Rivian may have figured out one of the hardest parts of building an affordable EV, as it has managed to reduce costs in producing one of its upcoming EVs. During the latest earnings call, the company said the upcoming R2 has achieved a cost reduction of more than 50% compared to the R1. With the R2 being made as the more accessible mass-market EV, this is a big deal.

How Rivian found ways to save up

According to InsideEVs, Rivian outlined several ways it brought costs down. The company reduced the R2’s wiring harness by 2.3 miles, cut the number of connectors by 60%, and reduced high-voltage cabling by 70% by consolidating multiple power conversion units into one. The company also simplified its new Maximus Drive unit, which has 41% fewer parts compared to the Enduro drive units used in R1 vehicles. Rivian mounted the inverter directly on the drive unit and used smarter cooling and packaging to cut parts and manufacturing complexity.

All of this sounds like boring manufacturing stuff, but it is making a real difference now. Fewer parts usually mean lower cost, easier assembly, fewer failure points, and better scaling.

The upcoming Rivian R2 is using a simpler mechanical setup, which reportedly helped achieve 70% cost savings on the front suspension by moving from a double-wishbone setup to MacPherson struts.

Meanwhile, large die castings also reduced the underbody part count by 90%, while rear door complexity was cut by 65%. CEO RJ Scaringe expects a reduction of more than 50% through design-for-manufacturing work and higher production volumes. He further added that this is how the company expects to ship the T2 profitably, while also keeping it at a more accessible price point without losing performance and utility.

So what about the buyers?

Rivian has positioned the R2 as a more affordable EV, with a target price around $45,000. But the T2 Performance is expected to kick things off at around $58,000 when deliveries begin. The expected price tag reveals that this isn’t an affordable car, though it is still more approachable than the R1S and R1T, which are positioned as premium models.

A $725bn hyperscaler capex ramp and a 20% HBM price rise have made the South Korean chipmaker the second-most valuable company on the KOSPI. The harder question is when supply catches demand.

There is a small joke in semiconductor circles about which part of an AI server is most expensive. The graphics processor used to be the obvious answer. For the past 18 months, increasingly, it has been the memory soldered next to it.

On Monday, the market priced that joke.

SK Hynix, the South Korean memory specialist that supplies the bulk of the world’s high-bandwidth memory for AI accelerators, climbed as much as 12 per cent in Seoul, with shares hitting roughly 1.4 million won, or about $970, in morning trading, according to Reuters.

The rally made SK Hynix the second-most valuable company on the KOSPI, behind only Samsung Electronics, and reflected, the wire said, foreign buying that followed strong earnings and reaffirmed AI infrastructure plans from US hyperscalers the previous week.

It is, by any measure, a remarkable run for a company most consumers have never heard of.

The trigger is straightforward. Big Tech’s combined 2026 capital expenditure is on track to land somewhere between $650bn and $725bn, depending on which analyst’s tally one trusts, an increase of roughly 77 per cent on 2025.

Microsoft has guided to as much as $190bn for the calendar year, with its chief financial officer publicly attributing about $25bn of that to rising memory-chip and component costs.

Meta, in its Q1 update, raised its own range to $125–145bn, citing similar pressures. Amazon’s Andy Jassy has committed roughly $200bn. Google has not been quieter. Practically all of this money flows, in one form or another, towards AI training and inference clusters; a meaningful share of it lands in the bill of materials for high-bandwidth memory, where SK Hynix dominates.

By late 2025, SK Hynix held an estimated 57 per cent of the global HBM market, according to figures cited by analysts at Counterpoint and others. That share is unusually concentrated for a commodity-adjacent business, and it is the structural reason the company’s earnings now look more like those of a software platform than a memory house.

SK Hynix’s first-quarter operating profit, reported on 23 April, was a record. Operating margins on its memory line, by some sell-side estimates, are running above 70 per cent.

Margins of that order do not last forever. They do, however, last as long as supply lags demand.

Why supply is not catching up

HBM is not ordinary DRAM. It is a stacked, 3D-packaged memory built to feed bandwidth-hungry GPUs, and producing it requires a specific set of advanced packaging steps that the industry, including Samsung and Micron, has been slower to scale than buyers would like.

According to TrendForce, both Samsung and SK Hynix have raised HBM3E prices by roughly 20 per cent for 2026, and supply is being booked years in advance by hyperscalers and accelerator vendors.

Samsung’s memory chief publicly warned earlier this year that significant memory shortages were likely to persist through 2027. The chairman of SK Group has gone further, telling investors he expects the wider chip-wafer constraint to last until 2030.

Whether or not those forecasts prove accurate, they explain why long-term supply agreements, in which a hyperscaler effectively reserves output years ahead, are becoming the norm. They also explain why Reuters reports SK Hynix and Samsung increasingly signing such deals with Microsoft and Google directly.

There are, in other words, two kinds of memory in 2026: the kind anyone can buy, and the kind your AI roadmap depends on. SK Hynix is in the second business.

There are, predictably, dissenters. The CAPE ratio on US equities now sits around 38, a level last seen at the height of the dot-com era, and TNW has noted the discomfort that comparison provokes, even as it argues that today’s leading AI-exposed companies are, unlike many in 2000, broadly profitable.

SK Hynix is squarely in that profitable cohort, but it is also unusually leveraged to a single product cycle. If hyperscaler capex moderates, or if competing accelerator architectures reduce HBM intensity per chip, the same operating leverage that has delivered record quarters could work in the opposite direction.

There is also the question of whether the AI build-out will keep producing the kind of returns that justify present capex levels. Meta has been simultaneously announcing record AI investment and cutting roughly 8,000 jobs as it restructures around the spending, a sequence that does not entirely fit the narrative of a healthy organic boom. Investors are, for now, willing to fund both ends of that equation. They have done so before in technology cycles, and they have changed their minds quickly.

The most useful way to read SK Hynix’s 12 per cent move is not as a forecast of where AI ends up, but as a real-time index of how confident the market currently is that AI training clusters will keep being built at this pace. Every dollar of hyperscaler capex announced becomes, eventually, an order book line at a memory supplier. SK Hynix sits in the chokepoint.

If the chokepoint loosens, through Samsung executing on HBM4 at scale, through Micron’s $25bn capacity push, or through architectural shifts that reduce HBM dependency, the dominant share gets shared and the multiples compress. None of those things has happened yet.

Until one does, the market’s read is that the second-largest company on the KOSPI is the one selling shovels to the AI gold rush, and it is not running out of buyers.

The harder, more interesting question is what the chart looks like a year from now. The answer will say less about SK Hynix than about whether the AI build-out has legs, and whether the people writing the cheques in Redmond, Menlo Park, and Seattle are still as certain as they were last week.

A Chinese creator has built a walk-in PC that turns desktop cooling into a human-scale spectacle. The fish-tank-style tower has enough room for a person, a compact desk, and a gaming setup, making the creator look like one of the tiny figures builders sometimes place inside flashy cases.

The build comes from TechTuber Soda Baka, who shared the project on Bilibili. It scales up familiar PC modding cues, including wall-sized fan housings, a huge graphics card prop, chunky cooler parts, and plenty of RGB lighting.

The size gets attention first, but the sealed enclosure soon becomes a heat trap.

The giant parts are mostly props

Soda Baka’s build starts like a serious PC project, with sketching, modeling, and fabrication before the frame comes together. The finished tower looks like an extreme version of the glass-heavy desktop cases that put every component on display.

Most of the human-scale hardware is built for show. The oversized fans, GPU, RAM sticks, and liquid cooler pieces appear to be nonfunctional, while working PC gear gives the creator something to use once he sits down at the desk.

The absurdity is carefully staged. Its size sells the illusion without pretending this is a practical computer anyone should recreate at home.

Why the AC matters

The air conditioner becomes the key component once the sealed enclosure starts heating up. To mimic a PC’s thermal problem at room scale, the creator uses hot-coal sauna gear and water to push the interior above 100 degrees Fahrenheit, or 38 degrees Celsius.

With the huge fans and coolers serving as set dressing, the build relies on a 12kW AC unit with claimed 820 cubic meters per hour of air circulation. The joke only works because the AC does the job the fake cooling hardware can’t.

What to watch next

The build looks more like a promo stunt than a mod worth copying. It may also be tied to an air conditioner sponsorship, though it isn’t fully confirmed.

For actual PC builders, the useful lesson is narrower and more practical. Glass-heavy cases, dense layouts, and sealed spaces all need a real airflow plan before heat becomes the limiting factor. Start with intake, exhaust, and component clearance before chasing the look.

Most tech roles saw salary growth, but not all benefited equally

2025 marked a turning point for the tech industry. After several years of layoffs, cautious hiring, and uncertainty, the sector is showing signs of recovery across the region.

At the same time, companies are rethinking how they prioritise and value talent. Artificial intelligence (AI) is no longer just a talking point—it’s now embedded in hiring priorities, day-to-day workflows, and compensation, as highlighted in Nodeflair’s latest annual tech salary report.

This year’s data, released today (May 4), shows that engineers with AI skills are earning meaningfully more than their peers, with some seeing pay bumps of up to 25%. At the same time, senior engineers continue to pull ahead, as companies place greater value on judgment, architecture, autonomy, and the ability to work effectively with AI tools.

In effect, AI is reshaping the tech career ladder and rewarding stronger problem-solvers at every level. Here is a breakdown of the report and the overall compensation changes for tech roles in Singapore, detailed by role and seniority:

[Disclaimer: Salary data are derived from Nodeflair’s proprietary database of over 130,000 data points, including user submissions that are verified by documents (payslips and offer letters) as well as job advertisements from various job portals. While a majority of the entries are backed by a sizeable amount of data, Nodeflair has flagged out entries with less than 200 data points as potentially lacking accuracy.]

Data scientists are the highest-paid, earning up to S$25K per month

Based on salary data from NodeFlair’s proprietary database, lead data scientists are the highest-paid role. Despite an overall 8.3% decline, they still command a median monthly salary of S$25,000.

Meanwhile, software engineer managers earn up to S$20,100 per month, marking a 10.8% increase as compared to the previous year.

Senior solutions engineers earn a median monthly salary of S$18,500, followed closely by senior product managers at S$18,100. Other top roles include lead mobile engineers at S$16,900, lead data engineers at S$16,000, and platform (DevOps & SRE) leads at S$15,200.

Roles at S$15,000 and below include lead cybersecurity engineers, lead data analysts, systems & IT leads, and senior quality assurance engineers.

On average, senior roles saw the strongest salary growth, with pay rising 12.6%. Lead roles followed at 11.6%, while manager roles increased by 10.8%.

By contrast, mid-level and junior roles recorded more modest gains, with average salary increases of 1.7% and 5.3%, respectively.

AI skills pay off

Beyond providing salary data, NodeFlair’s report highlighted a widening pay gap between engineers with and without AI skills.

Its analysis of 50th percentile software engineers shows that AI-native talent is being paid a clear premium.

The data found that junior software engineers (zero to two years of experience) with AI skills saw a 25% pay bump at the median, earning S$6,000 per month compared to S$4,800 for those without AI skills.

Among mid-level engineers (two to five years of experience), those with AI skills earned 13% more at the 50th percentile—S$8,000 versus S$7,100 for their non-AI peers.

At the senior level (more than five years of experience), engineers with AI skills earned 18% more at the median, taking home S$10,000 compared with S$8,500 for those without.

AI fluency is no longer a nice-to-have—it’s now a salary advantage

Ethan Ang, Nodeflair founder

Advertisement

“With the rise of tools like Claude Code and a broader wave of agentic coding workflows, engineering teams are now rethinking how software gets built around AI, including placing greater value on engineers who are truly AI-native,” he added.

• Read more articles we’ve written on Singapore’s job trends here.

Featured Image Credit: Shadow_of_light/ depositphotos

Trust in online payments has never been as important as it is right now. For anyone who spends money across digital platforms, the tools we use to pay can shape the entire experience. Digital wallets stand out as more than a payment method, they’re a foundation for how fast, safe, and convenient every transaction can feel. Convenience might look like a buzzword, but for online shoppers, it’s the difference between an immediate purchase and a drawn-out checkout that breaks the flow.

Speed and flexibility are at the center of this shift. Today’s gamers, content subscribers, and e-commerce buyers have come to expect instant access, not just to their products, but to funds as well. Digital wallets answer the call, letting users transfer, store, and spend money in ways that traditional cards or slow bank payments can’t match. For those looking to stretch their value further, methods to buy Razer Gold online show how funding a wallet can help unlock exclusive game items, bonus points, or discounted content across many platforms.

Traditional payment routes come with friction. Waiting on transfers, dealing with surprise foreign transaction fees, and entering card information again for each new site can stretch a simple moment into a tedious process. Digital wallets cut this down to a few taps or clicks. They unify purchase histories, hold multiple payment options, and mask your actual card number, drastically reducing fraud risk. These perks are not only attractive, they’re quickly becoming essential for buyers who value both privacy and speed.

When looking for digital games, players often do a quick search only to find that platform stores like the PlayStation Store may have high prices or regional restrictions. Eneba gives those shopping for new titles or DLC a much wider range of game keys, often below standard store prices. Game keys are unique codes that can be redeemed for full games or content, buy a PlayStation code on Eneba, redeem it in your account, and the game appears in your library instantly. The catalog is vast, with instant code delivery, clear info about global versus region-locked content, and a support system in place. Gift cards for Xbox, PSN, and Steam are also available, turning top-ups into a hassle-free option. Crucially, Eneba verifies its merchants and maintains compliance checks so the buying experience is safe and reliable.

Digital wallets don’t just serve gamers, though. They unlock new ways to handle subscriptions, buy digital art, or access streaming services. Their real appeal is in how they make every transaction less of a process and more of a click. This shift in user expectation is subtle but profound, echoing through every part of digital commerce.

Security has become a hot topic. Payment fraud, identity theft, and data breaches drive demand for alternatives to typing out full card details. Digital wallets blend strong encryption with multi-factor authentication. These layers of protection help users manage risk without adding complexity to already busy lives. For those spending on unfamiliar sites, using a wallet adds an extra safeguard, keeping personal details out of harm’s way.

The integration of reward systems is another reason digital wallets have become staples for online buyers. Topping up with game-specific currency or third-party credits can give you cashback, bonus points, or early access offers. This builds loyalty without forcing users to stick to just one shop or ecosystem.

Digital wallets continue to grow by adapting to what users want next: more currencies, more integrations, and fewer barriers between platforms and regions. Their evolution keeps driving innovation, not just for gaming but for all forms of online spending.

Digital marketplaces like Eneba offering deals on all things digital have helped refine what buyers now expect from every online purchase: instant, secure, and tailored to their needs. As competition heats up, it’s clear that digital wallets are here to shape the future of online transactions.

A joint venture with Blackstone, Hellman & Friedman, Goldman Sachs, and General Atlantic will sell Claude into the buyout firms’ portfolio companies. OpenAI’s DeployCo arrived first; this one is bigger.

There is a kind of business school question that has been quietly answered over the past month, without anyone formally asking it. The question is: which is more valuable to a frontier-model company, the next $50bn cheque from a venture investor, or a permanent distribution channel into the operating companies of the world’s largest private-equity firms? Anthropic has been working on the second answer.

On Sunday evening, the Wall Street Journal reported that Anthropic was finalising an approximately $1.5bn joint venture with a small group of Wall Street firms, with an announcement expected as soon as Monday. A

ccording to the WSJ, Anthropic, the buyout firm Blackstone, and Hellman & Friedman are anchoring the deal at roughly $300m apiece. Goldman Sachs joins as a founding investor at about $150m, with General Atlantic and other firms making up the rest. We wrote about the outline of this venture last month, when the structure was still scoped at $1bn or so; the final figure is closer to $1.5bn.

The investors will create a vehicle that operates as something between a consulting arm and a deployment factory: helping the portfolio companies of its private-equity backers integrate Claude across their day-to-day operations.

The pitch is straightforward. Buyout firms own thousands of operating businesses across health care, logistics, manufacturing, and financial services. Each is a potential Anthropic customer. Selling to them one by one, on the standard enterprise software cycle, would take years. Doing it inside a joint venture compresses that timeline into months.

It is, in other words, less a product launch than a sales infrastructure project.

OpenAI got there first, but smaller

The structural template will be familiar. OpenAI announced a similar joint venture, DeployCo, last month, anchored by TPG, Bain Capital, Advent International, Brookfield, and Goanna Capital. The five PE firms together committed about $4bn; OpenAI itself put in $500m, with an option for a further $1bn.

The DeployCo vehicle is expected to be valued at $10bn in a round closing in early May, with OpenAI guaranteeing its PE backers an annualised return of 17.5 per cent over five years.

Anthropic’s structure is different in important ways. The total commitment is smaller in absolute dollars but more concentrated, with Anthropic itself contributing roughly the same amount as its biggest financial partner. There is no public reporting of guaranteed returns.

The investor list is heavier on prestige and lighter on breadth: Blackstone is the largest alternative-asset manager in the world, Hellman & Friedman is among the most disciplined large-cap buyout houses, Goldman is Goldman, and General Atlantic gives the venture a growth-equity stake.

Each side is, in effect, betting on a different proposition. OpenAI’s DeployCo is a numbers play: pull as many PE portfolios as possible into a captive channel, fast. Anthropic’s venture is a credibility play: anchor Claude inside a smaller number of high-profile financial firms whose imprimatur, in turn, sells the model to the rest of the market.

The timing is not accidental. Anthropic has received pre-emptive offers for a roughly $50bn round at a valuation in the $850-900bn range, with the company’s board expected to decide in May and an IPO targeted as early as October 2026.

Anthropic’s annualised revenue run rate has, by its own disclosures, gone from approximately $9bn at the end of 2025 to around $30bn by the end of March 2026. A successful public listing at those numbers requires the company to demonstrate not only model capability but durable enterprise revenue at scale.

A joint venture that pumps Claude into the portfolio companies of three or four major buyout firms creates exactly the kind of revenue ramp public-market investors prefer to model.

It also has narrative value. Claude, in this telling, is not merely a chat product or a developer API but enterprise infrastructure, embedded inside the operating businesses that move significant chunks of the real economy.

There is precedent for the strategy on Anthropic’s books already. Goldman Sachs has spent the past several months piloting Claude internally as the basis for autonomous agents in accounting and compliance, with embedded Anthropic engineers reportedly spending six months inside the bank co-developing the systems.

JPMorgan Chase and Goldman, separately, have been testing Anthropic’s Mythos model under a Project Glasswing initiative focused on AI cyber-risk. The new joint venture is the commercialisation of those experiments.

What it gives Wall Street

For the buyout firms, the calculation is similarly transparent. Private equity returns increasingly depend on operational improvements at portfolio companies rather than financial engineering at the holdco level. AI deployment, in theory, is the next great efficiency lever, and one that the largest funds have struggled to roll out consistently across diverse operating businesses. Owning a stake in the deployment vehicle for one of the two leading model companies is a hedge: it gives the firms first-mover access, preferred pricing, and, plausibly, a financial stake in Anthropic’s broader commercial trajectory.

Goldman Sachs’s $150m position is smaller in dollar terms but particularly telling. It is the same bank rumoured to be co-leading Anthropic’s eventual IPO. A $150m anchor in this venture is less an investment than a relationship deepening.

The risks the structure does not solve

Joint ventures of this kind have a chequered history in financial services. They tend to underperform the most optimistic projections, particularly when the deployed technology is changing as fast as foundation models.

Claude as it exists today will not be Claude in three years; whether the venture’s organisational structure can keep pace with model upgrades, pricing changes, and rival offerings is a real question.

The DeployCo precedent is too young to assess, and Anthropic’s vehicle is, by design, more selective in its partner roster, which limits how quickly it can absorb shocks. OpenAI’s own valuation has come under scrutiny from its investors in recent weeks, a reminder that the model side of these arrangements is not above market discipline either.

There is also a more philosophical risk. Anthropic was founded by researchers concerned about the safety of advanced AI, and has consistently positioned itself as the more cautious of the two leading commercial labs.

A consulting arm that exists primarily to embed Claude inside the operating tissue of dozens of portfolio companies, each with its own data, regulatory, and labour profile, will test that positioning more rigorously than any external benchmark.

None of these is fatal. They are simply the costs of a structure that, until last month, did not exist as a category. Anthropic has decided it would rather pay them with Wall Street co-investors than continue to compete with OpenAI through traditional enterprise sales.

If the announcement lands as expected on Monday, that decision becomes its single largest commercial bet to date, larger, in distribution implications, than any of its model launches. Whether it works will be visible in revenue figures within a year, and in the IPO prospectus shortly after.

We trust GPS like we trust gravity. It just works and gets us where we want to go. But what if someone could trick it into lying to you, and you’d have absolutely no idea?

Unlike jamming, which floods your GPS with noise and at least lets you know something is wrong, spoofing sends fake signals that look completely legitimate. You might be tracking your car or a shipment and think everything is alright when actually the shipment has been routed to someplace else, with you none being the wiser. 

That’s GPS spoofing, and it’s a bigger problem than most people realize. This is what the Department of Energy’s Oak Ridge National Laboratory is trying to solve. 

How bad can it really get?

GPS spoofing might not seem bad on an individual level, but it’s a legitimate concern for companies and governments alike. International criminal networks are already using spoofing to steal loaded long-haul trucks. 

In one case, thieves spoofed GPS to hijack multiple shipments of specialty tequila from a brand co-founded by Guy Fieri and Sammy Hagar. They pulled it off repeatedly because the GPS tracking showed the truck heading exactly where it was supposed to go. 

Now imagine that same trick applied to a truck carrying radioactive materials or pharmaceuticals.

So, what’s the fix?

Researchers at the Department of Energy’s Oak Ridge National Laboratory have built the world’s first highly sensitive, portable GPS spoofing detector that works in real time, even while moving. What makes it special is that it can detect spoofing even when the fake signals are just as strong as the real ones, something no other known system can do.

“Ours is the best in the world,” said Austin Albright, who led the team. “Trucking needs a solution that works without special conditions or dependence on a trusted reference source.”

The detector works independently, without needing a GPS receiver or knowledge of available signals. The team is now working on making it more affordable for widespread use.

Albright sums up the urgency well. Like a carbon monoxide (it’s colorless and odorless) detector catches an invisible danger before it’s too late, this device does the same for GPS, before the cargo, or worse, disappears.