TL;DR
Firmus and DayOne will build a 360MW Nvidia DSX AI Factory in Batam, Indonesia, with 170,000 chips and up to $30B in expected offtake over six years.
Firmus and DayOne will build a 360MW Nvidia DSX AI Factory in Batam, Indonesia, with 170,000 chips and up to $30B in expected offtake over six years.
Firmus Technologies, an Australian AI infrastructure company valued at $5.5 billion, will build its first data centre in Indonesia through an eight-year partnership with Nvidia. The 360-megawatt Nvidia DSX AI Factory campus in Batam, an island just off the coast of Singapore, is being developed with Singapore-based DayOne and is set to go live in Q1 2027.
Firmus will access up to 170,000 Nvidia AI accelerator chips through 2027 and 2028 via a revenue-sharing and credit-support agreement. The company expects $25 billion to $30 billion in committed offtake agreements during the first six years of the partnership, according to Bloomberg.
The Batam project will be a multi-tenant facility for AI-native customers, unlike Firmus’s Australian projects, which focus on hyperscaler clients. Co-CEO Tim Rosenfield told Bloomberg that market volatility around AI stocks is “largely irrelevant” to how the company is building its business. “We’re building our business based on demand that we’re seeing from customers and contracts that we’re closing,” he said.
Firmus began as a Bitcoin mining operation in Tasmania in 2019. It raised $505 million in April at a $5.5 billion valuation in a round led by Coatue Management and backed by Nvidia. The company has a pipeline of data centre projects across Australia and Singapore, including a deal with CDC Data Centers to develop up to 1.6 gigawatts across Australia by 2028. Asia-Pacific data centre investment has been accelerating sharply, with Blackstone-backed AirTrunk committing $30 billion to India alone.
Rosenfield declined to comment on IPO plans, though the company is widely expected to list this year. The deal adds to Nvidia’s expanding DSX programme, which partners with data centre operators to deploy GPU infrastructure on a revenue-sharing basis rather than requiring upfront purchase. For Indonesia, the campus positions Batam as a regional AI compute hub, leveraging its proximity to Singapore’s financial and tech ecosystem. Demand for AI compute across the region is so intense that even Google has resorted to renting GPUs from SpaceX.
Looking for a different day?
A new NYT Strands puzzle appears at midnight each day for your time zone – which means that some people are always playing ‘today’s game’ while others are playing ‘yesterday’s’. If you’re looking for Sunday’s puzzle instead then click here: NYT Strands hints and answers for Sunday, June 28 (game #847).
Strands is the NYT’s latest word game after the likes of Wordle, Spelling Bee and Connections – and it’s great fun. It can be difficult, though, so read on for my Strands hints.
Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc’s Wordle today page for the original viral word game.
SPOILER WARNING: Information about NYT Strands today is below, so don’t read on if you don’t want to know the answers.
• Today’s NYT Strands theme is… The mark of a good composer
Play any of these words to unlock the in-game hints system.
• Spangram has 12 letters
First side: left, 5th row
Last side: right, 6th row
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON’T WANT TO SEE THEM.
The answers to today’s Strands, game #848, are…
The theme was initially confusing — were we searching for composers like Mozart? — and even spotting the spangram MUSICSTAFF didn’t make any difference.
Musical staff? Is that a posh way to describe an orchestra?
However, after a hint, I realized we were searching for words associated with the composer’s craft of writing music.
That said, ACCIDENTAL was a new one for me, as I did not know this was a musical term and it took me ages to connect — it is a symbol placed immediately before a note that alters its pitch, temporarily raising or lowering it.
Strands is the NYT’s not-so-new-any-more word game, following Wordle and Connections. It’s now a fully fledged member of the NYT’s games stable that has been running for a year and which can be played on the NYT Games site on desktop or mobile.
I’ve got a full guide to how to play NYT Strands, complete with tips for solving it, so check that out if you’re struggling to beat it each day.
Dell‘s EOFY laptop discounts can be a little less obvious than a normal retailer sale, because there’s no before price shown after a drop. Fortunately, I track Dell pricing, so I can spot when an easily missed price drop is actually worth paying attention to.
That’s why the Dell 14 Plus caught my eye: it’s a compact 14-inch Windows laptop with a pretty grunty Ryzen AI 7 350 processor, 16GB RAM and a roomy 1TB SSD. Just as importantly, there’s a TechRadar coupon code available that brings the price down to just AU$1,139.05.
Our review said the Dell 14 Plus “delivers pretty much exactly what you need to get the job done” for everyday use, school work and general productivity.
What more could you want?
One of the reasons I like this little laptop is that for the price, it doesn’t shortchange you on memory or storage space. And it might not feel as important now, but that 1TB SSD means you can store loads of photos and files without worrying about running out of space.
Our review gave the Dell 14 Plus 4.1 stars, and the scorecard sums it up neatly: “a solid general-use and productivity notebook that’s great for work or school”. The value score is a top-notch 5 out of 5, while performance and battery life each score 4 out of 5.
There are some caveats, of course. The display is only 300 nits, so it’s better suited to an office, home desk or lecture room than bright outdoor use. Our review also noted the keyboard, trackpad and speakers could be better, so I wouldn’t buy this expecting a premium MacBook Air feel.
But considering the price, those trade-offs make perfect sense if you mainly need a portable Windows laptop that handles everyday work without any fuss.
There’s also a related Dell deal that may suit some readers better: the Dell 14 Plus 2-in-1 with a Ryzen AI 7 350 processor, 16GB RAM and a 1TB SSD is AU$1,299.10 with the same FUTUREAU5% code, down from the early June price of AU$1,799.
The 2-in-1 makes more sense if you like using a touchscreen. Personally, I find that being able to fold the display back into tablet or tent mode is super handy for things like handwritten notes, reading on the couch, marking up documents or watching movies on a plane.
If you want to check out more laptop options before deciding, our main EOFY laptop deals hub is chock full of laptop deals. It covers everything from more Dell deals to MacBooks, Windows laptops, 2-in-1s and gaming laptops, so it has something for everyone.
For even more Dell discounts, check out our TechRadar Dell coupons page, or try one of the codes below.
YouTuber Matthew Trahan has made a career out of 3D printing increasingly unusual things. He has printed musical instruments, bedroom furniture, and, in one particularly memorable video, himself.
His latest project is a full outfit, from shirt to shoes, belt to glasses, because apparently nobody told him 3D printers are for creating engineering prototypes or structures that aren’t otherwise feasible, not for fashion week.
Trahan’s checklist for his latest video included ten items: a shirt, shorts, shoes, socks, a belt, a hat, a wallet, a bag, a tie, and glasses. He couldn’t print all of them successfully, though.
The shorts, in particular, look like they belong to a Minecraft character. Some of the results were genuinely interesting, especially the Waveform shoe design by Stephen Drunks.
On the cost side, however, the numbers tell a complicated story. The filament came to about $100, which might sound like a bargain until you add the Prusa Core 1L printer he needed specifically to print the shorts, which costs $1,999. He used several different machines across the project.

Trahan also spent 33 hours on modeling all the different items, but that wasn’t all. He also spent 560 hours, or about 23 days, printing all of them.
Using a power cost estimator, electricity for the 560 hours comes to about $13.30 at the US average rate of $0.16 per kilowatt-hour. The EIA’s April 2026 figure puts that average slightly higher at $0.19 per kWh, and Californians pay nearly $0.38 per kWh (via Gizmodo).
So, we’re looking at $100 in materials, $1,999 in equipment, 593 hours of your life, and a pair of questionable shorts. All the patterns are still available on MakerWorld if you want to give it a shot, but it might be an expensive affair.
networks
Malaysia ponders regulating management of IP addresses
Wants to revive the lost art of the National Internet Registry, which APNIC has deprecated and isn’t keen to bring back
The government of Malaysia has commenced a consultation on whether it should regulate management of IP addresses and autonomous systems numbers, over objections from regional internet registry the Asian Pacific Network Information Center (APNIC).
Malaysia announced its consultation in June, when the nation’s Communications and Multimedia Commission (MCMC) posted a paper [PDF] in which it explains that a lot has happened since passage of the 1998 Act that governs its activities – so it probably needs an update.
One of the proposed changes would see Malaysia create a statutory authority with the power to manage electronic addressing “including the management of IP addresses, AS numbers and associated fees.”
“This is to support the development of a National Internet Registry model and to ensure a transparent and sustainable administration of electronic addressing resources in Malaysia which will be overseen by the Commission,” the consultation paper states. “This will contribute to a more robust and well-governed digital infrastructure environment in Malaysia.”
APNIC says its talks with the MCMC saw the Malaysian entity express a desire for “full operational and technical autonomy over resource assignments” – powers that existing NIRs don’t have.
National Internet Registries (NIRs) are a relic of the time before regional internet registries came into being. Only APNIC and LACNIC, the Latin American and Caribbean Internet Addresses Registry, allowed NIRs – and only nine exist, covering China, India, Indonesia, Japan, Korea, Taiwan, Vietnam, Mexico, and Brazil.
APNIC stopped accepting applications for new NIRs in 2012, and in 2024 made the moratorium on new applications permanent.
In 2024, APNIC’s executive chair Kenny Huang explained: “NIRs are a historical feature of the APNIC membership structure, recognizing that some IP address registries were already operating at a national or economy level when APNIC started, and some were in formation.”
“In the past, particularly while IPv4 address space was being rapidly allocated and needed careful management, NIRs provided important support to a fast-growing Internet with high demand for number resources and registry services.”
The internet governance community long ago decided that internet resource distribution and management works best when handled by sizable organizations which operate at regional scale, and that if every country had an NIR it would create unhelpful risks and overlapping authorities.
If Malaysia presses ahead with its desire to create its own National Internet Registry (NIR) and have it assume some of APNIC’s functions, it will therefore challenge the status quo. If it actually gets an NIR into operation, that would likely revive debate about whether national governments should have a role in allocating internet resources given the potential for such power to be used for political purposes such as denying resources to groups that a government opposes.
The United Nations last year had its say on that idea by re-affirming its support for multi-stakeholder governance under which governments are one of many voices that participate in debate about the future of the internet.
Kenny Huang has written [PDF] to the Communications and Multimedia Commission (MCMC), pointing out that it’s currently not possible to create a new NIR and that APNIC won’t revisit its policy on the matter – but he also notes that it’s always possible to commence a consultation and policy process that would see APNIC debate a new position. But that process could only start after the conclusion of work on ICP-2, the major revision of the rules that govern the operation of RIRs. The current ICP-2 timeline calls for a revised document to be in place by the end of 2026.
If MCMC decides to pursue creation of an NIR, it will be in conflict on a collision course with APNIC. In the past, most collisions in the world of internet governance occurred at low speed and involved mostly civil debate that plays out over years. ®
After making waves with the launch of the R1T and R1S, Rivian is now in the process of rolling out its mass-market offering, the R2. We’ve already driven it, and we think it’s very good indeed. So good, in fact, that we argued that it looks to be the best EV at its price point so far. It also represents a critical point for the company, since its CEO, RJ Scaringe, has made no secret of the fact that he’s relying on the R2 to push Rivian into becoming a household name.
If everything goes according to plan for the company, plenty more drivers will be finding out about Rivian in the coming years, although most of them won’t be aware that Rivian operated for around a year under a different name. Speaking to the How I Built This podcast, Scaringe recounted how a lawsuit from Hyundai was responsible for Rivian’s current name, as the South Korean carmaker took issue with its original name, Avera.
When Scaringe founded Avera in 2009, he originally intended the company to be a sports car manufacturer. He managed to construct a prototype sports car focused on fuel efficiency, but eventually decided that the company needed to go in a different direction. With funds still very limited at the time, Scaringe had little choice but to change the name after Hyundai sued the company in 2010. The complaint alleged that the Avera name was too similar to the name of Hyundai’s flagship sedan, the Azera.
Scaringe told the podcast that “we said, okay, we’re not gonna fight Hyundai and use the very small number of dollars we have to fight over a name that very, very few people around the world had ever even heard of.” Instead, the founder and his team spent hours brainstorming other names for the company, eventually settling on Rivian.
The name was a portmanteau of the words “river” and “Indian,” because he’d grown up near the Indian River in Florida. The team reportedly chose the name because it was easy to say and didn’t mean anything in any other language. Plus, having a name based on the word river brought connotations of flowing and moving, according to Scaringe.
Settling on Rivian meant that Scaringe was happy and Hyundai was happy. Hyundai continued to sell its Azera sedan in America until 2017. The launch of its upmarket Genesis brand led to the nameplate being axed in the U.S., although Hyundai later launched a new generation of the Azera in other markets.
After its rebrand, Rivian would later disappear from the radar for a number of years as the team worked to develop the technology behind its current lineup. It built its first prototype truck in 2013, but continued to rework the design for several years until it finally settled on the design that would become the R1T. It remains on sale today, with the latest iteration of the R1 offering remarkable levels of power while still being comfortable both on the road and off-road.
Reachy Mini is a limbless desktop robot from Hugging Face made for human interaction experiments, and to give you an idea of what it’s like is a guide on how to implement expressive, local conversational AI complete with head movements and antenna wiggles. It’s conversational in the sense that it aims to feel natural, with low-latency responses and the ability to interrupt, with everything running on local hardware if one so wishes.

The software stack is essentially VAD (voice activity detection) → STT (speech-to-text) → LLM (large language model) → TTS (text-to-speech) which allows users to tweak things to their liking, or independently swap or modify pieces as things evolve.
This also allows users to tailor the services to match whatever their hardware is capable of. For example, one could easily use a frontier AI model via remote API for the LLM while keeping everything else local.
The local models in the example configuration are effective and relatively modest (Qwen3-4B-Instruct for the LLM, and even smaller models for the rest) but it’s nice to have the option to offload parts to remote providers if necessary.
Reachy Mini looked very interesting when it was launched as a kit last year, and since then Hugging Face has built up an impressive software suite and infrastructure through which users can easily share their applications. If you’re curious, there’s a simulator for Reachy Mini which should give you an idea of what it can do.

A fresh new fan film by Secondhand Movie Co takes Star Wars back to its roots and reworks them with absolutely no production value. The original 1977 film had a budget of little under eleven million dollars, which is roughly equivalent to sixty million today. However, this version must make do with microscopic fraction of that, or ten dollars. As a result, the sets, costumes, and the majority of the props are built out of cardboard.
Instead of attempting to hide the limited budget, the people behind Secondhand Movie Co simply went with it and made the most of it. Every aspect on screen, from C-3PO’s cardboard shell with a hasty gold paint job to R2-D2’s endearingly silly googly eyes, has a low-cost heart on its sleeve. The market stalls exhibiting the droids are as rudimentary as a pole with a piece of cardboard put on top, and some of them even wobble when knocked.
Sale
The market stalls segment defines the tone from the start; as Luke and his uncle tour the various stalls and C-3PO complains to R2 that he finally found a new family only for his companion to steal their affection. The cardboard walls behind them feature an Amazon box as improvised wallpaper.

A little while later, the holographic message appears, and R2-D2 projects Princess Leia’s call for aid onto a cardboard cutout illuminated by a flashlight shining through from behind. The famous statement is also somewhat changed: she screams out for Obi-Wan since she has no other options. The projection flickers and cuts out thanks to parental locks and screen-time limits. Aunt Beru mistakes the floating image for a girl trapped inside a rolling dumpster. Each addition stays close to the spirit of the original moment. The interruptions simply acknowledge that even a princess’s call for help must compete with everyday household rules in this version.

Tonal constancy is what makes the whole thing function. The actors are completely serious, and they deliver their modified lines with the same enthusiasm as the original ensemble. So, in the end, the cardboard and terrible makeup appear to be creative solutions rather than budget hacks.
[Source]
Chinese AI systems “have matched the performance of Anthropic’s powerful model Mythos in some cybersecurity scenarios,” reports the Wall Street Journal.
They call it “a development poised to reset the global tech race and pressure the White House in its overhaul of U.S. AI policy.”
Security researchers said that a new AI model, released this month by China’s Zhipu AI, also known as Z.ai, can match the latest U.S. models when it comes to finding security bugs, although it still lags behind Anthropic’s and OpenAI’s products in other tasks. Overall, the capability gap between top U.S. models and those built by Chinese companies has narrowed significantly, and use of Chinese AI systems has surged as businesses seek to rein in runaway costs. A host of companies, including Microsoft, are weighing how they can offer Chinese models on their platforms, a development that is set to alter the balance of power among tech companies…
Unlike models from Anthropic or OpenAI, Zhipu’s GLM-5.2 is open-weight. That means it can be downloaded and run on hardware operated by anybody and can be modified and used without supervision. Open-weight models are ideal for users who want unfettered access to systems they control, but they are also ideal for hackers, who can run them in the shadows. GLM-5.2 has ranked as one of the 10 most-used AI models, according to data from OpenRouter, a company that provides access to more than 400 AI models. In some benchmarking tests, according to the cybersecurity company Semgrep, GLM-5.2 bested Anthropic’s Claude Opus 4.8 model, which was released in May. When given further instructions, Opus 4.8 and GLM-5.2 can match Mythos in bug-finding ability, according to researchers…
“Banning Fable while selling chips China needs to develop its own version is a gift to China,” said Saif Khan, a distinguished technology fellow at the Institute for Progress think tank who worked on export restrictions in the Biden administration. The U.S. needs to maximize the use of Mythos and comparable models to harden its cyber defenses while it can, he added. Among the Mythos 5 and Fable 5 users that had lost access before Friday’s decision to restore Mythos 5 access for some trusted entities: the National Security Agency, which had been testing the tools and found them impressive in trials, according to people familiar with the matter… “It is incentivizing companies across the globe to use cheaper but very capable Chinese open-weight models, while at the same time undermining the U.S. AI industry,” said Niels Provos, a researcher who led security teams at Google and Stripe. “I don’t understand it.”
Thanks to long-time Slashdot reader schwit1 for sharing the article.
We have all had an Instagram feed go off track. A random Reel catches your attention for a moment, and before long, the app starts serving up the same kind of content again and again.
Instagram already has a way to fix some of that through Your Algorithm, a feature that lets users adjust the topics shaping their recommendations. Now, the company wants to make that tool easier to reach while people are actually using the app.
Instagram head Adam Mosseri has previewed new shortcut-like features that could bring Your Algorithm closer to the main feed and Reels. The changes could make it faster to correct repetitive recommendations, add fresh topics, or reshape a feed without digging through settings.
Instagram has been working on improving users’ algorithms and giving them more control over recommendations since last year. It recently expanded those controls to the main feed, allowing users to view and edit the topics Instagram thinks they are interested in.

Mosseri’s preview shows how Instagram may bring those controls closer to the places where recommendations actually appear. One version lets users pull down on the main feed to open Your Algorithm. Another shows the same control appearing after a swipe up from a Reel.
Instagram is also testing buttons under Reels that let users tell the app whether they want to see more videos like the one they are watching. In practice, recommendation tuning could become more immediate, since users would be able to correct the algorithm while browsing instead of opening settings later.
Your Algorithm could also become a more active tool for changing what appears in the feed. Users would not have to rely only on likes, pauses, searches, and watch time to steer recommendations over time.

Mosseri’s preview suggests users may be able to type what they want to see more often, such as positive content, fitness clips, travel videos, or recipe ideas. Instagram would then recommend related topics that users can choose from and add to their feed.
The feature could help users fix recommendations thrown off by accidental clicks on the fly, add new interests, and refresh a feed that has become too repetitive.
The new models will launch with the M5 Pro and M5 Max, according to Mark Gurman.
Apple may be skipping over the M6 generation of its Pro and Max chips, but that doesn’t necessarily mean it’s pushing back the release of its rumored touchscreen laptop. According to Bloomberg‘s Mark Gurman, the new MacBook will launch with the high-end M5 chips that came out earlier this year. The 14-inch and 16-inch models are still expected to be released between the end of 2026 and early 2027, as Gurman has previously reported. The next iteration of the touch MacBook will get the M7 chips not too far down the line.
According to Gurman, who spoke to sources with knowledge of the plans, the M7 versions are already in the advanced testing stage and could arrive by the end of 2027. The touchscreen MacBook will reportedly usher in a slew of changes on top of the touch display. That includes bringing over the Dynamic Island interface from the iPhone, an OLED screen and “an updated industrial design,” Gurman reports.
Apple is expected to introduce its M7 chip in early 2027, followed a few months later by the M7 Pro and Max. Gurman has also reported that we may see the M7 Ultra in 2028.
Two goals and an assist by sheer aura: Cristiano Ronaldo just entered the World Cup chat
Microsoft accidentally kills epic Outlook email threads
Weekend Open Thread: Staud – Corporette.com
The House | Manchesterism won’t survive the painful trade-offs unless it gets citizens on board
Potential 2028er World Cup attendee leaderboard
Asia stock markets slide as tech shares slump
A Look At A Gaggle Of Transputer Boards
Bitcoin (BTC) Dips Below $62K, Ethereum (ETH) Plunges 6% Daily: Market Watch
Dell (DELL) Shares Tumble Over 5% Following Analyst Downgrade to Hold
Securitize Wraps Roubini's SEC-Registered ETF as Dubai VARA Digital Security
Entergy settles forward sale agreements, raises $672 million in cash proceeds
Kraken's xStocks Opens Bending Spoons IPO Registration to EEA Retail
FIH Pro League: India defeat Pakistan 7-1, register biggest win of campaign | Other Sports News
Coinbase, Circle Deepen Crypto Stock Losses Despite Resilient S&P 500
RTX holders must register wallets before token distribution begins
Hyperliquid Named on Singapore MAS Investor Alert Register
Bluekit phishing kit adopts browser-in-the-middle for login theft
India vs Bangladesh LIVE Score, Women’s T20 World Cup: Bangladesh Opt To Bat; India Enter ‘Do-Or-Die’ Stage As Semi-Final Race Heats Up
The DATA Foundation Launches to Tackle AI’s Multi-Billion Dollar Training Data Bottleneck
21Shares Cuts 2026 Crypto Forecasts as Institutional Demand Rises
You must be logged in to post a comment Login