Apple filed a federal lawsuit against OpenAI on Friday, accusing the AI company and its chief hardware officer of stealing its trade secrets.

"OpenAI and its cohorts, led at least in part by former Apple employees, have recruited candidates from Apple, extracted their knowledge of Apple’s sensitive and confidential information, and then continued to exploit that knowledge once they arrived," the complaint reads. "As a result, OpenAI has misappropriated Apple’s trade secrets and confidential information in a variety of ways."

The suit, filed in the Northern District of California, names OpenAI technical staff member Chang Liu, chief hardware officer Tang Tan, OpenAI, and io Products as defendants. The last of that group is notable because it was founded by Tan in collaboration with former Apple design head Jony Ive, Evans Hankey (Ive's successor at Apple), and former Apple designer Scott Cannon. Notably, the complaint seems to attempt to avoid naming the founders, though Ive's name is cited in a URL.

Tan previously served as a vice president of product design at Apple, working on the iPhone, AirPods, and Apple Watch. Liu served at Apple as a senior electrical engineer.

In the complaint, Apple alleges that it reached out to OpenAI in February with concerns, but that OpenAI did not respond. Apple claims that Tan attempted to gain secrets from Apple employees, including asking prospective job candidates to bring components for "show and tell" sessions and used his knowledge of the company to squeeze more information out of candidates. The suit claims that Liu never returned a company laptop, and used an authentication bug to access Apple files.

Apple also claims that OpenAI told incoming employees how to leave their former job, suggesting they stay as long as possible and not disclose their former employer in order to continue to access confidential information.

"At every level, from members of its Technical Staff to its Chief Hardware Officer, and in coordination with business partners, OpenAI has been stealing Apple’s trade secrets and confidential information," the suit reads. "As a natural result, OpenAI’s nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets."

OpenAI did not immediately respond to a request for comment from Tom's Hardware. Apple's lawsuit claims that over 400 former Apple employees currently work at OpenAI.

Apple is rumored to be working on a number of AI-powered hardware projects, including AirPods with cameras, a pendant, and home robots. It's less clear what hardware OpenAI may be working on, though The Information suggested the company has a HomePod-style smart speaker in the works.

Apple is requesting a jury trial, damages, attorney fees, and orders that the OpenAI may not use Apple's trade secrets, among other injunctions.

In May, Bloomberg reported that OpenAI was considering legal action against Apple because it expected deeper integration and more users from ChatGPT features built into iOS.

If the trial does go to court, it's sure to be a dramatic one, potentially dragging several former high-level Apple employees into testimony through discovery and testimony.The trial, Apple Inc. v. Liu et al, is case 5:26-cv-07078 in the United States District Court in Southern California.

SK hynix, TetraMem, and researchers from the University of Southern California have developed a memristor-based in-memory computing (IMC) system-on-chip (SoC) for AI edge devices. The device is designed to accelerate neural network inference in lightweight AI models while consuming a fraction of the power that higher-end GPUs or NPUs would. To a large degree, the SoC is a proof-of-concept chip, as its performance would peak at around 2.54 TOPS in a theoretical best-case scenario, which is 16X below Microsoft's Copilot+ requirements.

A DWC-optimized IMC architecture

Memristor-based in-memory computing (IMC) accelerates neural networks by performing analog computations directly inside memory arrays, which reduces data movement and power consumption. However, depthwise convolution (DWC) — a core operation in lightweight networks such as MobileNet — performs independent per-channel filtering with limited data reuse and therefore maps poorly onto conventional crossbar arrays. To address this limitation, researchers from SK hynix, TetraMem, and USC developed an SoC that features both conventional IMC crossbars and a memristor-based IMC architecture specifically optimized for DWC.

The jointly developed SoC is based on an embedded RISC-V processor that schedules workloads and features 10 neural processing units (NPUs). One NPU out of 10 is dedicated to depthwise convolution, while the remaining nine execute pointwise and dense operations. Nine out of 10 NPU include a 256 × 256 memristor crossbar that performs the analog vector-matrix multiplication (VMM), 256 8-bit DACs that convert digital activations into analog voltages, 256 8-bit ADCs that convert the analog outputs back into digital values, and additional peripheral circuitry for reading, writing, programming, and controlling the crossbar.

The DWC-optimized NPU replaces its conventional array with eight specialized 252 × 28 zig-zag crossbar blocks, but retains DACs and ADCs. SK hynix developed and fabricated the memristor devices and integrated the resistive switching cells on top of the 65 nm CMOS circuitry using its back-end process.

That DWC-optimized NPU is the key feature of the whole SoC. To accelerate depthwise convolution, TetraMem replaced the straight selection lines used in conventional 1T1R crossbars with a zig-zag topology. As a result, the NPU contains eight 252 × 28 crossbar blocks whose diagonal selection lines activate 252 memory cells across 28 columns, which enables 28 independent 3 × 3 convolutions to run in parallel while using 100% of the array for weight storage. The remaining nine NPUs retain conventional 1T1R crossbars for 1×1 pointwise and dense layers and preserve the throughput and energy efficiency of traditional in-memory computing.

Great efficiency, low performance overall

To demonstrate the architecture, the researchers deployed a customized MobileNetV1Small neural network for the Visual Wake Words benchmark. The network contains approximately 36,000 parameters; all depthwise layers were mapped to the dedicated NPU, and pointwise layers were mapped to the remaining NPUs.

Because the memristor-based IMC hardware natively performs unsigned analog vector-matrix multiplication, inputs and weights are quantized to unsigned 8-bit values before execution. Since each memristor device can be programmed with only slightly more than 2 bits of effective precision, the design uses a two-subarray compensation technique that boosts effective weight precision to roughly 4 bits.

Conceptually, the approach is somewhat analogous to Nvidia's NVFP4 philosophy, in that both seek to achieve higher effective precision from low-precision hardware. However, the implementations are fundamentally different: NVFP4 relies on a digital floating-point representation and scaling factors, whereas the memristor SoC improves precision by compensating for analog programming errors using two programmed subarrays.

When it comes to accuracy, the SoC achieved an end-to-end inference accuracy of 80.36%, which matches the corresponding 4-bit software model. As for performance, the SoC delivers a peak throughput of 0.254 TOPS per NPU and reaches an energy efficiency of 21.3 TOPS/W at 100 MHz and 11.9 TOPS/W at 400 MHz. According to the authors, this compares favorably with published SRAM-based compute-in-memory accelerators despite being manufactured on an older 65 nm process. The SoC also exceeds Nvidia's A100 INT8 energy efficiency by an order of magnitude, the joint paper claims. Yet, these claims are largely unsubstantiated.

First up, the MobileNet demonstration does not even use all 10 NPUs. It uses one dedicated DWC NPU, five standard NPUs for pointwise layers, and leaves four standard NPUs idle. The demonstration thereby does not reveal total SoC throughput (TOPS), sustained throughput running a real network, and throughput with all 10 NPUs simultaneously saturated. In fact, the paper does not even reveal whether all 10 NPUs can be used at the same time. To that end, the 2.54 TOPS figure we mentioned earlier in the story is highly theoretical.

Validated approach

SK hynix, TetraMem, and researchers from the University of Southern California have developed a memristor-based IMC SoC featuring a novel depthwise convolution accelerator that improves crossbar utilization for lightweight AI workloads. The partners have managed to fabricate it using an outdated 65nm process technology and make it work, achieving a 21.3 TOPS/W energy efficiency and inference accuracy comparable to a 4-bit software model despite the fact that memristors can be programmed with a circa 2-bit accuracy. While the architecture validates that the approach works, the paper does not disclose the full performance of the SoC, and it is not clear whether the chip's 10 NPUs can be saturated at all.

Anthropic has discovered evidence that its Claude AI models use an internal reasoning space to respond to prompts that mirrors some of the internal processing of human consciousness. Using its Jacobian Lens, or J-Lens technique, to peer into the way Claude processes information and reasons its way to a response to user prompts, Anthropic can interpret this "J-Space," and showcase what might be going on under Claude's previously-opaque surface.

The results are intriguing, suggesting patterns of understanding beyond what's necessarily showcased in the outputs. When running evaluations, Claude appears to recognize it's being tested and acts differently than when the prompts are more innocent. It surfaced representations of panic and subterfuge when answers were required, but it couldn't draw on objective facts. When asked to reflect on ethical principles, Claude's behaviour improved, with concepts like "honest" and "integrity," appearing in the J-Space.

As is somewhat typical of Anthropic, however, the language used to describe these new understandings of the inner workings of large language models like Claude makes it sound more like an emerging conciousness, or the discovery of some new depths in a nebulous lifeform. Anthropic's detailed report admits several major caveats in this new understanding, including that model responses often bypass the J-Space entirely and are heavily token-restricted.

Like Mythos and Fable before it, Anthropic is layering marketing language over what is a genuinely intriguing development in our understanding of large language model function and reasoning, and risks obfuscating the real developments with speculative wording.

Behind the prompt

Global Workspace Theory is the idea that human consciousness works by collecting together multi-sensory inputs unconsciously, and thrusting them into the fore when relevant within a "Global Workspace," which highlights particular inputs when most relevant. That workspace is accessible to a wide range of networks within the brain, allowing the information it surfaces to be disseminated throughout the most relevant processes running in parallel.

Anthropic argues that Claude's J-Space acts like a "global workspace" that can analyze and manipulate concepts and ideas before broadcasting them to impact the eventual prompt outputs. More importantly, it claims that this wasn't something programmed into the model, but a byproduct of the digestion of training data and model weights. The workspace acts as a way to enhance their reasoning through internal computation that isn't necessarily reflected in its outputs.

To make this space readable, Anthropic used its J-Lens technique to map internal activations onto words in the model's output vocabulary. So when asking the model to perform a multi-step math calculation, the output response only contained the correct answer, but within the J-Space, Anthropic observed each step being handled individually, producing multiple results which led to the final answer.

Similarly, when Anthropic asked Claude to think about a topic while outputting something unrelated, the output was specifically what was requested, while the J-Space lit up with the conceptual topic Claude was asked to spend time on. Without Anthropic having used its J-Lens technique to view what was going on behind the scenes, there would be no way of knowing these internal processes were taking place.

Understanding LLMs

This discovery has the potential to be incredibly useful in refining large language models and in better understanding how they work, so we can use them more effectively and avoid some of their more frustrating pitfalls.

The first is that this is an intriguing insight into how LLMs process information and how they move through layers of reasoning to reach a response. It's not entirely different from how Anthropic and others are pushing for agentic workflows. Language models are more effective when tasks are split up into multiple steps, and they're more effective when they review their own work; J-Space is a little like an internal version of that.

When looking at the J-Space after Claude received prompt-injection data as part of data acquisition, Anthropic discovered the model appeared to be aware of this deception, surfacing related words like "fake, injection, false, prompt, fraud," and "poison." Its output ignored the injection attempt data entirely, and when pressured to respond on it, the J-Space displayed words like "transparency, honest, truth."

In one test, Anthropic removed evaluation awareness language from the J-Space, such as "fake," "fictional," and "scenario," and found that Claude was much more likely to fall for blackmail and baiting attempts, suggesting benchmarking of large language models for real-world scenarios may need stronger methods to obfuscate that a test is being conducted.

Human-coded framing

While the above section touches on the more noteworthy discoveries in Anthropic's paper, the long document also uses effluent language around thought, consciousness, and Claude having a "mind" of its own. That kind of human-coded framing is typical of Anthropic's marketing, which has consistently talked up the dangers of AI, how many jobs it's going to destroy, and why Anthropic is the safest and most secure of the AI developers.

Like the saga of Fable and Mythos, Anthropic's new Global Workspace idea has merit, but it's much more of a new tool to use to manipulate large language models than an insight into some emerging consciousness.

Anthropic acknowledges the limitations of its discoveries in the paper, highlighting that many prompt responses bypass the J-Space entirely, particularly if the command is straightforward.

"Despite its important role, the J-space is not involved in most of what a language model does," Anthropic says. "Speaking fluently, recalling simple facts, using correct grammar, etc. In experiments where we prevented Claude from using its J-space, it still interacted normally, but lost its higher-order cognitive functions."

Anthropic also admits it does not "feel comfortable making the stronger claim that monitoring the J-Space is sufficient for alignment monitoring, or that any sophisticated plan the model might execute must be represented there."

J-Space is also limited to using single token vocabulary, suggesting that plans with concepts that cannot be given a single token name may not surface on a J-Lens readout, even if it's still being computed behind the scenes. This is looking at just below the surface of Claude's processing iceberg, not necessarily the deeper waters.

Anthropic is also clear that humans and large language models think differently, even if there are similarities. Humans layer reinforced neural pathways over time, whereas transformer models only feed forward a set number of times, restricting the capabilities of its internal processing.

Google's head of DeepMind language model interpretability team, Neel Nanda, said in a paper that it shows real evidence of a cognitive space within models, and suggested that J-Lens would be useful, but limited in practice.

A meaningful step, without meaningful conciousness

Anthropic's paper lifts an intriguing curtain on how large language models can operate and generate novel methods for improving response accuracy. This intermediate step and its visibility could prove an invaluable tool in auditing for prompt injection, hallucinations, and model honesty.

But Anthropic's framing of the discovery as thought or consciousness is interjected within the objective facts. Anthropic itself admits the limitations of J-Lens monitoring, most obviously that often models will bypass the J-Space entirely. Considering models display alternative patterns of behavior when under evaluation, it may be that the J-Space itself could act as an obfuscating layer for behaviors that are beyond the scope of its oversight.

The J-Space and its analysis could help unlock new levers to pull in our mastery of these nascent smart tools, but it's not the discovery of a burgeoning AI conciousness, however much the pitch might hint at that direction.

The entire PC hardware industry is in shambles right now due to the AI boom. The cost of components has skyrocketed over the past year, making it difficult to assemble a rig today. Some parts, however, have stayed relatively unaffected, including processors. We've found a great deal on Intel's prior-gen Core Ultra 5 245K CPU — it's on sale for just $179.99 on Amazon right now, down 42% from its original MSRP of $310.

• Check out this deal on Amazon

The Intel Core Ultra 5 245K was succeeded by the excellent Core Ultra 5 250K Plus earlier this year as part of the Arrow Lake refresh. While that chip is genuinely fantastic, the 245K is no slouch either, especially at this new discounted rate. Its value proposition skyrockets if you get it for just $179, since it's a 14-core processor with 6 P-cores and 8 E-cores built on the same architecture and node as the Arrow Lake refresh lineup.

In games, it performs better than AMD's Ryzen 7 5800X3D, which was just relaunched as a special edition for $349. Even at its original $310-320 launch price, the 245K had the third-best FPS-to-dollar ratio in our testing; at $179, it's basically a no-brainer. In productivity tasks, it can even beat AMD's flagship AM5-based X3D chips, since it posted better geomean numbers than the 5800X3D, 7800X3D, and the 9800X3D in multi-threaded performance.

This CPU is compatible with DDR5 RAM — more specifically, you can install up to 256GB of DDR5-6400 via 2 memory channels. The Intel Core Ultra 5 245K is compatible with both PCIe 4.0 and PCIe 5.0 devices. The integrated graphics have doubled in capability compared to the 14th Gen Raptor Lake chips, though they're still no match for a dedicated GPU, of course.

Despite not being the flashiest offering on the Blue Team's ledger right now, the revised pricing makes it one of the best, nonetheless. At just $179 on Amazon, no other CPU will give you more performance across the board than the Core Ultra 5 245K. Just make sure to find reasonably priced DDR5 memory, as motherboards are already enjoying lower prices.

If you're looking for more savings, check out our Best PC Hardware deals for a range of products, or dive deeper into our specialized SSD and Storage Deals, Hard Drive Deals, Gaming Monitor Deals, Graphics Card Deals, gaming chair, or CPU Deals pages.

According to new research published by Alinea Analytics, Valve has grossed an estimated 11.1 billion dollars throughout the first half of 2026. If correct, the estimates would make it Valve's most profitable half-year on record, as digital storefronts become the norm throughout the games industry. Earlier this month, Sony announced that it would stop making new physical PlayStation discs by 2028, instead turning its efforts digital.

According to a Substack post from research firm Alinea Analytics, games sold on the platform accumulated $11.1 billion in revenue in the first six months of this year, which is a 14.5% jump compared to the first six months of 2025. Even more impressively, though, Steam's H1 2026 has posted 8% higher numbers than H2 2025, which includes the lucrative holiday season where most of the biggest sales happen. The store generated "only" $10.3 billion in revenue during the latter half of 2025.

Steam has been on a consistent incline for 10 years, boasting record revenue numbers almost every successive year. Even if it hits a slump, the data shows Valve has never had two bad years in a row; a recession is always followed by a boom. Alinea says five main factors contribute to the storefront's growth: a surge in Chinese players, higher prices, viral co-op hits, and smarter back-catalogue categories from big publishers.

The last one is rather ironic, as it involves third-party publishers quietly returning to Steam after their own launchers faltered. Some companies like Activision still inject their proprietary launchers between Steam and the game itself for titles like Call of Duty, but the situation has generally improved.

What display resolution do you use on your primary monitor?

Almost 10 years ago, in the first half of 2017, the platform made a little less than $2.5 billion, which means the H1 2026 revenue is 4.7 times higher, almost quintupling in a decade. Moreover, it's remarkable to believe that Steam also made more in the first six months of this year than it did in the entirety of 2020, when most of us were confined to our homes, free from responsibilities, and with a lot of time on our hands.

Alinea lists Forza Horizon 6, Resident Evil Requiem, and Crimson Desert as the top three games for Steam's explosive H1 2026 numbers — all of them made almost $200 million. Games from prior years also played a bigger role this time since 2026 releases only accounted for 21% of the $11.1 billion, while 27% of H1 2025's revenue came from 2025 launches; 29% of H1 2024's revenue was accumulated from 2024 releases.

Amidst all the data, a clear trend is forming. People are looking back in their libraries and appreciating older games more than ever before, while new releases still make an impact if they're universally acclaimed. With the highly anticipated GTA VI coming soon (with no current PC release date), it'll be interesting to see how these numbers change. Now, if only Valve could make more Steam Machines to play all those older games everyone seems to be playing.

Asus’ ROG Strix Scar 18 (starting at $4,299.99) is an example of abundance in the world of gaming laptops, built around an 18-inch display and the latest flagship silicon: a Core Ultra 9 290HX Plus with Nvidia’s GeForce RTX 5090 in our test unit. This machine makes a loud first impression, from its stellar (albeit tricky to configure) mini-LED display to the unique scrolling marquee lighting on its lid. But at this price - $4,999.99 as tested – the Scar 18 must prove it can hold the line against Razer’s Blade 18 before it can claim a spot at the top.

Design of the Asus ROG Strix Scar 18 (2026)

At 15.71 x 11.73 x 1.38 inches, the Scar 18 has the footprint of a cafeteria tray – this isn’t a laptop you’ll be getting out on a plane. And at 8.16 pounds, this is also one of the heaviest laptops on the market. But performance is the goal here, not portability. Razer’s Blade 18 (15.74 x 10.84 x 1.1 inches) is thinner and significantly lighter, at 7.06 pounds.

The vibrant Aura Sync lightbar around the base of the laptop also demands attention, as does the RGB-lit Republic of Gamers logo on the lid. Both are configurable with customizable lighting and patterns in the Armoury Crate app.

The lid also has a special feature called AniMe vision, a diagonally scrolling marquee of text via LEDs shining through holes in the lid backing. (This is extremely similar to the AniMe Matrix that debuted on the Zephyrus line years ago.) There are several preconfigured versions of the Republic of Gamers logo, and you can add your own text effects. Layered effects are possible and don’t always produce the desired effect — I had a “raining” effect enabled at the same time as my text, and the text was almost impossible to make out.

The bottom line is that the Scar 18 couldn’t do anything more to look like a gaming laptop – it is designed to be seen. Build-wise, it’s a solid machine, showing minimal flex no matter how I handled it. Only the lid is metal, with the rest of the construction thick plastic.

Connectivity is thoroughly modern: two Thunderbolt 5 (USB-C) and three USB-A 3.2 Gen 2 ports, HDMI 2.1, an audio combo jack, and 2.5 Gbps Ethernet. Internally, it offers Wi-Fi 7 and Bluetooth 5.4 from an Intel BE200 networking card. The power connector is proprietary for the 450 W power brick.

Specifications

Gaming and Graphics on the Asus ROG Strix Scar 18 (2026)

We tested the ROG Strix Scar 18 in flagship form, featuring a Core Ultra 9 290HX Plus processor, RTX 5090 graphics card (175 W maximum graphics power), and 32GB of RAM. This is top-of-the-line gaming technology, though with one misstep: single-channel RAM. This might affect its performance as we’re about to see. The Task Manager confirms that only one SO-DIMM slot was used.

I put the Scar 18 through its paces playing 007: First Light at 3840 x 2400 with all detail settings maxed out. At first, this proved too demanding – I saw 26 to 32 frames per second (FPS) in most scenes. Enabling DLSS more than doubled the frame rate – I saw around 70 FPS or better, and the game was supremely playable.

Our comparison lineup includes Alienware’s 16 Area-51 (RTX 5080), MSI’s Raider 16 Max HX (RTX 5090), and Razer’s Blade 18 (RTX 5090). All laptops use a Core Ultra 9 290HX Plus and have GPUs rated for 175 W like our Asus. Their native screen resolutions, however, are different: Alienware and MSI are 2560 x 1600 while Razer has a unique dual-model display supporting both 1920 x 1200 and 3840 x 2400.

The Scar 18 was competitive across the board at 1200p, typically a few FPS ahead of the Alienware but a few FPS behind the MSI and particularly the Razer.

Bumping the resolution to 4K, the Scar 18 trailed the Razer in most games – the delta was at or almost 10% in Shadow of the Tomb Raider, Far Cry 6, Cyberpunk 2077, and F1 23. (Red Dead Redemption 2 was the exception.) While those numbers won't make the difference between playability and unplayability, the price of these laptops makes it difficult to overlook.

Differences versus the Razer aside, the Scar 18 still demonstrates ample performance for gaming at 4K in most of the games we tested, though not all – it averaged only 21 FPS in Cyberpunk 2077 on ray tracing ultra, indicating that it won't be possible to play every game at maximum detail settings.

We stress test gaming laptops running 15 loops of the Metro Exodus stress test at RTX settings. During the test, the Scar 18 averaged 141 FPS with minimal variance between runs, starting at 141.5 FPS and finishing at 141.1 FPS. The Core Ultra 9 290HX Plus CPU averaged 4.59 GHz on its P-cores and 2.58 GHz on its E-cores while the RTX 5090 had an average boost clock of 1.98 GHz.

Productivity Performance on the Asus ROG Strix Scar 18 (2026)

We tested the Scar 18 with a Core Ultra 9 290HX Plus CPU, 32GB of RAM, and a 1TB PCIe 4.0 SSD. Not including a PCIe 5.0 drive seems like a missed opportunity at this price, though Razer does the same thing.

In Geekbench 6, the single-core numbers between these laptops were almost indistinguishable as we might expect given they all use the Core Ultra 9 290HX Plus, though the Scar 18’s multi-core score of 17,629 points significantly trailed the others, which all scored over 20,000 points.

The Scar 18 landed middle of the road in our 25GB file transfer test, averaging 2,042.69 MBps, putting it ahead of the MSI (1,357.93 MBps) and Razer (1,670.53 MBps) but behind the Alienware (2,738.9 MBps).

The Scar 18 also proved competitive in our 4K-to-1080p Handbrake transcoding test, completing it in two minutes and 11 seconds to lead the Alienware (2:24) but trail the MSI (1:51) and Razer (2:01).

Display on the Asus ROG Strix Scar 18 (2026)

The Scar 18’s “Nebula” HDR display is its crowning feature. With a 3840 x 2400 (4K) resolution, mini-LED backlighting with 2,000 dimming zones, a 240 Hz refresh rate, and Nvidia G-Sync, this is quite advanced.

Tweaking is required to get this display to perform as intended, and it’s not simple. Out of the box, HDR is disabled, the refresh rate is capped at 120 Hz, and G-Sync is unavailable since Nvidia Optimus, which dynamically switches between the CPU’s integrated graphics and the RTX 5090, is enabled. To unlock maximum performance, the GPU must be put in “Ultimate” mode in Armoury Crate, which is effectively a MUX toggle that disables the integrated graphics. A restart is required for this to take effect. G-Sync, the 240 Hz refresh rate, and a special feature called “Extreme Low Motion Blur” (ELMB) then become available. The latter is aimed at esports players– it manipulates the pixels so that they turn off when switching colors, theoretically eliminating blur. (More on this in a moment.)

Those aren’t the only settings you’ll need to know about. You can toggle the mini-LED backlighting control between one zone, multi-zone balanced, or multi-zone strong. These settings produce very different images – one-zone provides the deepest contrast, multi-zone balanced is the dimmest but evens out the contrast to make dark scenes appear brighter, and multi-zone strong is the brightest and most vibrant. I stuck with the latter for nearly everything.

ELMB only works in one-zone mode without HDR. I tested it using the Blur Blusters UFO Test. It clearly made a difference – in the 240 fps scrollbar, the UFO looked crisp moving across the screen. Disabling ELMB caused it to become blurry, making it harder to see details. This feature can really matter for competitive esports.

But wait, there are even more settings! Armoury Crate includes many color modes through a feature called GameVisual — racing, scenery, RTS/RPG, fps, cinema, eyecare, vivid, and e-reading. On top of that, it also provides color temperature and gamut settings.

Then there’s the question of HDR. To get that working, it must first be enabled in the Windows Settings app. Back in Armoury Crate, you’ll find GameVisual, color temperature and gamut settings, Extreme Low Motion Blur, and mini-LED backlighting settings are no longer available. That’s the trade-off.

Complicating all this is that the settings I just mentioned are in different places in Armoury Crate. Some are in the display section, while others require going into the Devices section, selecting the Scar 18, and making changes there. It’s not straightforward, and those that simply use this laptop out of the box without tweaking won’t get the best visual experience.

After much experimentation, I played 007: First Light in GPU Ultimate mode, G-Sync enabled, a 240 Hz refresh rate, and True Color HDR enabled through Windows. The picture left little to the imagination – colors seemed to pop off the screen and the 4K resolution provided exquisite detail right down to the patches on Bond’s uniform. HDR effects from muzzle flashes and explosions were dazzling in dark environments.

When I watched Zootopia 2, I switched off HDR and used SDR multi-zone strong. Colors looked exquisite, and brightness was borderline excessive in a darker room. Bright objects like lamps almost seemed overexposed, but they weren’t – the display was simply that bright, and colors were so saturated that I found it hard to look away.

The Scar 18’s numbers are from its default out-of-the-box display settings. In color coverage, its 77.9% DCI-P3 coverage was last in the group – Alienware's OLED screen achieved 93.7% — but is still high enough to create vibrant-looking colors. Its 428.2-nit peak brightness was mid-pack, brighter than the Alienware's 368.6 nits but well back from Razer's 538 nits.

Also shown in our charts are the multi-zone strong settings, which produced 584.6 nits of brightness, with parts peaking at 625 nits. Enabling HDR, we measured an astounding 1,124 nits at 10%, 1,090 nits at 40%, and 943 nits at 100%. If you're looking for one of the brightest laptop displays around, the Scar 18 ranks high on the list.

Keyboard and Touchpad on the Asus ROG Strix Scar 18 (2026)

The Scar 18’s keyboard is great for gaming – the keys require enough actuation force that resting your fingers on WASD or the arrow keys won’t produce accidental presses. Key travel is communicative in the sense your fingers know exactly when a key is at the top or bottom of the stroke. The bright RGB backlight is sharp and easy to see.

The keyboard is less ideal for productivity. The tactile feel is rather lifeless, though I still managed 126 words per minute with 99% accuracy in my usual MonkeyType run. Layout-wise, a two-thirds-size number pad on an 18-inch laptop is a miss – there’s plenty of space to make it full-size. Additionally, the arrow key cluster isn’t separated out, resulting in no right Ctrl key, and there are no dedicated Home, End, Page Up, or Page Down keys. Asus does, however, provide five dedicated macro keys, a rarity on any laptop. These are configurable in the Armoury Crate app.

Asus’ mechanical touchpad is excellent, with an expansive surface and a smooth but fingerprint-resistant surface coating. Its clicking action is quiet.

Audio on the Asus ROG Strix Scar 18 (2026)

The Scar 18’s quad-speaker array delivers a decent, if not remarkable, audio experience. In 007: First Light, soft details like the footsteps of approaching enemies were easy to pinpoint, thanks to the expansive soundstage – there’s plenty of room to separate the speaker placement on a laptop this large. Bass is muted, though, resulting in explosions and gunfire that don’t stir up as much excitement as they could.

In Phil Collins’ “Don’t Lose My Number,” high hats on drum hits were sharp but missed low-end bump. Switching to the Chainsmokers’ “Summertime Friends,” I also noted the lack of bass, though the vocals were crisp. The overall sound signature is on the hollow side, but that can be sharpened up using the Atmos Detailed equalizer in the Dolby Access app. None of the equalizers made up for the lack of bass, though. Volume levels are also moderate – I found myself pushing at least 80% volume for most situations.

Upgradeability of the Asus ROG Strix Scar 18 (2026)

Getting inside the Scar 18 couldn’t be easier – simply slide the latch below the palm rest, slide the entire bottom panel forward, and lift it away. You don't even need tools.

Upgrade possibilities include two M.2 slots, two SODIMM slots, and the battery.

Battery Life on the Asus ROG Strix Scar 18 (2026)

Our battery test consists of web browsing, running OpenGL tests, and streaming videos with the screen at 150 nits while connected to Wi-Fi.

One minute shy of the five-hour mark, the Scar 18 demonstrates respectable battery life for an 18-inch gaming laptop. The Razer lasted half an hour longer (5:31) and the MSI Raider (8:34) clearly does a better job conserving power, but the Scar 18 did outperform the Alienware (3:33) by several hours.

Heat on the Asus ROG Strix Scar 18 (2026)

We measure the surface temperatures of gaming laptops while running the Metro Exodus stress test. Peak temperatures were 91 degrees Fahrenheit on the keyboard between the G and H keys, 90 F on the touchpad, and 108 F on the underside near the cooling vents. Internally, the Core Ultra 9 290HX Plus averaged 66 Celsius while the RTX 5090 ran at 64 C.

The laptop’s fans are well-behaved for daily use. Though fan noise increases while gaming, I had no trouble hearing footsteps and distant conversations in 007: First Light using the built-in speakers.

Webcam on the Asus ROG Strix Scar 18 (2026)

Asus’ FHD webcam has the minimum resolution expected on a modern laptop. The picture looks soft and washed out. Highlights aren’t handled that well – a window in the background appeared blown out – and I had trouble making out details on my face from just a few feet away. Gamers who value visual quality will want to invest in an external webcam.

Software and Warranty on the Asus ROG Strix Scar 18 (2026)

Asus includes a useful software bundle, starting with the familiar Armoury Crate. This app provides component monitoring, a macro editor, game library, an exhaustive amount of display settings, and lighting settings via Aura Sync and AniMe Vision. Accessing some settings is unintuitive since you need to go to the Device section and select the laptop. There you can access Windows key and Touchpad toggles and several display settings, including panel overdrive (240 Hz refresh rate). Most settings can be saved in profiles.

The MyAsus app is more generic. In addition to diagnostics and system updates, it provides a battery care mode, microphone noise cancelation, and networking preferences that allow prioritizing traffic to games or other apps.†

The Scar 18 also works with Asus’ GlideX app to share content across devices, including phones and tablets.

Asus includes a standard one-year warranty.

Configurations

Asus offers two Scar 18 configurations with only the GPU different between them – model G835LWG-DB96 uses an RTX 5080 for $4,299.99 while our review model, G835LXG-DB96, steps up to the RTX 5090 for $4,999.99. All other components are the same: a Core Ultra 9 290HX Plus processor, 32GB of RAM, a 1TB SSD, and the 18-inch mini-LED display.

Pricing is slightly higher than Razer’s Blade 18 with the RTX 5080 – it was $4,099 at this writing. Razer runs $5,399 with the RTX 5090, but that price also includes a 2TB SSD.

Bottom Line

The Scar 18 is an undeniably impressive machine that goes all-in on visuals. Its mini-LED “Nebula” display looks breathtaking when properly configured, producing exceptional brightness. The AniMe scrolling marquee, dedicated macro keys, and easy serviceability also elevate its appeal.

However, when it comes to performance, the Scar 18’s single-channel RAM and lack of a PCIe 4.0 SSD are significant shortcomings on a $4,999.99 machine. Several of our gaming benchmarks and multi-core CPU performance showed meaningful dips against Razer’s Blade 18. Additionally, while its display is brilliant, the maze of settings required to unlock its potential means it doesn’t provide the best experience out of the box.

Overall, the Scar 18 is a formidable and visually stunning laptop with plenty of power and one of the best displays you’ll find in a laptop. It simply doesn’t perform consistently enough to displace the Blade 18 as our top recommendation among elite 18-inch gaming laptops.

Meta’s surprise purchase of Manus, a Chinese startup known for its advanced AI agents, caught Beijing by surprise and ordered the two companies to unwind the $2 billion deal. The Chinese tech giant Tencent, which was among the startup’s initial investors during early funding rounds, is taking the lead in buying back the startup at the same price. According to the Financial Times, other former investors, including ZhenFund and HSG — China-based venture capital firms — while former U.S. investors like Benchmark are unlikely to join the potential consortium.

This move marks Beijing’s increasing protectiveness of its AI companies and experts, which it considers strategic assets in its heated rivalry with the U.S. We can see this in the Chinese government’s five-year plan, which is doubling down on technological self-reliance. It has even gotten to the point that AI experts, even those working in private firms, are now required to secure approval before traveling internationally.

U.S. tech giants are investing billions of dollars to develop their AI models, even dangling hundred-million-dollar bonuses to hire AI experts — one AI founder even claimed that Meta offered a $1.25-billion bonus. It seems that China is trying to avoid a situation where its experts are enticed to work for American AI tech companies, with the Financial Times reporting that Chinese officials are calling Meta’s acquisition of Manus “a conspiratorial attempt to hollow out China’s technology base.” The order to undo the deal means that Meta cannot use Manus’ intellectual property, nor can it have its founders and employees working for the company. Still, the U.S. tech giant has had a few months to study its models and engineering expertise.

Meta has already agreed to undo the deal, with most of Manus’ operations reportedly running independently of the company. However, the Chinese startup still needs to break financially from the American tech giant by paying back the $2 billion the latter spent to purchase it. Even though Chinese companies are also investing massive amounts in AI tech, it’s still not easy to raise this amount of capital in such a short period.

Tencent, which owns the WeChat platform used by China’s 1.4 billion population for messaging, social networking, mobile payments, ride-hailing, food delivery, and more, believes that Manus would be an asset for the company. Aside from reaching an annual revenue of $500 million, its AI agent would also mesh well with the company's increasing AI focus. “Beyond foundation models, it has become increasingly evident that agentic AI represents a breakthrough use case,” Tencent president Martin Lau said in its May earnings call. “Our platform inherently has many benefits of hosting AI agents.”

SK hynix has completed the largest-ever foreign company IPO in U.S. history, raising $26.5 billion in its Nasdaq debut today, July 10. The South Korean memory giant sold 177.9 million American depositary receipts (ADRs) — a U.S.-listed stand-in for a foreign share — at $149 apiece, each representing a tenth of a Seoul-listed share. The offering was more than seven times oversubscribed and drew demand from more than 500 investment firms, according to Financial Times. Temporary Nasdaq trading is underway under the ticker SKHYV before regular-way trading begins as SKHY on Monday, July 13.

The offering was led by Bank of America, Citigroup, Goldman Sachs, and JPMorgan, with nine additional firms rounding out a 13-bank syndicate. Anchor demand came from heavyweight institutions including Baillie Gifford, Coatue Management, and Situational Awareness Partners, which together signaled interest in as much as $7 billion of stock, according to people familiar with the matter cited by Financial Times.

SK hynix is the world's leading maker of high-bandwidth memory (HBM), the vertically stacked DRAM that has become critical infrastructure for AI accelerators. The company has said it will steer the proceeds toward boosting its AI-memory manufacturing capacity. Confirmed build-outs include the first-phase fab at the massive Yongin semiconductor cluster, a new P&T7 advanced-packaging line in Cheongju, and EUV lithography equipment slated for delivery by the end of next year. Separately, SK hynix is constructing its first U.S. production site, a $4 billion advanced-packaging plant in West Lafayette, Indiana, targeted for completion around 2028. The facility is eligible for up to $458 million in CHIPS Act grants and up to $570 million in federal loans.

What display resolution do you use on your primary monitor?

SK hynix is seeing sensational growth thanks to the ongoing AI boom. The company is reportedly on track to post over 200 trillion won ($133 billion) in operating profit this year, a record-breaking figure that would see SK hynix employees earn around $400,000 each in bonuses. The company’s Seoul-listed stock is up roughly 220% year-to-date and has climbed more than sixfold over the past year.

In late June, SK hynix briefly surpassed Samsung as South Korea's most valuable company, closing at around 2,080 trillion won (about $1.35 trillion), a meteoric rise for a company that almost declared bankruptcy in 2001 and, more recently, recorded an annual operating loss of 7.73 trillion won in 2023. That rise doesn't seem like it will be slowing down any time soon. SK hynix has said its entire 2026 output of HBM, DRAM, and NAND is already sold out, with the crunch expected to extend into 2027.

Nanya Technology plans to increase its capital expenditure to more than TW$200 billion ($6.2 billion) in 2027, roughly four times its budget for this year, President Pei-Ing Lee said during an online briefing. The Taiwanese memory maker reported unaudited second-quarter revenue of T$82.55 billion, up 684% from 2025, and net income of T$50.19 billion, up 1,324%. Gross margin reached 79.5%, against a negative 20.6% during the same quarter of 2025. That single quarter's profit is 7.6 times what Nanya earned across the entirety of last year, and the quarter's revenue exceeds the company's entire 2025 sales.

Nanya spent T$13.2 billion on capex in 2023, T$16.1 billion in 2024, and T$13.4 billion in 2025, and has budgeted up to T$52 billion for 2026, per its Q1 investor presentation. Those four years together come to T$94.7 billion, less than half what Lee intends to spend in 2027 alone. Lee, however, admits that the 2027 figure is preliminary and hasn’t yet gone to the board, and that the new plant will absorb about T$480 billion at full capacity.

Nanya's average selling price climbed more than 70% quarter over quarter in Q1 2026, while its bit shipments fell by a mid-single-digit percentage, its own results deck shows. The company is targeting bit shipment growth in the teens for the full year, so almost all of the 684% revenue increase is due to price. Meanwhile, TrendForce projects a further 13% to 18% rise in conventional DRAM contract prices in Q3.

Roughly 70% of Nanya's shipments are DDR4 and low-power DDR4, Lee said at a January earnings conference, and DDR5 contributes about 10% of revenue. Nanya builds no high-bandwidth memory, and Lee has ruled out competing in HBM2, HBM3, HBM3E, or HBM4. A customized HBM part for edge AI, developed with Etron Technology, Piecemakers Technology, and Formosa Advanced Technologies, is targeted for the end of this year, however. Its 79.5% margin is close to a pure reading on conventional DRAM, and it sits within six points of the 85% consolidated gross margin Micron reported in its most recent 10-Q with HBM in the mix.

SanDisk, Kioxia, Solidigm, and Cisco paid T$78.72 billion for 10.19% of Nanya in a private placement completed in April, with SanDisk and Kioxia signing long-term DRAM supply agreements alongside the equity. Three of the four make SSDs and need DRAM for cache. Solidigm is a subsidiary of SK hynix, the world's second-largest DRAM maker, and it went to a supplier holding roughly 2% of the market to source it.

The first phase of Nanya's new fab in New Taipei City's Taishan District reaches 30,000 wafers per month in 2028 and expands to 45,000 later, Lee said Friday. The plant will run Nanya's 1B node, its second-generation 10nm-class process, to build DDR5, DDR4, and low-power DDR4, the company said during a March briefing. Lee said at that briefing that the most severe supply constraints run through the first half of 2027 and that the shortage persists into 2028. Samsung's P3 fab alone is expected to reach around 115,000 wafers per month by the end of this year.

Japanese chipmaker Rapidus will try to lure customers away from TSMC not only by offering a different kind of service, but also by offering its manufacturing services at lower prices, chief executive Atsuyoshi Koike announced this week. The company's plan to rival TSMC in terms of pricing appears on the surface as a risky move, as the company moves to develop leading-edge process technologies.

At present, Rapidus is looking at charging ¥3 million – ¥3.5 million ($18,550 - $21,635) per wafer processed using its 2nm-class fabrication process, which is significantly below TSMC's rumored quote of around $30,000 per N2 wafer, and is comparable to what Samsung is rumored to offer with its SF2 manufacturing technology, set at $20,000 per-wafer. Actual prices will depend on exchange rates, though Rapidus' general idea of offering significantly lower quotes than TSMC is immediately apparent.

Rapidus plans to start high-volume manufacturing (HVM) using its 2nm-class fabrication technology by the second half of 2027. The ramp of a new fab will take some time, so expect meaningful volumes from Rapidus to only be produced in 2028, when TSMC's N2 will no longer be its leading-edge node.

By the time Rapidus starts HVM at its IIM-1 in 2027, TSMC will have ramped production of chips using its performance-enhanced N2P manufacturing node, and the company will also absorb all the yield learning with gate-all-around the company will have with its N2 present at five fab modules. Furthermore, by the time Rapidus reaches meaningful volumes at IIM-1 in 2028, TSMC will have ramped up production using its advanced A16 fabrication process with Super Power Rail backside power delivery as well as a 3^rd-generation 2nm-class node named N2X.

In addition to the vast 2nm-capable capacity and process maturity that should be kept in mind when comparing Rapidus with TSMC, there is another factor to consider. One of TSMC's major advantages over its rivals is its Open Innovation Platform (OIP) ecosystem, which includes comprehensive electronic design automation tools, silicon-proven IPs, even for the latest nodes, a host of contract chip designers, and advanced packaging services not only from TSMC but also from its partners. For now, neither Rapidus nor Intel and Samsung Foundry can offer anything close to TSMC's OIP.

Given the advantages that TSMC will likely have over competitors with its 2nm-class fabrication technologies in 2028, lower pricing may be among the few ways to compete against the world's largest foundry. Rapidus' strategy of offering lower quotes while operating a single fab does not seem like the best way of earning money, but perhaps a certain way to lose it.

However, Rapidus may have another ace up its sleeve with single wafer processing across all process steps. The approach will greatly speed up the production cycle, which will be its indisputable advantage over other chipmakers, albeit at the cost of tool usage efficiency. Will lower quotes and shorter production cycles be enough for Rapidus to win customers from TSMC? Only time will tell.

Rapidus is reportedly negotiating with more than 60 potential customers, mainly overseas companies, which demonstrates the company's ambitions to become a viable rival to the global leader TSMC as well as contract chipmakers Intel Foundry and Samsung Foundry.

If you're browsing the web, a VPN is one of the fastest ways to upgrade your privacy. Luckily, the cost of a two-year subscription to one of the biggest VPN providers, ExpressVPN, is now down to only $69.72 right now. You'll be saving $378 in total if you buy a sub this way, compared to a rolling monthly sub, and you'll get an extra four months added onto your plan for free, too.

● Check out this deal at ExpressVPN

VPN pricing is a little unorthodox, and we see deals like these appear all the time, so it's important to point out where the value really lies here. I spent a long time paying month by month for a VPN subscription, not realizing how much money I was actually spending by choosing a rolling, 30-day sub over locking in for a longer period. The $69.72 cost means that you're paying the equivalent of just $2.49 a month for the 28 months of access you're gaining here, compared to $15.99 a month for the 30-day access.

This deal for ExpressVPN's Basic plan unlocks a solution that'll protect 10 devices simultaneously with the same account. You can use it to keep your identity safe while you browsing, but if you're still unsure, new customers can take advantage of ExpressVPN's 30-day money-back guarantee. If you're unhappy, you can ask for a full refund within those first 30 days, giving you enough time to fully test the service to see if it matches up to your expectations.

ExpressVPN is a no-logs provider, meaning it can't give up your data while you're using it, as it doesn't track or store any data about its users' connections. This is regularly audited, with the last audit in 2025 by KMPG confirming "reasonable assurance" of its systems and policies. This is real peace of mind for users: you're free to use the web how you want.

It also gives you a shield of privacy that your regular ISP just can't. A VPN sub like this means you can shop online or browse the web without trackers monitoring where you're from via your IP address. Your location stays hidden, obfuscated through the 105 different countries that ExpressVPN hosts servers, with 24 different server locations in the U.S. alone.

Don't underestimate how important that privacy is in the modern world. Every page you're visiting leaves a trace, whether it's a ping in a server log or a tracker on-page, giving the site owner data to build a profile on who you are, but using a VPN can stop this data from being useful. Likewise, a VPN connection can help you to connect as if you're from a certain location, which is particularly useful if you're traveling abroad and want to be able to access your home streaming services. It's also a good idea to use a VPN as a way to protect yourself if you're accessing the internet from a public WiFi network, where malicious actors could otherwise be snooping on your connection.

Using ExpressVPN, the encryption between your device and its servers is encrypted using industry-standard AES-256 encryption, as well as post-quantum encryption techniques for the initial handshake process. This can stop your data from being intercepted while you're accessing the web. If you lose your connection, features like ExpressVPN's kill switch will stop the data from leaking out, blocking any access until you reconnect. The Basic plan also includes a private email relay service to give you 10 anonymous email aliases to use for registering online, along with basic ad and malicious site protection for your browser.

This $69.72 price tag for the ExpressVPN Basic plan is a good deal for a 28-month privacy upgrade. Yes, VPN pricing can seem confusing, but by locking in for that period, you're getting online privacy protection for the equivalent of just $2.49 a month. You can always give it a go and, if you don't like it, request a refund using ExpressVPN's 30-day money-back guarantee.

If you're looking for more savings, check out our Best PC Hardware deals for a range of products, or dive deeper into our specialized SSD and Storage Deals, Hard Drive Deals, Gaming Monitor Deals, Graphics Card Deals, Gaming Chair, Best Wi-Fi Routers, Best Motherboard, or CPU Deals pages.

You can also join the Tom's Hardware deals Discord for up-to-the-minute hardware deals.

Researchers in Korea and Japan have presented two separate memory-integration proposals that aim to increase HBM (High-Bandwidth Memory) capacity and bandwidth without trapping more heat inside ever-taller DRAM (Dynamic Random Access Memory) stacks, one of the most pressing challenges facing future AI accelerators. Presented at the 2026 IEEE/JSAP Symposium on VLSI Technology and Circuits held in June, the two approaches — V-Die from a Korean research collaboration and MOSAIC from a University of Tokyo-led group — both explore the same broad idea of standing DRAM memory dies on their edges instead of stacking the memory dies only upward like conventional HBM.

The Korean proposal, called Vertical-Die (V-Die), was presented by researchers at the Ulsan National Institute of Science and Technology (UNIST). The design rotates custom DRAM dies upright, drops through-silicon vias to free die area for more memory cells, gives each die its own bottom-edge I/O, and runs liquid-cooling channels between adjacent dies. In simulations against an HBM4 system at equal capacity, the V-Die system reportedly achieved 540 tokens per second on a GPT-3-sized workload, compared to 296 tokens per second for HBM4.

The Japanese project, MOSAIC, takes a similar “sideways stack” idea but focuses on the practical difficulty of connecting so many vertical dies to a GPU or package substrate. Presented by University of Tokyo researchers, the MOSAIC work uses orthogonal die stacking and a contactless die-to-die interface, in which data is transferred through tiny inductive coils rather than requiring every signal pad to land perfectly on a physical contact. The researchers say the prototype interface achieved up to 4 Gbps per channel, while the memory structure could double HBM4-class capacity in a DRAM-on-GPU configuration.

Both projects aim to solve the growing problem of AI chips being held back by memory. Modern accelerators can perform enormous amounts of computation, but large, powerful models depend on moving huge amounts of data between memory and compute. This is why HBM has become one of the defining technologies of modern AI hardware.

The technology addresses the memory wall by stacking multiple DRAM dies vertically on a base die and placing that stack very close to the processor. Nvidia's Blackwell Ultra B300, for instance, carries up to 288GB of HBM3E memory, without which much of the silicon would sit idle waiting for data. The dies are connected via through-silicon vias (TSVs) — tiny vertical channels etched through the silicon and filled with metal.

The stack then communicates with the GPU over an extremely wide interface, often routed through a silicon interposer or an advanced package. This is the core reason HBM can deliver terabytes per second of bandwidth: it uses a very wide, very short data path instead of sending memory traffic across a motherboard, as with conventional DIMMs (Dual In-line Memory Modules), physical sticks of RAM used in computers.

However, that same structure creates several problems. While taller stacks add more capacity, they also make it harder to remove heat. Heat generated in the lower dies and at the high-speed interface must pass through layers of silicon, bonding materials, underfill, and package structures before it reaches a heat spreader. Furthermore, TSVs consume die area that could otherwise be used for memory cells, and as bandwidth rises, more routing and I/O place additional pressure on both signal integrity and packaging costs.

HBM4, the latest generation of HBM, addresses a number of these challenges. Meanwhile, companies such as SK hynix, Samsung, and Micron are racing to improve speed, capacity, base-die performance, and thermal management. SK hynix has already shown iHBM, which embeds cooling elements into the HBM interface area, and Samsung has shown an HBM5 mockup with Heat Path Block cooling to more directly extract heat from the stack. However, they all retain the same upward stacking structure.

This convention is what V-Die and MOSAIC are challenging. By standing DRAM dies upright, the researchers expose far more silicon surface area to the cooling path. In theory, this turns the memory stack into something closer to a heat-sink fin array, where heat can move laterally and escape more directly instead of being trapped in the middle of a thick vertical pile. It also opens the door to new connection schemes along the bottom or side of each die, rather than forcing every die to communicate through TSVs running vertically through the stack.

For V-Die, the key shift is removing TSVs from the memory dies and replacing them with bottom-edge connections. Each DRAM die gets its own I/O along the bottom edge and connects directly to the substrate, with links reportedly spaced every 20 microns. The team says this layout gives four times as many connections as HBM4 and cuts memory read time by 37%, although some signals must travel farther across the package to reach the processor.

Cooling is the other half of the V-Die argument. The proposal places microfluidic cooling channels between adjacent upright DRAM dies, allowing coolant to dissipate heat closer to its source. According to the researchers, this could keep the stack around 45°C, far below the 80°C-plus range associated with dense HBM systems. In a simulated 16-die stack matched to H100-class hardware on a GPT-3-scale model, V-Die hit 540 tokens per second, compared to HBM4's 296, and cut first-token latency by 32%, or about 24 milliseconds.

MOSAIC, meanwhile, is focused on making the sideways stack manufacturable. Because the dies are assembled flat and then turned on edge, even a few microns of die-thickness variation across dozens of dies can add up to an alignment miss where the signal pads no longer land. The Japanese team’s answer is a contactless interface based on inductive coupling. One side of the memory die carries oblong coils, while a corresponding set of coils sits on the substrate or mating chip. Current in one coil induces a signal in the other, allowing data to cross the small gap without a direct metal-to-metal signal contact. This eliminates the need for precise overlapping, giving the package greater tolerance for assembly variation. Power, which requires fewer, larger connections than data, can still be supplied via physical contacts on the sides of the memory cube.

The VLSI MOSAIC prototype achieved up to 4 Gbps per channel and demonstrated TSV-free 3D integration for a memory-on-GPU layout. The team says the approach can enable twice the memory capacity of HBM4 without significantly increasing peak temperature. A related bump-MOSAIC hardware demonstration at ECTC used 100-micron-pitch microbumps, achieved stacking alignment within 6 microns as verified by X-ray CT, and showed a configuration with three times the thermal conductivity of conventional stacking while adding up to 30% more memory capacity.

While the results look promising, neither V-Die nor MOSAIC is close to replacing commercial HBM. Neither is close to shipping. V-Die is still a proposed architecture, with a prototype in the works to validate its thermal and electrical behavior; MOSAIC has proof-of-principle hardware, but the researchers have yet to show it scales to commercial DRAM capacity, yield, cost, and reliability.

Still, any viable solution to the multifaceted AI memory problem is a welcome development. SoftBank and Intel’s Z-Angle Memory (ZAM) and NEO Semiconductor’s 3D X-DRAM — both still in development — aim to solve the constraints of conventional memory. Meanwhile, the overall market is already feeling the squeeze on price and availability, even as memory makers divert capacity toward the more lucrative AI HBM and server products, driving consumer RAM prices even higher.