solar farms wherever and whenever it is most valuable.

“There’s a big difference between being in any one energy market, and being in all of the energy markets,” Berte told TechCrunch.

The round is led by Headline, with new investors Bullhound Capital, Felix Capital, and Daphni. Valuation is undisclosed. Sereact’s vision language action models already run in BMW, Daimler Truck, and logistics customers. The $110M is more than four times the €25M Series A raised just 15 months ago.

Sereact, the Stuttgart-based AI robotics software company, has raised $110 million in a Series B round led by Headline, the international venture firm with offices in Berlin, San Francisco, and Paris. New investors Bullhound Capital, Felix Capital, and Daphni joined alongside several existing backers.

The company declined to disclose its valuation. Funds will be used to develop Sereact’s core AI model, one that “makes robots smarter and more adaptable to different tasks”, and to scale deployment across logistics, manufacturing, and, increasingly, humanoid robot platforms.

Sereact was founded in 2021 by Ralf Gulde (CEO) and Marc Tuscher (CTO), both former AI researchers, at the University of Stuttgart.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

The company’s technical approach is grounded in Vision Language Action Models (VLAMs): AI systems that combine computer vision, natural language understanding, and action planning into a single model, allowing robots to perceive their environment, interpret instructions, and execute physical tasks without requiring complex programming or environment-specific pre-training.

A robot picking a fragile object can, in principle, evaluate whether its planned grip will cause damage before its gripper closes.That capability is the meaningful differentiator in a market where most industrial robotics still operate on pre-programmed sequences that assume a controlled, predictable environment.

Warehouses, manufacturing floors, and logistics facilities are not controlled environments: objects arrive in unpredictable orientations, packaging varies, and edge cases are constant.

Sereact’s software-first approach, explicitly positioned against the hardware-first strategies of most robotics companies, is designed to make robots adaptable to this variation without requiring engineers to reprogram them for each new object type or layout change.

In Gulde’s formulation from the Series A announcement: “with our technology, robots act situationally rather than following rigidly programmed sequences.”

The commercial record behind the Series B is substantive. Customers include BMW Group, Daimler Truck, the Dutch e-commerce fulfilment company Bol, and logistics specialists MS Direct and Active Ants.

The deployment at automotive OEMs is editorially significant: BMW and Daimler Truck are not pilots or proof-of-concepts, they are production environments where the economic cost of a robot failure is measured in line stoppages.

Sereact’s technology reaching production at that tier of customer is the validation signal that distinguishes it from the large number of AI robotics companies still operating at the demonstration stage.

The funding trajectory makes the ambition of the round clear. Sereact raised $5 million in seed funding in 2023, €25 million (approximately $26 million) in a Series A led by Creandum in January 2025, and now $110 million in April 2026, a more than four-times step-up from the Series A in fifteen months.

Creandum’s Johan Brenner captured the investment thesis at the Series A: “most AI robotics companies are currently hardware-first. What sets Sereact apart is their software-first, foundational approach which means they have the potential to become the brain of any robot that requires vision and autonomous capabilities.”

That thesis, a software-first robotics intelligence layer deployable across any hardware platform, is essentially the same thesis that has made Mobileye valuable in autonomous vehicles and that Nvidia is pursuing through its Isaac robotics platform: the idea that the highest-margin position in robotics is not the robot itself but the intelligence running it.

The broader market context is accelerating. Humanoid robot deployments by Figure AI, Boston Dynamics, and Unitree are moving from controlled tests to commercial production at warehouse and manufacturing customers.

The global humanoid robot market, valued at under $1 billion in 2023, is projected to exceed $38 billion by 2030. Tesla’s Optimus production ramp, targeting volume output from July 2026, will require robotics intelligence software at scale.

Sereact’s explicit intention, stated at the Series A, to expand beyond logistics into humanoid robot platforms positions it to compete for that market. The $110 million Series B is the capital raise that makes that expansion credible.

Do you remember 3D TVs? They had a roughly six-year run from 2010 to 2016, a timespan that witnessed the demise of plasma TVs and the ascent of OLED TV technology. At that same time, there was a 3D movie boom in theaters, with nearly every major release getting a 3D-format screening, whether the content benefited from the addition of a third dimension or not.

Booms are usually followed by a bust, and these days most folks don’t even give a second thought to 3D movies. In an ironic twist, however, recent advances in display tech have created fertile ground for 3D viewing, particularly for watching 3D movies at home.

To 3D or not to 3D

I’ll admit to having geeked out on 3D during the last boom period, partly because I was regularly reviewing TVs at the time, and also because I had young kids who wanted to see the new animated movie releases, most of which were shown in 3D. Consequently, I spent many a weekend wearing 3D glasses in theaters, and also amassed a decent-sized Blu-ray 3D disc library for home viewing.

Looking back at that time, my recollection is that the 3D kinda sucked, both at the theater and at home. In theaters, it was the pre-laser projector era, when the brightness of a typical digital cinema projector was significantly lower than what you get in today’s venues. Put on 3D glasses, and the picture put out by that already brightness-challenged projector could look almost unacceptably dim.

Along with being dim, many movies shown in 3D weren’t actually shot using 3D cameras but instead converted to 3D. Not surprisingly, any 3D effect in these titles was minimal, or even accidental.

On the home front, the 3D TVs of the time had smaller screen sizes and were significantly less bright than today’s models. Today, a typical lineup from any TV brand ranges from 55 inches up to 98 inches, and peak brightness for higher-end models is in the 2,000-3,000 nits range. Back in 2015, screens generally maxed out at 65 inches, with 55 inches being a more common size. As for brightness, one of the last 3D TVs I measured, a Panasonic LCD model, topped out at a mere 132 nits.

(Looking back at that 2015 review, I was reminded that Netflix at one time streamed 3D content, and that I had actually watched a Wim Wenders 3D movie on the Panasonic. Crazy.)

Most home projectors in that earlier era also supported 3D, which let you watch 3D movies on 100-inch-plus-size screens. But while a bigger screen provided a better and more immersive experience, the brightness of pre-laser, lamp-based home projectors was much less than what you can expect today. To cite one example, the Sony VPL-VW350ES, a 4K SXRD projector that cost $10,000 when I tested it in 2015, had a specified 1,500 lumens brightness. Today, the Sony Bravia Projector 7, the company’s entry-level 4K laser model, lists for the same price and has a specified 2,200 lumens brightness, a near 50% increase.

Surveying the 3D landscape

Today, there are still some 3D theatrical releases, though most are kids-oriented animated movies like the recent Pixar hit Hopper and The Super Mario Galaxy Movie. New Blu-ray 3D discs also continue to trickle out, with the latest entry in the Avatar franchise, Avatar: Fire and Ash, which is scheduled to ship in May 2026, being one notable release.

Anyone who owns a 3D-capable Blu-ray player and picks up Avatar: Fire and Ash on disc will find the 3D at home landscape much improved compared to the 3D boom years, and that comes down to one product category: projectors.

I recently got my hands on the Hisense PT1, a 3D-capable projector. The PT1 is an ultra short throw (UST) model designed for close placement to a screen and it supports screen sizes up to 150 inches. At $2,500, it’s affordably priced for a 4K UST projector, and its 2,500 lumens specified brightness exceeds that of considerably more expensive long throw projectors such as the Bravia Projector 7, a model that, despite its relatively high price, doesn’t offer 3D support.

Having the Hisense PT1 in-house gave me an opportunity to do something I’ve long been wanting to do: revisit my Blu-ray 3D disc collection to see how the experience holds up on a bigger, better, and brighter display than what I was used to in the old days. For my viewing, I paired the Hisense with a 100-inch Elite Pro AV Floor Riser Pro DarkUST 3 projection screen using an Oppo UDP-203 4K Blu-ray player as a source. Active 3D glasses were AWOL Vision DLP Link.

3D Revisited

To get the 3D TV party started, I selected three movies that I know were actually shot using a stereoscopic process and not upconverted to 3D: Avatar, Hugo, and Coraline. It didn’t hurt that these are all good movies – very good in the case of Coraline – and that I remembered them as having a pronounced 3D effect when I first watched them, both in the theater and at home.

Avatar is the king of modern 3D spectacles, and it was the disc I was most eager to check out on the Hisense projector. (The next chapter in the series, Avatar: The Way of Water, was also a 3D banger when I watched it at my local IMAX theater, and it looked great on an Apple Vision Pro headset that I briefly had on loan.)

To quickly sum things up, Avatar looked great on the system. The picture had strong contrast and bold color, especially in the luminescent forest scenes early on, and the 3D effects were even more pronounced than I remembered. There really was a sense of being there in the forest with those blue aliens. I found that brightness could have been better – I guess I got spoiled by watching Avatar: The Way of Water in optimal 3D viewing conditions – but the picture wasn’t so dim as to prevent me from enjoying it.

Hugo on the Hisense was another case where the 3D effects were more pronounced than I remembered – the picture was almost too three-dimensional. Unlike Avatar, Hugo is a good enough movie (it was directed by Martin Scorcese) that you don’t need 3D to fully enjoy it, so this is one case where I’d forego the third dimension to get the added picture brightness in 2D.

Coraline (2009) was the first stop-motion animated feature to be shot in 3D, and it received a 4K restoration and 3D theatrical re-release in 2024. Unfortunately, the remastered version didn’t make it to Blu-ray 3D, though a 4K disc version is available. 

Watched on the Hisense UST projector, my original Coraline disc lacked the 3D depth swagger of Avatar and Hugo, though there were some scenes that strongly benefited from the extra dimension, particularly the ones where Coraline crawls through the passage to and from the Other World. I had caught the remastered 2D version in theaters back in 2024, but watching it in 3D on a 100-inch screen at home was a compelling enough experience that I found myself sitting again through the entire movie.

Plowing through select scenes from the rest of my collection, I was surprised to see how good some of those discs, which I previously hadn’t been impressed by, looked on the big screen. Disney’s Tangled (which even has previews in 3D) and How to Train a Dragon both had subtle, yet impressive depth effects, and the IMAX documentary Deep Sea 3D was a visual delight, particularly the jellyfish segment. The Tsui Hark historical martial arts drama Flying Swords of Dragon Gate, in contrast, was almost too much in 3D given its non-stop over the top, gravity defying action scenes. And the same could be said for Pacific Rim, which had a near-wobbly look due to the shifting spatial perspectives.

Some things are better left in two dimensions.

A 3D TV revival?

The popularity of 3D movies may come and go, but the format will never completely disappear – it’s got way too much sticking power. What I don’t expect to see revived at any time in the near future is 3D TVs. Set manufacturers, and the viewing public, seem pretty much done with that concept.

Fortunately for 3D movie fans, 3D lives on in projectors. That’s not to say that all new projectors support 3D. In fact, Epson, Sony, and JVC have all dropped 3D support in their latest models. But other projector brands such as Hisense, Optoma, AWOL Vision/Valerion, and XGIMI are bullish on 3D, providing broad support for the format across their respective product ranges.

Looking back on my 3D experiment/nostalgia trip, the one thing I would have changed was to use a brighter projector. With a rated 2,500 lumens, the Hisense PT1 has good brightness for the price, and is a great value for an RGB laser projector. But the company’s flagship UST model, the Hisense L9Q, which is rated at 5,000 lumens, twice the brightness, would have been a better choice. The L9Q costs $6,000, and while that price might be high, it is apparently the cost of doing proper 3D business.

Where to buy:

Related Reading:

CNET just launched People’s Picks, a series of surveys where actual real-life humans like you vote for the products and services they use and can’t live without. For our first survey, we want you to weigh in on your favorite headphones and earbuds. We’ll choose winners based on your recommendations and publish a roundup featuring your top picks. So be sure to check back in a few weeks to see if your favorites made it to the list.

Watch this: AirPods Max 2 vs. Sony WH-1000XM6: Which Headphones Are Better?

09:43

Why we want to hear from you

Our writers and editors test hundreds of products each year, but your real-world experience with these devices is something we can’t replicate in our labs. You’ve used these headphones at the gym, on your commute to work and on long flights, and that perspective is invaluable. Your voice helps others know about the headphones or earbuds you love, too.

“I review a lot of headphones and earbuds for CNET, and there are plenty of great models from the top brands in this survey that I rate highly. I’m always curious about what models people ultimately choose and why, so I’m excited to get your feedback and learn the results of this survey,” says David Carnoy, CNET’s executive editor and headphones expert.

With our survey, we’ll collect answers from real-world users like you. The headphones and earbuds chosen through our 3-minute survey will be featured in our People’s Picks roundup of the top picks based on your recommendation.

Watch this: AirPods Pro Alternatives. Great Sound for Less, but Is That Enough? | All Things Mobile

10:51

Make your voice heard

Whether you swear by a pair of $25 earbuds or love a pair of high-end headphones, your pick counts. The survey takes just a few minutes to complete, and after we gather enough information, we’ll tally the results and publish the winners.

Not sure what to pick? Check out our list of the best headphones to revisit your favorites before voting.

The round amount is undisclosed. Backers include Nordic Makers, Emblem, the Lovable company itself, the Sequoia Scout Fund, and the Andreessen Horowitz Scout Fund. Lovable CEO Anton Osika personally endorsed the team. Atech lets users describe a hardware concept in natural language and receive a working prototype.

Atech, a Copenhagen-based AI hardware startup, has raised an undisclosed pre-seed round from Nordic Makers, Emblem, Lovable, the Sequoia Scout Fund, and the Andreessen Horowitz Scout Fund.

Founded by Vladimir Baran (CCO), Tomas Erik Harmer (CEO), and David Stålmarck (CTO), the company is building what it calls “vibe-engineering” for hardware, a platform that lets users describe a physical device concept in natural language and receive a working prototype, with all underlying technical complexity handled by the platform. 

The conceptual anchor is the parallel with what Lovable did for web application development. Lovable, the Swedish startup that allows non-developers to build full-stack web applications through natural language prompts, is valued at over $1 billion following rapid growth from its 2024 launch.

Anton Osika, Lovable’s CEO, backed Atech directly and provided the company’s most significant endorsement: “I am seeing the same patterns Lovable had but for hardware. I’m really excited to see Atech’s journey. The team is one of a kind.”

That institutional stamp of approval, from the founder of the software-vibe-coding category, is the most newsworthy element of the announcement and frames Atech explicitly as the hardware equivalent of an already-validated category.

The problem Atech is addressing is genuine and well-documented. Building a hardware prototype has traditionally required either years of specialised engineering expertise or significant capital to hire that expertise.

A developer can build and deploy a web application in a weekend using modern tools; the equivalent end-to-end hardware experience does not exist. This asymmetry has kept hardware innovation concentrated among a small group of specialists and within well-capitalised companies, while software development has been progressively democratised over the past decade through layers of abstraction that removed the need to understand compilers, memory management, or network protocols.

Harmer, Atech’s CEO, described the gap: “Software has an entire stack of tools that lets a teenager build an app in a weekend, hardware doesn’t, and we’re still working at the first level of abstraction. Atech is building the missing layers, so creating in the physical world can feel as fast and joyful as writing code.”

The investor syndicate is notable for a pre-seed round and warrants careful characterisation. The Sequoia Scout Fund and the Andreessen Horowitz Scout Fund are scout programmes operated by Sequoia Capital and a16z respectively, through which scouts, typically founders and operators in the funds’ networks, make small investments (often $5,000 to $100,000) in early-stage companies on behalf of the firm.

Scout fund participation does not constitute a direct investment by Sequoia or a16z in the traditional sense, and does not imply follow-on commitment from the parent firms. For Atech, the value of scout participation is primarily signalling, proximity to two of the world’s most prominent venture capital firms, rather than the capital amount itself.

Nordic Makers is a Copenhagen-based angel investor collective with deep ties to the Danish and broader Nordic startup ecosystem. Emblem is a European seed fund.

Lovable’s participation as a corporate backer, rather than just as an endorser, gives the round a strategic dimension: Lovable has a direct commercial interest in seeing hardware development democratised, as it would expand the surface area for Lovable-style interfaces.

The broader thesis Atech is operating within, which the company refers to as “Physical AI”, is gaining traction across the industry. Nvidia’s 2025 annual report framed Physical AI as its next major market opportunity: intelligent systems that sense, interact with, and act upon the physical world, including robotics, autonomous vehicles, drones, and industrial systems.

As these applications proliferate, the ability to rapidly prototype physical hardware becomes a competitive capability rather than a niche skill.

Whether a natural-language-to-prototype platform can genuinely close the hardware-software gap depends on how far the abstraction stack Atech is building can reach: PCB layout, component selection, firmware, and manufacturing considerations are all domains where the penalty for getting things wrong is a physical object that does not work, not a bug fix pushed in a pull request.

That is a harder problem than Lovable solved, and it is the question the company’s first customers will answer.

Elon Musk called violent crime in San Francisco “horrific” and moved the offices of his social media business X outside the city in 2024 because of safety and business considerations. Other local tech companies have attempted to address their security concerns by partnering directly with cops.

Airbnb and Salesforce are among businesses that for years have contracted San Francisco police to protect their offices on a regular basis, according to public records obtained by WIRED. Airbnb, for example, spent roughly $428,443 for the presence of uniformed, armed officers in 2024, the most recent year for which complete data was received. Salesforce shelled out about $727,907 through a security vendor. The payment amounts have not been previously reported.

Salesforce hired police to protect its offices in San Francisco’s tallest structure, known as Salesforce Tower, as well as a nearby building in the city’s busy downtown area. It also spent nearly $41,000 for officers at its TrailblazerDX 2024 conference hosted at the city’s convention center.

The security practices of tech companies in San Francisco have received renewed attention in recent weeks after a man allegedly threw a molotov cocktail toward the home of OpenAI CEO Sam Altman and tried to barge into the company’s headquarters by ramming a chair into the building’s glass doors. Authorities allege the suspect wrote a document criticizing AI technologies that outlined a goal to kill Altman and referenced the names of other AI executives. He is facing state and federal prosecution but hasn’t entered formal pleas yet.

OpenAI and Anthropic, the two leading generative AI model developers based in San Francisco, have not been regular customers of the city rent-a-cop program, according to police spokesperson Allison Maxie.

Salesforce, Anthropic, and Airbnb declined to comment. OpenAI did not respond to requests for comment.

The contracting program is known locally as 10B, which is also the section of the city code authorizing it. Any person, company, or organization that desires extra personnel or equipment for “law enforcement purposes” can request “such personnel to perform such services,” as long as the police chief signs off on it first. Under the law, companies pay the same hourly rates for officers that the city would, including overtime. In early 2024, the rate for a standard officer was $135 an hour during the day, while a lieutenant fetched up to nearly $190 at night, records show.

The program is often used by organizations hosting concerts, events, and conferences, as well as by sports teams that need additional security. The largest spender in 2024 was the San Francisco Giants baseball team, with a bill approaching $1.9 million. That year, at least four National Basketball Association teams paid a collective sum of about $16,500 for police escorts.

Several tech companies used the program on one-off bases in 2024. Records list OpenAI as paying $813.43 for unspecified coverage at the Asian Art Museum, Microsoft having a single bill of $1,622.16, and Zoox running a tab of $838.43. Occasional or one-time customers in prior years have included Affirm, Cruise, Datadog, and Fanatics.

Zoox spokesperson Marisa Wiggam said police protected a large offsite gathering for employees and that it is open to using the program again if a need arises. Microsoft and Affirm declined to comment. The other companies did not respond to requests for comment.

The program has been used more regularly by houses of worship, office building landlords, retail stores, and bank branches, including Apple, Bank of America, Best Buy, Bloomingdale’s, Chase, Lululemon, and Sephora, records show. The firm Security Industry Specialists paid over $1.2 million in 2024 for what police records described as coverage at three Apple stores, making it the year’s second-largest customer.

An estimated 80 percent of police departments across the country allow moonlighting by officers informally or through formal policies like San Francisco’s, according to a survey from over a decade ago by Seth Stoughton, faculty director of the University of South Carolina’s Excellence in Policing & Public Safety Program. Some cities have raised concerns about the conflicts of interest and liability risks posed by the arrangements. But agencies that permit them say they benefit community relations and officers’ wallets, Stoughton’s survey found.