Taiwan’s Keelung District Prosecutors Office raided Super Micro Computer’s local office on Monday, widening an investigation into the alleged smuggling of Nvidia chips to China through the company’s servers. The raid also targeted the residences of six individuals and the sites of two other affiliated companies, according to Bloomberg. Super Micro shares fell more than nine percent on the news.

Taiwanese data centre operator Chief Telecom and Super Micro distributor Albatron Technology were also searched, according to a person familiar with the investigation cited by Bloomberg. Albatron confirmed in an exchange filing that it had been searched but did not explain why. Chief Telecom did not immediately comment.

Super Micro said in a statement that it is working closely with Taiwanese authorities and remains committed to protecting its technology and intellectual property. The company said its products continued to be targeted in the smuggling cases and that it is cooperating with law enforcement in Taiwan and other jurisdictions. Super Micro itself has not been charged in the investigation.

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 raid extends Taiwan’s first formal crackdown on AI chip diversion, which began in May when prosecutors detained three individuals accused of using forged documents to export Nvidia-equipped servers to China. The three suspects, including Super Micro co-founder Wally Liaw, allegedly routed at least one shipment through Japan before it reached the mainland. Around 50 servers were seized before they could leave the island.

Taiwan does not currently treat AI chip exports to China as a crime. Prosecutors have instead been charging suspected smugglers with violations of existing laws such as document forgery. Taipei is now considering criminalizing the exports themselves, which would give local prosecutors a more direct tool to pursue the illicit trade.

The case ties back to a broader scheme that US prosecutors have valued at roughly two and a half billion dollars. The US Department of Justice charged Super Micro co-founder Liaw and two others in March with conspiring to divert Nvidia-equipped servers to China through a front company in Southeast Asia, using heat guns to swap serial numbers and dummy servers to fool auditors. Liaw has pleaded not guilty, and the case is set for trial in November.

Nvidia CEO Jensen Huang addressed the smuggling problem last week, telling shareholders that data centres built with diverted chips are a dead end because the company will not provide support or repairs. The raid on Monday suggests Taiwan is increasingly willing to enforce that message from its end, extending the investigation beyond the original suspects and into the companies that handled the hardware.

Most iPhone leaks are predictable. You’ll see a case show up or an early leak showcase the mold of the upcoming iPhones. But the leak this time around is a bigger deal, since it has basically just revealed the device as it’s being tested.

A new post from leaker Ice Universe claims to show the iPhone 18 Pro undergoing a drop test. The clip itself is short, but it gives us a proper look at an unreleased Pro iPhone in a controlled test environment. This reveals that it is durable enough to handle a basic fall, though the thickness is still surprising, and the weight is still unknown.

Why this is bigger than your typical early leak

The drop-test clip appears tied to a much larger data leak involving Tata Electronics, one of Apple’s key manufacturing partners in India. Reuters reports that files posted on the dark web included iPhone 18 Pro supplier lists, component maps, and photos from drop tests at Tata facilities.

A leaked exterior photo can spoil a design, while supplier maps and component lists can reveal how the phone is built and who is building it. It can even reveal which companies are tied to key parts such as chips, batteries, and camera components. Just like the current iPhone 17 Pro models, the next-gen Pro iPhone also features the same three-camera layout and large camera island. The drop test is being performed on a gray color variant.

The iPhone 18 Pro may be thick for a reason

The claimed drop-test footage does not give us a proper spec sheet. But Apple’s Pro models are expected to get major internal upgrades, with the stolen Tata files mentioning logic board schematics, A20 Pro data sheets, and references to Apple’s C2 modem.

Though its clear that the iPhone 18 Pro isn’t the highlight this year. The big new thing is likely Apple’s first foldable iPhone, which could be called the iPhone Ultra.

Base44, the vibe coding platform that Wix acquired for $80 million just one year ago — when the company was barely six months old and had a team of eight — has started rolling out its own AI model to support its users in creating apps with natural language.

The move comes as the discussion in AI circles has intensified over whether frontier models are best suited for all use cases. A related question is whether businesses built on top of someone else’s models are truly defensible long-term. The latest move of Base44, based in Tel Aviv, speaks to both.

While its custom LLM is only just rolling out, Base44 hopes that it will eventually outperform frontier models. According to its founder, Maor Shlomo, “training and owning the model as part of [our] entire stack allows us a lot more optimizations on latency, cost, and efficiency.”

At first glance, this could be a way to stay ahead of competitors such as Swedish startup Lovable, which reached unicorn status in its Series A round last summer and that relies on external LLMs. However, Shlomo expects that others will train their own models — “at least the players that have gotten enough scale and velocity to have enough data.” 

According to Jonathan Userovici, a general partner at VC firm Headline — whose portfolio includes AI companies like Mistral AI, but not Base44 — data is one of three key ingredients of defensibility for AI startups, alongside distribution and tech stack. 

The upshot is that players with strong brands are now leaning into their data and infrastructure to increase their defensibility, and Base44 fits that pattern. The company says the first iteration of its LLM, Base1, was developed and trained on a dataset generated from “tens of millions of real user interactions on the platform.” 

This dataset will keep on growing with the company; but so will its rivals’. The bigger competition may not be vibe-coding startups at all but instead come from frontier AI labs that are getting closer to Base44’s home turf — Cursor and Grok’s parent company xAI now both belong to SpaceX, and Claude Code has become a vibe coding player in its own right. 

This gives Anthropic and other foundational AI providers access to data and feedback loops they can use to improve models for app creation, but Shlomo thinks specialization gives Base44 a leg up. “Models are progressing, but they’ll stay very general in what they can do,” he predicted.

Userovici, for his part, cautioned against underestimating frontier models, citing the example of the legal tech startup Harvey, which abandoned plans to train its own model. He doesn’t expect applied AI companies to become frontier labs en masse but frames Base44’s move in a broader context — one in which inference costs have become a meaningful part of the equation.

That cost pressure, Userovici says, has driven change that enterprise customers are now demanding. “They don’t necessarily see a [return on investment] when using the latest models for all use cases, so an entire infrastructure is being set up to do orchestration and optimization to select the right models for them so that costs don’t skyrocket while maintaining the same or similar performance across the majority of use cases.”

Enterprise companies still are a minority among the audience of the vibe coding platforms, but they represent a growing share of platform revenue, and users of all sizes are starting to express concerns over the cost of using AI. Base44’s decision to develop its own LLM stemmed from multiple factors, but cost reduction is likely among the benefits.

“We want to get a model that is going to be more aligned to what we think is the right thing, is going to be more optimized to what we see users like in terms of the results we’re getting, and is going to be faster and cheaper for customers eventually than using the frontier models like Opus,” Shlomo said.

As for Base44 itself, cost reduction isn’t as clear cut. In a press release, the company explained that “ownership of the model gives Base44 direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time.” 

Even with a delayed payoff, improved margins would be good news for Base44’s parent company, which recently announced it would lay off 20% of its workforce. In contrast, Base44 has been growing in headcount since the acquisition — and announced it had passed $100 million in annual recurring revenue a few months ago.

That’s still less than Lovable, which said it hit $500 million in ARR earlier this month. But Shlomo is betting that the “huge engineering effort” to develop Base1 will cement Base44’s positioning as the “only vertically integrated vibe-coding application — meaning, in Userovici’s terms, a player that owns its distribution, data, and infrastructure all at once.

This article was updated to correct Base44’s location.

As a parent, I would have readily paid for access to a tool that could draw from evidence-based ideas for my child’s specific challenges.

Michele Van Valey has had a rather unusual entry into science and research. She said she initially intended to begin a master’s degree following her bachelor’s in English, but instead “accidentally landed a waitressing job in the heart of a burgeoning music scene”.

She worked on several creative projects during this time, spanning music documentaries and indie record labels.

Later, her life turned towards yoga following some health issues. “I spent the next decade in more unconventional education studying the body – yoga therapy, Pilates, massage therapy, nutrition and meditation,” she says.

She returned to formal education with a masters in ‘mindfulness based interventions’ at University College Dublin (UCD) in 2019, which eventually evolved into another masters in counselling and psychotherapy at IICP.

Her masters dissertation led her to her current doctoral programme at UCD studying children’s mental health. She is also a psychotherapist by practice.

What inspired you to become a researcher?

Growing up in California, I owned my first Stanford University sweatshirt by the time I was 8 years old. My father was a teacher and higher education was instilled in me from a very young age. I did not know what I wanted to do in college but I enjoyed words, stories and writing so I opted for an English degree.

Drawn to contemporary realism, I met the Shakespeare requirement reluctantly until Professor Corum shared his animated interpretation of the wind instruments in Othello.

Two hundred and fifty freshmen held their collective breath, rapt, as he described the bawdy possibilities. “Am I reaching?” he asked. The room exploded. And so, my interest in words evolved into an interest in language; framing, meaning, tension, subtext, and all that is not being said. Qualitative research fits well for me.

Years later, Professor Gardiner introduced me to the work of anthropologist, Maria Gimbutas and asked me to write a feminist take on Chaucer’s Legend of Good Women for my thesis (so much for contemporary realism). Crossing disciplines, mapping historical landscapes and the process of discovery lit me up. After four years, I was ready for a break in study but always knew I would be returning. Joining the research team at UCD is a lucky opportunity that feels a lot like home.

Can you tell us about the research you’re currently working on?

The PhD I’m working toward is a doctoral scholarship in child mental health and the digital age, supported by UCD Foundation through a charitable donation from Cycle Against Suicide.

It came about after my friends and I put together a series of events hosted by UCD called, Neuroconvergence, where we brought neurodivergent voices together as a way to learn from each other, play together and perhaps even move the policy needle.

Dr Blanaid Gavin participated on a panel with me and I shared the dissertation I had written with her. That was the lucky coincidence. I didn’t know that she was compiling a team of researchers to design digital tools for parents and families in need of support.

I had personal experience of long waitlists to access services and the expense of paying for them privately. I also knew that there were tens of thousands of parents in Facebook groups advising each other around how to support their children when there was little chance of accessing a professional. The idea of an evidence-based parenting tool that offered in the moment support to families waiting for professional access really appealed to me. It took me longer to warm to the AI piece.

My project will become an AI-enabled, adaptable parenting tool that can draw from various empirically evidenced, theoretical disciplines in the moment of need. A cursory look at the literature revealed that most clinical, AI informed applications rest on behaviourist interventions. The physiological responses which often fuel behaviour is where my area of interest lies. So, neurobiology and the relational sciences will be cantered in my design.

Despite my broad reservations about using AI (sycophant-y, Tech-lords, plagiarism and water consumption), I know it’s coming and my individual protest will have no impact. The literature on LLMs [large language models] suggests they are unreliable therapists without constraint or oversight, so designing and testing for safety and efficacy will be important.

Perhaps contributing to the growing body of research that calls for ethical guardrails is a more effective use of my voice. As Dario Amodei of Anthropic recently pointed out to Oprah, “we can’t stop the train but we can steer it”. I’m only a few months in, so the biggest evolution has been integrating my friend Claude into the team. There is a lot more to AI than chat and I have become enchanted by the possibilities. Let’s hope we keep the train from jumping track.

In your opinion, why is your research important?

Parents are desperate for advice, especially when they have a child who is struggling in some way. Often the need is immediate, thus the success of Facebook groups for information sharing.

Someone is always there to offer ideas and solidarity. It’s a wonderful resource. Yet nearly every parent there is waiting on services because people want to hear from trained professionals when it comes to their child’s mental health.

Additionally, western culture leans heavily toward behaviourist approaches due to the large body of evidence that supports them. Many families have benefitted from helping a child face their difficulty to overcome it.

Sometimes, however, there is something systemic or traumatic prompting the behaviour that requires a different lens. Parents who are trying to help their children while waiting for professional support will benefit from access to a variety of empirically evidenced ideas to further their understanding of their child’s experience.

What commercial applications do you foresee for your research?

There are only a handful of parenting apps in the market that are supported by research, most of them behaviour focused. Very few, if any, integrate Interpersonal Neurobiology.

As a parent, I would have readily paid for access to a tool that could draw from evidence-based ideas for my child’s specific challenges.

Addressing thoughts and changing behaviours can be helpful. It is also possible that there may be other areas to investigate. Access to different ideas for effective, individual decision making will be invaluable to families.

What are some of the biggest challenges you face as a researcher in your field?

Speed. The volume of information coming out about AI is relentless. Staying organised in my work and balancing that with my role as a parent is also top of mind.

Are there any common misconceptions about this area of research? How would you address them?

I suppose people like myself will be worried about AI misuse, governance, plagiarism and climate concerns. It might sound like an oxymoron to create a relational AI tool, but what if it could help families access calm connection in difficult moments? What if tech brought us closer to understanding each other rather than pulling us in different directions? This feels worthy of investigation.

What are some of the areas of research you’d like to see tackled in the years ahead?

With respect to AI the field is moving fast. It’s actually hard to know how it will be researched at all unless AI is employed to speed things along. Gathering and analysing data takes years.

With respect to parenting, I hope that more families will learn about attachment needs, stress reactivity and coregulation. Every tier on my educational journey has taught me that healing is relational. The late Psychologist, John Welwood, even proposed that relationship is “the leading edge of human evolution at this time in history.” I’d love to see much more work in this area.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Tesla is facing an unusual security problem in the US, and it is happening before many of its batteries even make it onto the road. According to an investigation by WIRED, multiple truckloads of Tesla batteries have allegedly been stolen directly from the company’s Nevada Gigafactory, highlighting a growing wave of organised cargo theft targeting high-value technology shipments.

Cargo theft is becoming a serious problem for Tesla

The report claims that at least nine major suspected cargo thefts took place at Tesla’s Nevada battery factory in January alone. Investigators say the issue is much larger than those incidents, with authorities tracking at least 17 cargo theft cases involving Tesla and other businesses in Nevada’s Storey County this year.

Storey County Sheriff’s Detective Sam Hatley described the situation as “an epidemic,” suggesting the recorded cases likely represent only a fraction of the total number of thefts. Industry researchers estimate cargo theft losses in the US have nearly doubled between 2022 and 2024, costing businesses close to $18 million every day. Electronic components and batteries have become especially attractive targets because of their high resale value.

The investigation also reveals that some of the early thefts exploited weaknesses in transport verification procedures. Authorities believe organised groups took advantage of fake identities, illegitimate freight carriers, and gaps in logistics security to collect shipments before legitimate transport companies arrived.

Tesla has reportedly responded by tightening security at its Nevada Gigafactory, including stricter driver identity verification at facility gates. Investigators say the changes have reduced the frequency of successful theft attempts.

Powerwall batteries, GPS trackers and multiple arrests

One of the biggest incidents reportedly involved two trailers carrying Tesla Powerwall residential battery systems worth more than $475,000 each. The trailers were allegedly collected by an illegitimate logistics company before later being recovered hundreds of miles away, although the cargo had already disappeared.

Investigators also uncovered several additional cases involving trailers carrying roughly half a million dollars’ worth of Powerwall batteries. In some instances, GPS trackers helped authorities locate stolen trailers, while detectives even installed their own tracking devices in an effort to catch suspects returning for abandoned cargo.

The investigation eventually led to the arrest of three suspects following another attempted theft in late January. Prosecutors allege the group travelled from California using forged commercial driver’s licences to collect Tesla shipments.

The report also notes that lawmakers are beginning to respond to the broader rise in cargo theft. A bipartisan bill recently passed the US House of Representatives aims to strengthen enforcement against organised retail and cargo theft while improving coordination between law enforcement agencies.

For Tesla, the incidents underline that protecting electric vehicle batteries no longer ends at the factory floor. Increasingly, securing the supply chain has become just as important as building the batteries themselves.

Verizon will pay BT an equalisation payment of $625m for the joint venture.

UK’s BT Group and US’ Verizon have decided to combine their international units into a joint venture aimed at multinational connectivity. The parent companies would be better able to focus on their domestic markets, while providing support to the new venture, they said.

The 50:50 joint venture between the communication giants will serve more than 3,000 customers across 180 countries, representing some $4bn in combined annual revenue. Verizon will pay BT an equalisation payment of $625m, the companies said in a joint statement.

The two companies aim to support local compliance and sovereignty needs with their joint venture. Customers are promised “secure and resilient” connectivity, designed for data, operational and regulatory requirements.

The joint venture will be led by Martijn Blanken; former CEO at Neo Space Group and Exa Infrastructure, and a current non-executive director at the Cordiant Capital-owned Irish investment vehicle Speed Fibre Group.

Clive Selley, appointed to his position in April, will continue to lead BT International as CEO, while Verizon’s leadership also remains unchanged.

Speed Fibre Group purchased BT’s Irish wholesale and enterprise business unit in a €22m deal last year.

“Customers will benefit from new, secure and resilient connectivity platforms, which are designed for the age of AI and sovereign where it matters,” said BT Group CEO Allison Kirkby.

“It will create new opportunities for our people and long-term value for our owners. Today’s announcement marks a major milestone for BT International, and an important step forward for BT as a whole, as we deliver on our UK-focused strategy.”

Dan Schulman, the CEO of Verizon, added: “Our international customers require secure, flexible connectivity that works seamlessly across borders and cloud environments.

“When we thought about how to best support them, this joint venture was the clear answer: a cutting-edge, AI-ready and secure platform run by a single global organization dedicated to their needs.

“At the same time, our relationship with those customers will stay equally strong as we continue to directly provide them with the connectivity they need in the US”

Verizon shares fell 5.2pc yesterday (29 June) following the news, rising marginally in after-hours trading, while BT share prices rose 0.6pc.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Looking for the most recent Mini Crossword answer? Click here for today’s Mini Crossword hints, as well as our daily answers and hints for The New York Times Wordle, Strands, Connections and Connections: Sports Edition puzzles.

Need some help with today’s Mini Crossword? I thought 5-Across was actually an interesting fact, and one I didn’t know. Read on for all the answers. And if you could use some hints and guidance for daily solving, check out our Mini Crossword tips.

If you’re looking for today’s Wordle, Connections, Connections: Sports Edition and Strands answers, you can visit CNET’s NYT puzzle hints page.

Read more: Tips and Tricks for Solving The New York Times Mini Crossword

Let’s get to those Mini Crossword clues and answers.

Mini across clues and answers

1A clue: Big party
Answer: BASH

5A clue: Pricey variety of beef that translates to “Japanese cow”
Answer: WAGYU

6A clue: Unwelcome sound at 7 a.m., perhaps
Answer: ALARM

7A clue: Really make excited
Answer: REVUP

8A clue: Get ready (for)
Answer: PREP

Mini down clues and answers

1D clue: Farm machine that bundles hay
Answer: BALER

2D clue: Plant used to make mezcal
Answer: AGAVE

3D clue: Pancake topper
Answer: SYRUP

4D clue: Prominent feature of a camel
Answer: HUMP

5D clue: Bend out of shape, as wood
Answer: WARP

Swift has watched the sky since 2004, catching some of the universe’s biggest explosions. Now it is sinking, and time is short. NASA is paying Katalyst Space Technologies about $30mn to save it, the Associated Press reported. Liftoff could come as early as Tuesday.

The plan sounds simple and is anything but. Reach a satellite nobody designed for capture, grab it, and lift it higher.

Why Swift is falling

Every satellite in low orbit fights a slow drag from the thin air up there. The Sun makes it worse. Intense solar activity has puffed up the atmosphere, and the extra drag now pulls Swift down faster than NASA expected.

The telescope now orbits at about 360km. Left alone, it would drop below 300km by October, past the point where a rescue could still work. After that comes re-entry and a fiery end for a working observatory. NASA has already switched off Swift’s instruments to slow the fall, and science observations stopped in February.

That would be a real loss. Swift ranks among the fastest eyes on the sky, swinging within minutes onto gamma-ray bursts, the brief, violent flares that mark dying stars and colliding neutron stars. “If we let Swift reenter, we would lose that telescope,” NASA science chief Nicky Fox told the AP. “We don’t currently have the budget to build another one to replace that.”

The grab

Katalyst’s answer is Link, an autonomous spacecraft about the size of a small fridge with a 12-metre solar wingspan. It carries three arms, each tipped with two pinching grippers. It rides up on an air-launched Pegasus rocket, dropped from a plane over the Marshall Islands in the Pacific.

From there it has to chase down its target. NASA expects Link to take about a month to reach the 1.4-tonne observatory and grab it, then a further couple of months to raise the orbit from roughly 360km to about 600km. If it works, Swift could be back at work by September.

The catch is the hard part. Swift has no docking port and no grip points, because nobody designed it for servicing. Astronauts once fixed Hubble by hand, but that took the space shuttle and a crew. This time a robot works alone.

A new kind of mission

The speed of it is striking. NASA signed the contract only last September with two instructions: hurry, and do not make things worse. Nine months later, Katalyst is ready to fly.

It is also a first for the US. China nudged a dead satellite into a higher graveyard orbit in 2022, but catching a working telescope that was never built to be caught is a harder job. “No one thought it was going to be possible,” said Shawn Domagal-Goldman, NASA’s astrophysics director.

That matters far beyond Swift. A young industry wants to service, refuel and move satellites in orbit rather than let them die. A real rescue, on a real deadline, is the proof such firms have been waiting for.

Why it matters

The maths is part of the appeal. Thirty million dollars is a fraction of the cost of building and launching a fresh space telescope. If a tug can add years to a healthy instrument, the case for saving hardware over scrapping it gets stronger.

Hubble could be next. Katalyst says a bigger robot, due to fly next year, could reach satellites far higher up and give the ageing Hubble its own boost around 2028. Further out, the firm imagines fleets of orbital robots fixing, fuelling and even building in space.

There is a tidier future hiding in this too. Today most spacecraft simply fall and burn when they reach the end of their lives. A working tug fleet could lift the valuable ones, deorbit the dead ones on purpose, and start to clear the junk crowding low orbit.

For now, all of that rests on one launch and one delicate grab. NASA and Katalyst will know within months whether Swift keeps watching the cosmos or becomes a cautionary tale. The countdown has already started.