Business, Financial & Legal

Google is developing an upgraded variant of its next-generation Tensor Processing Unit, and MediaTek has secured the exclusive contract to build it. The chip is codenamed Triggerfish and sits within the TPU v9 family as an upgraded follow-on to the base Humufish design. Analyst Ming-Chi Kuo confirmed the arrangement on June 22, with production expected to begin in late 2027 and volume ramping through 2028.

Triggerfish builds on Humufish, though with some AI-targeted changes. SRAM capacity is two to three times higher than that of Humufish, allowing complex AI tasks to run on-chip rather than accessing slower memory. A new simulation die has been added, and the memory has been upgraded from HBM4 to HBM4E for higher bandwidth. This directly addresses the memory wall bottleneck that limits how fast AI accelerators can feed data to their compute units.

A dedicated CPU tile on the same package as the main compute die handles workload switching between training and inference. This makes Triggerfish the first Google TPU designed to handle both in a single chip rather than splitting them across separate silicon.

TSMC's fully booked production capacity is no longer just an industry talking point, as the ripple effects are becoming increasingly difficult to ignore. According to a report from Nikkei Asia, a growing list of major chip customers, including AMD, Google, Nvidia-backed Groq, Tesla, and BYD, are in discussions with Samsung Foundry about manufacturing future chips as demand continues to outstrip TSMC's available capacity.

AMD is reportedly in talks with Samsung about manufacturing certain future CPUs starting in 2028. Google is discussing two separate Samsung engagements. One covers its next-generation Axion processors expected around 2028, while the other involves components tied to its 10th-generation Tensor Processing Unit, codenamed Icefish, which is being co-developed with MediaTek.

Meanwhile, Tesla has already confirmed that its next-generation AI6 chip will be produced at Samsung's Texas facility. BYD, the world's largest electric vehicle maker, is also in discussions with Samsung about manufacturing future autonomous driving chips. Groq, which develops language processing units and is backed by NVIDIA, is already producing chips at Samsung and may also use the foundry for its next-generation products.

Google is reportedly in talks with Samsung Electronics to manufacture part of its next-generation AI processor. According to The Information, citing people familiar with the matter, the chip in question is Google's 10th-generation Tensor Processing Unit, codenamed "Icefish", with mass production potentially beginning as early as 2028.

The arrangement under discussion would split manufacturing between two foundries. TSMC would handle the main compute die on its 1.4nm process, while Samsung would produce the memory I/O die using its 2nm technology. Google is co-developing Icefish with MediaTek, and the chip is still under development.

The Information noted that Samsung's deep understanding of memory characteristics, including HBM, appears to have been a key factor in Google's consideration, and Samsung's ability to handle memory production, foundry work, and advanced packaging under one roof gives it a compelling integrated pitch.

Elon Musk has officially become the world's first trillionaire after SpaceX stock began trading on the Nasdaq at $150 per share, valuing the company at over $2 trillion.

Forbes reported that, based on SpaceX's stock price, Elon Musk's net worth has now exceeded $1.1 trillion as of Friday morning. Musk made a substantial jump on Thursday evening when the SpaceX IPO went live at $135 per share, which caused Musk's net worth to rise from $188 billion to $982 billion. Yes, $794 billion in a single evening.

Musk's wealth is divided between shares and stock options, with Musk owning 4.8 billion shares of SpaceX, worth $715 billion, and an additional 350 million stock options, worth $50 billion. In total, Musk's ownership of SpaceX sits at 38%.

Microsoft is reportedly cutting hundreds of jobs within its Azure cloud business in China, marking at least the third round of downsizing the company has carried out in the country over the past two years. According to the South China Morning Post, between 200 and 400 employees are expected to be affected. Those impacted are set to leave Azure on July 6 and receive severance packages based on tenure plus up to seven months of salary.

Five affected employees cited by the report said Azure staff in Beijing and Shanghai received emails last week informing them their roles would be terminated. Some employees have also been offered the option to relocate to Canada. A Microsoft representative told the South China Morning Post that the company had shared an optional internal transfer opportunity with eligible employees, adding that Microsoft remains focused on serving customers and growing globally.

The layoffs come as both Washington and Beijing tighten oversight of cross-border data flows. The US Department of Justice implemented its Data Security Program last year, restricting American organizations from transferring certain datasets to employees, vendors, and investors in countries of concern, including China. Meanwhile, Beijing has continued to strengthen its data governance framework following the introduction of the Data Security Law and the Personal Information Protection Law in 2021.

Valve has been accused by a Dutch consumer group of using Steam's market power to maintain high PC game prices, higher than the consumer watchdog believes they should be.

The claim is being run by the Stichting Consumenten Competition Claims under the GameClaim banner, alleging that Dutch PC gamers have unfairly paid too much for PC games, downloadable content, and microtransactions due to Valve's restrictions on competition through Steam. The claim states that the group isn't opposed to Steam as a platform, but is seeking fair PC game prices and compensation for Dutch gamers.

The lawsuit claims that Steam is so dominant in the market that developers simply cannot avoid it, and since Valve knows this, it uses its position to maintain high prices and prevent cheaper pricing elsewhere. Additionally, the claim alleges that Valve forces developers to sell their titles at a minimum price and prevents them from offering lower prices or better terms through other online PC game stores. Essentially, the lawsuit claims Valve blocks developers from selling their games outside of Steam for less than they sell them on Steam.

Despite a rough couple of years for the company, Intel could be set to make a turnaround after its foundry just received what could be its biggest validation yet. According to a report from The Information, Google has placed an order with Intel to manufacture more than three million Tensor Processing Units in 2028. Intel's stock jumped more than 9% on the news, adding to a gain that has already seen the stock climb nearly 169% this year.

Google's TPUs are its in-house AI chips, designed to reduce reliance on NVIDIA's GPUs while powering its own AI services and cloud offerings. Google is reportedly expected to produce more than six million TPUs between 2027 and 2028, with Intel handling roughly half that volume. Previously, TPUs were co-designed by Google and Broadcom with TSMC handling production.

TSMC holds around 73% of the global foundry market and remains the dominant producer of advanced chips for NVIDIA, Apple, and Google. But the AI boom has pushed demand well beyond what TSMC can comfortably handle. TSMC's own chairman recently acknowledged that customer demand is reaching the limits of what the company can manage, and the company is reportedly fully booked out until at least 2028. That capacity crunch has created an opening for Intel that would have been difficult to imagine a few years ago.

If you have been holding off on a PC upgrade, hoping DDR5 prices would come down soon, AMD has some bad news. David McAfee, AMD's VP and GM of Client Channel Business, told 4Gamers during Computex 2026 that DDR5 memory prices are not expected to return to normal levels for around two years, putting any meaningful recovery around 2028.

In the interview, David pointed to the AI boom that has shifted supply away from consumer DDR5 and toward HBM for AI data centers. DDR4 production capacity has also been declining over the past year or two as manufacturers shifted investment heavily toward DDR5. This has left users on older platforms squeezed from both directions.

To give you an idea of the impact, a 32GB Corsair Vengeance RGB DDR5-6000 kit that cost around $100 in late 2025 is now selling for around $440. McAfee expects prices to recover gradually as new capacity comes online. Samsung, Micron, and China's CXMT are expanding their DDR5 production capacity. The problem is that building new memory manufacturing facilities takes years, meaning the relief is real but slow.

At Computex 2026 in Taipei, SK Group Chairman Chey Tae-won confirmed that the company plans to double its total wafer production capacity within five years, driven by AI demand that he expects to keep memory supply tight through 2030.

"We are going to double the whole capacity over the next five years. There are a lot of obstacles and hurdles, but we will get over them and expand," Chey told reporters. He added that the company would provide whatever funding was required to support the expansion, saying simply, "Whatever we need, we'll provide it."

According to Bloomberg, SK Hynix currently produces around 550,000 DRAM wafers per month, including approximately 200,000 from its Wuxi fab in China. The target is to reach roughly 1 million wafers per month by 2030 to 2031.

For roughly six decades, the semiconductor industry has followed a simple and reliable formula: make transistors smaller, pack more of them onto a chip, and watch performance climb. That formula, commonly known as Moore's Law, is now running into hard physical limits. A research team at the University of Illinois Urbana-Champaign thinks the next gains will come not from going smaller, but from going vertical.

Led by materials science and engineering professor Qing Cao, the team has developed a method for stacking multiple active layers of silicon circuits directly on top of one another on a single chip, achieving device yields between 98% and 100%. The results were published in Nature.

Building high-performance silicon circuits typically requires temperatures approaching 1,000 degrees Celsius. Once the team installs the first layer of circuitry and metal wiring, they must keep any subsequent layers below 400 degrees to avoid damaging what is already in place. Previous attempts to work around this used alternative materials for the upper layers, but those devices consistently underperformed compared with standard silicon transistors.

The AI boom has driven extraordinary demand for computing power, but it is now creating a constraint that raw performance alone cannot solve. Speaking at a conference in Amsterdam, TSMC Senior Vice President of Business Development Kevin Zhang said energy efficiency has overtaken computing performance as the defining priority for customers across the chip industry.

"The area customers most want improvement in is energy efficiency. This is true across the board, whether you are the edge guy, smartphone, mobile, IoT application, or high-performance AI data center," Zhang told Reuters. The shift marks a major change for the chip industry as the era of simply packing more transistors onto a chip and calling it progress appears to be drawing to a close.

TSMC expects its A14 chips, due around 2028, to deliver more than 20% higher computing performance while cutting power consumption by up to 30% compared to its current N2 technology. Zhang added that while transistor density remains central to TSMC's plans, technologies such as advanced packaging, chip stacking, and photonics are becoming increasingly important in driving efficiency gains beyond what transistor scaling alone can achieve.

Samsung has begun shipping samples of its HBM4E high-bandwidth memory to major global customers, making it the first company to deliver the next-generation AI memory product. The announcement sent Samsung shares surging as much as 6.51% before settling at a 3.67% gain, closing at 310,500 won.

The new 12-layer HBM4E delivers a stable pin speed of 14 Gbps, with performance that scales up to 16 Gbps, representing more than a 20% increase over HBM4. Memory bandwidth reaches up to 3.6 TB/s per stack, which, for context, is roughly equivalent to the combined memory bandwidth of two GeForce RTX 5090 GPUs in a single stack. The 12-layer configuration ships with a 48GB capacity, a more than 30% increase over the previous generation, with 32GB eight-layer and 64GB sixteen-layer variants also in the works, depending on customer requirements.

Beyond raw speed and capacity, Samsung has also made meaningful efficiency gains. Advanced low-power design techniques and an optimized packaging architecture improve energy efficiency by 16% and reduce thermal resistance by more than 14% compared to HBM4, translating into better heat dissipation and longer-term reliability in demanding data center environments.

TSMC employees are reportedly pushing back over rumors that the company may cut employee bonuses, even as the chipmaker continues to post record profits driven by the global AI boom. The company's net profit jumped 58% year over year in the first quarter of 2026, making the timing of any potential bonus reduction particularly tone-deaf from an employee perspective.

The frustration has spilled onto social media, with dedicated Facebook communities for TSMC staff reportedly flooded with angry complaints. Employees are venting about working high-stress, exhausting shifts while the company prioritizes investor returns and capital expansion over workforce compensation. Some are now openly discussing tougher action, with ideas of deploying Samsung-style strike tactics gaining rapid traction across the company's ranks in Taiwan.

The Samsung comparison is not accidental. This week, Samsung Electronics narrowly avoided a catastrophic factory shutdown by signing a last-minute deal with its union, creating a record $26.6 billion performance-based bonus pool. The deal prevented a strike that could have cost Samsung upwards of $66 billion and disrupted global memory chip supply chains. TSMC employees appear to be watching that outcome closely and drawing their own conclusions about what organized pressure can achieve.

Huawei has produced a 122TB SSD using its proprietary chip-packaging technology known as Die-on-Board (DoB), without access to the latest 100-plus-layer 3D NAND from mainstream suppliers. According to BLOCKS&FILES, US restrictions have forced Huawei to rely more heavily on domestic suppliers for its NAND needs.

Once Huawei's existing stockpile of US 3D NAND chips was exhausted, it turned to Chinese-made NAND from suppliers such as YMTC. Rather than accepting the capacity disadvantage of existing TSOP or BGA packaging, Huawei focused on board-level packaging innovation to improve capacity density by enabling tighter NAND integration.

DoB is a wafer-level packaging technology that mounts semiconductor dies directly onto a base PCB. Mainstream SSD suppliers like Samsung with V-NAND, Kioxia and Sandisk with BiCS, and Micron typically use multi-die stacking inside TSOP, BGA, or other packages before mounting them onto a base PCB. DoB enables more flexible die stacking and shorter interconnects, improving both performance and power efficiency compared to traditional packaging approaches, with a 33% improvement in capacity density, though the number of stacked layers remains undisclosed.

Samsung Chairman Lee Jae-yong has reportedly slipped into Taiwan for a meeting with MediaTek, in an effort to steal the chip designer as a foundry client from TSMC.

The move underscores Samsung's aggressive push to expand its foundry business and challenge TSMC's dominance in the sector. According to a report from WCCF Tech, Lee made the trip to Taiwan, on May 21 with a high-level entourage with the goal of meeting with MediaTek's CEO Cai Lixing.

According to reports, Lee was offering MediaTek incentives, including lucrative memory deals, or priority access to its memory. This is reportedly a common tactic by Samsung to attract new foundry partners, as it was the strategy Samsung used to attract Qualcomm as a foundry client. Furthermore, the conversation between Samsung and MediaTek comes at a time when the relationship between TSMC and MediaTek is fairly strained.

SpaceX has revealed its financials for the first time, showing a $4.9 billion loss in 2025 despite $18.7 billion in revenue. However, that isn't the biggest news here, as the Elon Musk-led company is preparing for an IPO, and its filing shows it could raise as much as $80 billion, and value the company somewhere around $1.75 trillion.

The filing, required ahead of its IPO, marks a major shift for SpaceX, which has long operated under a veil of secrecy, at least when it comes to its financials. Revenue for 2025 rose 33 percent year-over-year, but the company swung from a $791 million profit in 2024 to a staggering $4.9 billion loss.

First-quarter 2026 losses matched the full-year 2025 loss, raising questions about long-term financial sustainability. Capital spending nearly doubled to $20.7 billion, with a large chunk going toward AI development and infrastructure.

Elon Musk has lost his lawsuit against OpenAI and its CEO Sam Altman after a jury unanimously ruled he waited too long to file his claim, meaning the statute of limitations prevented a ruling in Musk's favor.

The verdict dismisses Musk's allegations that OpenAI's pivot to for-profit violated its founding mission and committed unjust enrichment. Musk alleged that OpenAI strayed from its original mission of being a non-profit company to a profit-driven model that prioritized commercial success. The trial centered around Musk's 2024 lawsuit, which made the aforementioned accusations.

However, a nine-person jury found that the statute of limitations had expired, with Musk having left the company in 2018. District Judge Yvonne Gonzalez Rogers agreed, dismissing all claims as "untimely."

Global memory semiconductor giants are undergoing a profound structural shift. According to a report by the Chosun Ilbo citing research from market firm Omdia, Samsung Electronics, Kioxia, and Micron Technology are either shutting down or slashing production on their legacy 2D NAND flash production lines.

Samsung has slightly reduced its NAND Flash wafer production forecast, expecting a decline from 4.9 million wafers in 2025 to 4.68 million in 2026. Similarly, SK Hynix is forecasted to reduce output from 1.9 million wafers in 2025 to 1.7 million in 2026. Meanwhile, 64Gb MLC NAND spot prices have surged more than 300% from end-2025 levels and are currently trading in the $20-$28 range. MLC NAND, which stores 2 bits per cell, offers better data retention and durability than 3-bit TLC and 4-bit QLC designs, yet weak profitability has pushed it to the margins.

The main driver behind the production cuts is profitability. Thanks to the AI boom, high-bandwidth memory and advanced 3D NAND exceeding 300 layers have become the industry's focal points, while low-margin legacy 2D NAND is being rapidly sidelined and deprioritized for capital investment.

World-famous singer Dua Lipa has recently filed a $15 million (£11 million) lawsuit against Samsung for using her likeness to sell TVs without her permission or consent. The complaint was filed on May 8th, 2026, in the US District Court for the Central District of California. The complaint accuses Samsung of copyright infringement, trademark infringement, and misappropriation of Lipa's image and likeness.

According to the lawsuit, Samsung is using an image of Dua Lipa taken at the 2024 Austin City Limits Festival on its TV product boxes. The complaint claims that Ms. Lipa owns the rights to the image, and Samsung's packaging was "designed to improperly capitalize on Ms. Lipa's hard-earned success" for the promotion of their products. Samsung was thus sent multiple "cease and desist" letters from Ms. Lipa's legal team, all of which were ignored according to the lawsuit.

The suit alleges that Samsung was profiting financially by using Dua Lipa's likeness and quotes several social media comments from fans of Ms. Lipa who were apparently convinced to buy a Samsung TV just because of her appearance on the box. According to Ms. Lipa's legal team, Samsung has violated California's right of publicity statute, as well as trademark and copyright infringement claims. The singer is demanding $15 million in damages, as well as punitive damages and legal costs.

eBay's board of directors formally opposes GameStop's $56 billion share buyout proposal and shares major points of concern with the deal.

A bit ago, GameStop made an offer to acquire eBay for $125 per share in a half-cash, half-stock deal worth around $56 billion. On the same day the proposal was announced, GameStop CEO Ryan Cohen went live with CNBC in a quizzical interview and was unable to answer exactly where the remaining buyout funds would come from--GameStop may be $15 billion short.

Fast-forwarding to today, and eBay has now rejected GameStop's offer. Paul Pressler, the chairman of eBay's board of directors, tells the GameStop CEO: "We have concluded that your proposal is neither credible nor attractive."