Kandou AI, a Swiss semiconductor company that builds chip-to-chip interconnect technology, has raised $225 million in what it calls a Series A round, led by Maverick Silicon with strategic participation from SoftBank, Synopsys, Cadence Design Systems, and Alchip Technologies. The round values the company at $400 million. The label is worth pausing on: Kandou was founded in 2011 and previously raised more than $163 million across Series B and C rounds under the name Kandou Bus. The “Series A” designation reflects a rebrand and leadership change, not a fresh start.
The company’s new chief executive, Srujan Linga, a former Goldman Sachs managing director, took over in 2025 from founder Amin Shokrollahi, an EPFL professor of mathematics and computer science who invented the core technology. Shokrollahi’s contribution, a signalling method called Chord that sends correlated signals across multiple wires to increase bandwidth by a factor of two to four while halving power consumption, remains the technical foundation. The rebrand to Kandou AI and the repositioning toward artificial intelligence infrastructure is Linga’s doing, and it appears to have worked: the $225 million raise is the largest in the company’s history and brings SoftBank, one of the most aggressive AI infrastructure investors, onto the cap table.
What makes Kandou AI’s position unusual is not the problem it is trying to solve but the material it proposes to solve it with. The AI industry’s interconnect bottleneck is real and well documented. As models scale to hundreds of billions of parameters and training clusters expand to tens of thousands of GPUs, the speed at which data moves between processors and memory has become the binding constraint on performance. At signalling speeds of 224 gigabits per second, traditional copper interconnects consume roughly 30 per cent of total cluster power, with signal degradation so severe that reach is limited to less than a metre without amplification.
The prevailing industry response has been to move to optics. Ayar Labs raised $500 million in March 2026 at a $3.8 billion valuation for its co-packaged optical interconnects. Marvell completed a $3.25 billion acquisition of Celestial AI in February, buying photonic fabric technology that claims 25 times the bandwidth of copper alternatives at a tenth of the latency. The optical interconnect market for AI data centres is projected to grow from $3.75 billion in 2025 to $18.36 billion by 2033.
Kandou AI is betting that copper is not finished. Its Chord signalling technology, the company claims, can achieve path-to-Shannon-capacity efficiency, reducing power consumption and system costs by a factor of ten while extending copper links to 448 gigabits per second and beyond. If that claim holds, it would mean that the billions being spent on optical interconnect transitions are at least partially premature, and that existing copper infrastructure can be made to work for several more hardware generations at a fraction of the cost.
The composition of the investor syndicate matters more than the headline figure. Synopsys and Cadence are the two dominant providers of electronic design automation tools. Their participation is not purely financial; it signals potential integration of Kandou AI’s serialiser/deserialiser intellectual property into the design flows that chip architects use to build processors and memory controllers. Alchip, a Taiwanese ASIC design services company, provides a path to manufacturing. SoftBank, which has invested more than $100 billion in AI-adjacent companies through its Vision Fund and direct investments, adds the scale capital and the strategic network.
The practical implication is that Kandou AI’s technology could appear inside chips designed by other companies rather than requiring customers to adopt Kandou’s own silicon. This is a licensing and IP model, similar in structure to Arm’s approach in mobile processors, and it is a more capital-efficient path to market dominance than manufacturing and selling chips directly. Whether Kandou can execute on that model with a $400 million valuation and $225 million in fresh capital, against optical competitors valued at ten times as much, is the central question.
At $400 million, Kandou AI is valued at roughly a tenth of Ayar Labs and an eighth of what Marvell paid for Celestial AI. That gap could reflect market scepticism about copper’s longevity in AI infrastructure, or it could reflect the fact that Kandou’s technology, if it works as claimed, does not require the industry to rip out its existing wiring. Copper is already in every data centre. If Kandou’s signalling technology can make it fast enough for another generation of AI workloads, the adoption curve would be faster and cheaper than an optical transition.
The risk is that “another generation” may not be long enough. AI model sizes and training cluster scales are growing at a pace that consistently outstrips infrastructure predictions. What is adequate at 448 gigabits per second today may be inadequate at the terabit-per-second speeds that next-generation models will demand within two to three years. Optical interconnects, for all their cost and complexity, offer a higher theoretical ceiling.
Kandou AI’s $225 million buys it time to prove that the ceiling can wait. The company’s 15-year history and the technical credibility of Chord signalling, which has been deployed commercially in consumer electronics since the mid-2010s, lend substance to the bet. But the AI infrastructure market has a pattern of rewarding ambition over incrementalism, and a company arguing that the existing material is good enough faces a harder narrative sell than one promising to replace it entirely. The investors on this round appear to be betting on engineering pragmatism. Whether the market agrees will depend on how quickly the optical transition matures, and whether Kandou’s copper can keep pace with an industry that has shown little interest in waiting for anything.
Five years years ago AT&T effectively stopped selling DSL and started hanging up on DSL and copper phone line customers. While killing landlines and DSL is understandable given the limitations of the dated copper-based tech, the problem is that thanks to concentrated telecom monopolization, many of these customers were left without any replacement options due to a lack of competition.
There are other issues at play too. AT&T has, for decades, received countless billions in tax cuts, subsidies, merger approvals, and regulatory favors in exchange for building infrastructure it either didn’t complete, didn’t maintain, or didn’t upgrade. There’s a rich back history of AT&T taking taxpayer money and then failing to deliver upgrades that were promised local municipalities.
Many of folks impacted by AT&T’s decision to hang up on copper are rural or elderly folks who relied on traditional landlines for reliable 911 access but are either outside the range of cellular, or find cellular to be less reliable and significantly more expensive on fixed budgets. The system has a tendency to downplay or ignore these folks.
So you can see how there’s a tension between private telecom monopolies and public interest regulators (the few we still have) tasked with protecting taxpayers and the public interest.
In 20 of the 21 states AT&T operates in, its lobbyists have managed to sell lawmakers on eliminating Carrier of Last Resort (COLR) obligations requiring it provide landline telephone service to any potential customer in its service territory. It’s easy to lobby lawmakers on the idea that the company needs to “move forward past outdated regulations,” and ignore the actual real-world impact or AT&T’s rich history of subsidy fraud or limitations of wireless as a fixed-line alternative.
But they’ve run into trouble in California, after the California Public Utilities Commission (CPUC) told AT&T in 2024 it can’t just hang up on these unwanted (taxpayer subsidized) connections. The CPUC said it’s not blocking AT&T from retiring its aging copper networks, but it wants some AT&T dedication to upgrading failing infrastructure to more modern fiber, not just throw “good enough” wireless at the problem.
Last week AT&T sued California and CPUC (full lawsuit here). AT&T is also asking the Trump FCC to intervene and prevent the CPUC from doing its job. AT&T, for its part, sells this as a story of California leveraging outdated regulations to block AT&T from embracing modernization:
“The federal government and virtually all States where AT&T historically offered POTS [Plain Old Telephone Service] have now eliminated outdated regulatory obstacles, allowing AT&T to begin powering down its POTS network and increasing its investments in modern communication technologies. California stands alone in resisting this progress.”
CPUC counters by saying they don’t want customers who used to have reliable landline service shoveled off to costly and less reliable wireless services instead of fiber. Or left without any connection whatsoever after spending the last four decades slathering AT&T with subsidies.
But it’s worth noting that AT&T’s legal assault is about more than just the fate of dying copper landlines.
California’s CPUC has been filling the void left by Trump regulators and attempting to ensure U.S. broadband is somewhat affordable. That’s involved conditions affixed to grants, affordability conditions applied to recent telecom mergers, and public safety requirements in response to climate-related risks. AT&T, as you might expect, doesn’t like that. Their goal is, with no hyperbole, no oversight at all.
So in addition to this lawsuit, they appear to be leveraging Dem politicians (like Assemblymember Tasha Boerner) in the state to push amendments to the state constitution that would strip the CPUC of its independence, ensuring that AT&T would have more direct lobbying control over the CPUC’s makeup through its robust lobbying control of state legislators.
The changes, which were approved by a California State Assembly vote (67-1), would need to be voted on by California residents later this year. As such, they are being sold to local state folks as a way to keep CPUC focused on soaring electrical utility rates. But the timing of the effort to limit CPUC’s oversight of broadband, just as AT&T tries to deliver the killing blow to the agency, is hard to miss.
Ultimately the broader narrative in the press sold to voters will be that California regulators are engaged in broad over-reach and hampering AT&T’s potential innovation. Downplayed or ignored will be the fact that federal consumer protection has largely been destroyed, and semi-independent regulators like the CPUC in a handful of states are the last line of defense in a country being devoured by corruption.
It’s a lopsided fight that historically telecom monopolies tend to win, which is why, as you can see with your own eyes, most U.S. broadband is patchy, expensive, sluggish, with abysmal customer service. Instead of empowering regulators that protect affordability and competition, we have a nasty tendency to lobotomize them on behalf of “free market competition” that isn’t real, and that monopolies don’t want.
Filed Under: affordability, california, california public utilities commission, cpuc, dsl, fiber, telecom, upgrades
Companies: at&t
US export controls are pushing China’s AI chip industry away from general-purpose GPUs and toward custom ASICs. Huawei leads with 62% projected market share, while Alibaba and Cambricon pursue alternative architectures that may create a structurally different ecosystem from the Nvidia-dominated West.
TL;DR
China’s AI chip industry is no longer trying to build an Nvidia clone. Under sustained US export controls that block access to the most powerful general-purpose GPUs, the country’s largest technology companies are pivoting toward application-specific integrated circuits, custom chips designed to do one thing extremely well rather than handle any workload. The shift is creating a domestic semiconductor ecosystem that may end up architecturally distinct from the Nvidia-dominated model that powers AI in the West.
At the centre of this divergence is a design choice that export controls have accelerated. General-purpose GPUs, the kind Nvidia sells, are flexible and programmable, making them ideal for the fast-moving research phase of AI development where model architectures change constantly. ASICs sacrifice that flexibility for raw efficiency, delivering faster performance at lower power consumption for specific AI tasks. In a market where the best Nvidia hardware is unavailable, the economics of custom silicon become far more compelling.
Chinese companies are pursuing three distinct ASIC architectures. Huawei is betting on neural processing units through its Ascend series, including the widely deployed 910C and the upcoming Ascend 950. Cambricon Technologies is building domain-specific architectures with its Siyuan 590 and 690 series. Alibaba is taking a third route through its semiconductor unit T-Head, which launched the Zhenwu M890 parallel processing unit at its annual cloud computing summit last week, claiming three times the performance of its predecessor.
On the GPU side, Moore Threads leads the domestic effort. Founded in 2020 by Zhang Jianzhong, Nvidia’s former China executive, the company has dedicated itself to general-purpose chips like the MTT S5000 series. Biren Technology, Enflame, and Iluvatar CoreX are also competing in the space, but none has achieved the scale of the ASIC leaders.
A Morgan Stanley report published on 8 May made the market dynamics clear. Huawei is projected to capture 62% of China’s domestic AI accelerator market in 2026, followed by Cambricon at 14%. Among big tech firms building proprietary chips, Baidu and Alibaba are each expected to take roughly 5%. The ASIC heavyweights are winning on volume and momentum.
The performance gap between Chinese chips and Nvidia’s export-compliant hardware has narrowed significantly. Morgan Stanley data shows that Huawei’s Ascend 950 cards and Cambricon’s Siyuan 690 can outperform Nvidia’s H20, the most powerful chip Nvidia is currently permitted to sell to China, by 50 to 150% as measured in tokens per second.
Huawei expects AI chip revenue to reach roughly $12 billion in 2026, up from $7.5 billion in 2025. Nvidia’s share of the Chinese AI accelerator market has effectively collapsed to zero, a development that CEO Jensen Huang has described as a “horrible outcome” for the United States because it breaks the software dependency on Nvidia’s CUDA ecosystem that took two decades to build.
For China’s highly commercialised AI market, which focuses on deploying applications to hundreds of millions of users rather than conducting frontier research, the ASIC approach makes particular sense. Inference, the process of running a trained model at scale, rewards the kind of narrow optimisation that custom silicon provides. Training new models still benefits from GPU flexibility, but the revenue is in deployment.
Hardware performance is only half the equation. The deeper challenge for China’s chip industry is breaking the lock-in created by Nvidia’s CUDA platform, the software layer that millions of AI developers worldwide use to write code for Nvidia hardware. CUDA’s network effects are enormous. Virtually every AI framework, every research paper, and every pre-trained model assumes CUDA compatibility.
Huawei is building CANN as its alternative, while Moore Threads has developed MUSA. DeepSeek has spent months rewriting its core code to work with Huawei’s CANN framework, moving away from the CUDA ecosystem. But semiconductor analyst Zhang Haijun notes that as AI models grow more complex, the boundaries between custom ASICs and flexible GPUs are “becoming increasingly blurry,” suggesting that the winning architecture may eventually combine elements of both.
Omdia chief analyst Su Lian Jye frames the choice practically: enterprises with robust AI engineering capabilities and a clear roadmap benefit from ASICs, while those running mixed workloads still lean toward general-purpose GPUs. For now, market momentum in China favours the specialist approach, partly by choice and partly because the general-purpose option from Nvidia remains either unavailable or restricted.
The long-term consequence of this divergence may be more significant than the near-term performance benchmarks. If China’s AI industry standardises on a mix of Huawei NPUs, Alibaba PPUs, and Cambricon domain-specific chips, each running its own software stack, the result will be a fragmented but domestically self-sufficient ecosystem that operates on fundamentally different architectural assumptions from the West.
That fragmentation carries costs. Developers building for the Chinese market may need to support multiple hardware platforms simultaneously, increasing complexity. Cross-border AI collaboration becomes harder when the underlying compute stacks are incompatible. And the lack of a single dominant platform means no Chinese chip maker benefits from the kind of ecosystem lock-in that made Nvidia’s CUDA so powerful in the first place.
But the direction is set. US export controls intended to slow China’s AI progress have instead accelerated a structural redesign of its chip industry, pushing it toward custom silicon, domestic software stacks, and an architecture that no longer depends on American hardware. Whether that ecosystem can match the pace of innovation in the Nvidia-powered West is the defining question of the AI chip race.
Photo credit: PC World
MSI will release the Claw 8 EX AI+ handheld gaming PC on June 23. The company sets the device at a premium level, with early pricing talk centering on $1,500 for the loaded version that includes 32GB of LPDDR5x-8533 memory and a one-terabyte drive. That figure places it well above many current Windows handhelds and gives it clear ambitions against established names in portable PC gaming.
The new body stretches out the handles wider, giving them the feel of a normal controller. The preview units weigh approximately 785 grams, but they are balanced in such a way that they are less of a handful even when held for extended periods of time. Hall effect sticks and triggers are making a comeback since they are dependable and do not float around like older ones did. A redesigned D-pad and button layout completes the controls, and the small RGB lights around the sticks give a pop of color without causing too much confusion.
Haptics have also been thoroughly redone, as new linear motors replace the old vibration units. They provide crisp, precise feedback with significantly less noise and power draw. Early users have reported that the improvement is noticeable in compatible games, since it adds texture to what you’re doing and eliminates the unpleasant buzzy rumbling that prior versions were prone to.
Next, and most importantly, the Claw 8 EX AI+ has power, thanks to its Intel Arc G3 Extreme processor. It’s a processor that accomplishes the fundamental tasks well and has a graphics component on top of 12 Xe cores. MSI believes it’s fairly efficient and can compete with some of the other mobile CPUs on the market while using significantly less power. Dual fans suck air throughout the entire device, which is approximately 25% better than the last one, and to be honest, it’s quite quiet even when working hard.
When users got their hands on some prototype gear, they were able to see just how well the chip handled recent games at the screen’s native resolution. It also supports Intel’s XeSS 3, which is a fancy term for upscaling and frame generation that helps boost frame rate when needed. You can also plug this device in and turn up the performance if you want, although there’s an endurance mode to keep the power drain low. The eight-inch screen is the same as previously, but with upgraded internals that allow it to display 1920 by 1200 resolution, refresh at 120 times per second, and use VSync to reduce screen tearing. It covers the entire sRGB range and gets bright enough to see at around 500 nits.
An 80-watt-hour battery powers the show, and those who have tried it say it will last a long time, possibly even longer than its predecessor. It charges via one of the two Thunderbolt 4 connectors, which also support external displays and rapid data transfers. Storage is a single M.2 2280 slot, so if you want to upgrade to a larger drive later on, you can simply swap it out without the need for any tools. Memory is fixed at 32GB in the launch model. There’s also a microSD slot and an audio jack, since who doesn’t appreciate a little extra flexibility?
MSI’s Xbox Mode overlay is linked with the operating system, Windows 11. This device allows you to adjust performance parameters, monitor the battery level, and switch between games without leaving the handheld experience. It aims to be like a console, but you can still access the complete desktop if you need to connect a monitor. Other connectivity options include Wi-Fi 7 and Bluetooth, as well as a fingerprint scanner embedded into the power button that eliminates the need to constantly enter in a password.
[Source]
Nvidia has stepped into the processor market with its RTX Spark, but at first glance, it’s clearly behind Apple Silicon by a considerable margin.
Computex 2026 is underway, and Nvidia has formally stepped into the processor ring with its own chip. Nvidia calls the RTX Spark a “superchip” for Windows PCs that have massive AI performance.
This chip consists of an ARM-based Nvidia Grace CPU with 20 cores, as well as an Nvidia Blackwell RTX GPU with 6,144 CUDA cores. There’s also fifth-generation Tensor cores, up to 128GB of unified memory, and a 600GB/s Nvidia NVLink-C2C interconnect providing high-bandwidth communications between the elements.
According to Nvidia, it is “designed for AI, creating, and gaming,” with the intention of it being used to help create slim Windows notebooks with all-day battery life, but massive performance capabilities. This includes rendering massive 90GB 3D scenes for games, generating 4K AI video, and 12K video editing.
More handily for AI researchers, it will also be capable of running a 120 billion-parameter large language model with up to a million tokens of context, using local agents.
“The PC is being reinvented,” said Nvidia CEO Jensen Huang, referring to users predominantly launching apps and manually doing work. Instead, RTX Spark is made to enable “local agents, frontier models, creative workflows, RTX games” on a notebook.
“This is the new PC,” he declared in a press release. “The personal AI computer.”
Undoubtedly, this is a big move for Nvidia, and a major chip introduction that can dramatically affect the Windows notebook market in general. However, Nvidia is still working to catch up to Apple.
Nvidia’s chip shares many of the same core concepts as Apple Silicon, in that it combines a CPU, GPU, neural processing elements, and high-speed unified memory on a single chip. Viewed from a high level, the architectures follow the same approach.
Evidently, when Apple Silicon stunned the world at its launch, it made an impression on the PC industry.
However, despite Nvidia’s bluster about its chip being extremely fast and powerful, it does need to be more directly compared against Apple Silicon to see whether it truly stands up. There’s no real official benchmark result from Nvidia to compare against at this time, but there was a pre-release benchmark that’s noteworthy.
Posted to Geekbench in June 2025 and subsequently removed, but archived by Wccftech, the listing for the Nvidia N1x is believed to be an early version of the GPU maker’s chip.
The version listed includes an ARMv8 chip with 20 cores, a base clock speed of 2.81GHz, and 128GB of unified memory.
Geekbench single-core score comparison between the unreleased N1x and a selection of Apple Silicon scores
When it comes to performance, the single-core score is listed at 3,096 points, with the multi-core score reaching 18,837.
The immediate comparison made by the publication was to the M3 Max chip in a 16-inch MacBook Pro. On checking Geekbench’s listings, the 16-inch MacBook Pro with M3 has a single-core score of 3,128 and a multi-core of 20,969.
For reference, the highest M3 Pro result in a MacBook Pro is 3,105 for the single-core score and 15,255 for the multi-core.
Under the current M5 generation, the 14-inch MacBook Pro with M5 gets a massive 4,224 for the single-core score and 17,465 for the multi-core.
Geekbench multi-core scores compared between N1x results and various Apple Silicon versions.
This doesn’t seem massively impressive, until you check the core counts of Apple’s chips. The M3 Max in question has 16 cores, the M3 Pro has 12 cores, and even the M5 result involved just 10 cores.
The current Apple Silicon leader, the 18-core M5 Max, is seen setting scores at around 4,200 again for the single-core, but the multi-core is hovering within touching distance of 30,000.
Admittedly the N1x result is a pre-release listing and a year old. It’s entirely possible that Nvidia has updated the design, increased the clock speed, and made other changes since that time.
However, with modern manufacturing lead times being extremely long, there probably hasn’t been much change since then.
The main takeaway here is that Nvidia has seen Apple Silicon as a threat, and believes it can do something better for the Windows market.
It probably could. Eventually.
The Computex launch of Nvidia’s RTX Spark
If the N1x result actually reflects the capabilities of the first chip, it’s a good start for the company’s initial release. But that said, it’s up against some considerable competition.
As much as Nvidia boasts about the AI capabilities of RTX Spark, Apple’s already got a counter for it. Aside from the Neural Engine, the M5 generation has neural accelerators in each GPU core, making it massively more capable of AI tasks.
As it stands, on the CPU front, it is trailing behind a chip from Apple that’s more than two years old. Nvidia has some considerable catching up to do.
The Centre for Cybersecurity Belgium (CCB), the country’s national authority for cybersecurity, warned on Friday that threat actors are now exploiting a recently patched critical Windows Netlogon vulnerability in attacks.
Netlogon is a remote procedure call (RPC) interface and a core Microsoft Windows Server background service that authenticates services and users on Windows domain-based networks.
Microsoft patched this vulnerability (CVE-2026-41089) during the May 2026 Patch Tuesday, describing it as a stack-based buffer overflow in Windows Netlogon that allows attackers without privileges to gain remote code execution on targeted domain controllers.
“An attacker could send a specially crafted network request to a Windows server that is acting as a domain controller,” it said. “If successful, this could cause the Netlogon service to improperly handle the request, potentially allowing the attacker to run code on the affected system without needing to sign in or have prior access.”
CVE-2026-41089 impacts all currently supported Windows Server versions, including the latest release, Windows Server 2025.
According to a security advisory published by the company on May 12, the vulnerability was discovered by Windows Attack Research & Protection (WARP), an internal offensive cybersecurity and engineering research team at Microsoft.
On Friday, Belgium’s national cybersecurity authority (CCB) warned that attackers are now actively exploiting the CVE-2026-41089 security flaw in the wild and urged admins to immediately patch vulnerable servers.
“CVE-2026-41089 in #Windows #Netlogon is now actively #exploited in the wild and could lead to #RCE. CVSS(3.1): 9.8,” the CBC warned in a Friday tweet. “Patch as quickly as possible.”
However, the CCB didn’t provide further details on these ongoing attacks and didn’t respond to a BleepingComputer request for more information.
Microsoft has yet to update its advisory, and a company spokesperson didn’t reply to an email from BleepingComputer requesting confirmation that CVE-2026-41089 is now actively exploited.
Two weeks ago, Microsoft shared mitigation measures for YellowKey (CVE-2026-45585), a Windows BitLocker zero-day vulnerability that grants access to protected drives, described as a backdoor by anonymous security researcher ‘Nightmare Eclipse,’ who also disclosed it and published a proof-of-concept (PoC) exploit.
Over the past several months, Nightmare Eclipse also disclosed the BlueHammer (CVE-2026-33825) and RedSun (CVE-2026-41091) privilege escalation zero-day flaws (both now being exploited in attacks), the GreenPlasma and MiniPlasma zero-day privilege escalation flaws that provide SYSTEM privileges, and UnDefend (CVE-2026-45498), another zero-day that attackers with standard user permissions can exploit to block Microsoft Defender definition updates.
Initially, Microsoft has reacted to Nightmare Eclipse with thinly veiled threats of legal action, followed by a tweet saying that the company “will work with law enforcement as appropriate” when “an individual breaks the law and engages in malicious activity causing real harm to our customers.”
Automated pentesting tools deliver real value, but they were built to answer one question: can an attacker move through the network? They were not built to test whether your controls block threats, your detection rules fire, or your cloud configs hold.
This guide covers the 6 surfaces you actually need to validate.
During his keynote at GTC Taipei, NVIDIA chief Jensen Huang laid out a clear shift. People will soon treat their computers less like tools that wait for commands and more like capable partners that can take on real work when asked. RTX Spark sits at the center of that change. The new processor marks NVIDIA’s first complete silicon solution built specifically for Windows PCs. It combines a powerful graphics engine with an efficient central processor in one tightly linked package. This setup brings the full range of NVIDIA’s software tools directly to laptops and small desktops without requiring a separate graphics card.
On the graphics side, the chip is based on the Blackwell architecture, with 6,144 CUDA cores that use fifth-generation FP4 precision to conduct AI calculations with ease. A 20-core proprietary Grace processor, co-developed with MediaTek, handles general computing tasks and connects to the graphics side via a super-fast chip-to-chip interface. The design shares memory between the two, allowing configurations of up to 128GB of unified memory. NVIDIA claims that this arrangement will increase AI performance beyond one petaflop.
That tech combination enables some real-world advantages. For example, systems can run massive language models with 120 billion parameters and context windows containing up to a million tokens on the device from the start. Users can also work with personal AI agents that run locally, keeping their data private and allowing them to navigate between apps without constantly sending queries to the cloud. Creative tasks benefit as well, as video editing now supports 12K resolution with full decoder support, 3D scenes render more smoothly, and Adobe and other toolsets experience significant improvements in their AI-powered capabilities.
Gaming performance improves significantly, as laptops can now achieve high frame rates at 1440p while utilizing ray tracing and frame generation technology familiar from NVIDIA’s other platforms. Furthermore, this hardware is designed to handle new rendering approaches coming to games and creative apps. NVIDIA collaborated extensively with Microsoft to make Windows compatible with this architecture, which includes optimizations for workload scheduling, power management, and increased support for neural rendering in DirectX. A new runtime dubbed NVIDIA OpenShell gives consumers and developers more control over how their AI agents act, including rules for task routing and data protection. To top it all off, new Windows security capabilities have been added to keep agent activity under control while maintaining user oversight.
The RTX Spark laptops are designed to be exceedingly thin, with some measuring only 14mm thick and weighing 3 pounds, but still providing all-day battery life. Screen sizes range from 14 to 16 inches, with aluminum enclosures and high-end screens that use sync technology. Desktops will also be available in compact sizes for individuals who prefer to use a stationary system while maintaining the same capabilities.
Multiple manufacturers, including ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI, intend to debut products this autumn, with Acer and GIGABYTE following closely behind. HP, for example, will offer the super-slim OmniBook Ultra 16 and OmniBook X 14. The first batch of systems are focused at the high end of the market, but NVIDIA and its partners anticipate that lower-end configurations with less memory will become more affordable in the future.
“The company best known for powering the AI boom is coming for the PC,” reports Axios.
Nvidia’s CEO unveiled a new ARM-based “N1X processor made alongside Microsoft,” reports CNBC, that “will be incorporated into a new RTX Spark superchip, debuting in the fall on a fresh line of Windows PCs from Microsoft, Dell, HP, ASUS, Lenovo and MSI.”
It was only a matter of time before NVIDIA released a powerful system-on-a-chip (SOC) to take on AMD’s Ryzen AI Max and Qualcomm’s latest Snapdragon X2 chips. At Computex today, NVIDIA unveiled the RTX Spark, a “superchip” meant to give both laptops and small desktops fast AI and graphics performance…
The company says it offers 1 petaflop of AI computing power, and that it has 6,144 Blackwell RTX cores and 20 Mediatek Arm CPU cores. NVIDIA claims it’s similar to the RTX 5070 laptop GPU but with much lower power draw. RTX Spark also has an NPU that’s fast enough to be part of Microsoft’s Copilot+ initiative, which requires a 40 TOPS NPU, but NVIDIA says it’s mainly touting the tensor cores as part of the chip’s Blackwell GPU for AI performance. RTX Spark’s GPU can directly draw on the chip’s large pool of unified memory, which can span from 16GB to 128GB, and the chip itself can use anywhere from single-digit wattage up to 80W…
NVIDIA CEO Jensen Huang positions RTX Spark as a complete reinvention of the PC, eventually turning them more into devices meant for AI agents than manual human input… NVIDIA has been working together with Microsoft for “several years” while designing the RTX Spark, according to NVIDIA representatives… In a blog post provided to media, Microsoft head of Windows and devices, Pavan Davuluri, noted that the company optimized Windows 11’s workload profile scheduling for the RTX Spark. “Whether you’re checking your email or running an agent locally to debug code, the Windows scheduler on RTX Spark will ensure you get the best performance and efficiency out of your CPU,” he wrote.
Flight 236 left Newark Liberty International Airport on Saturday night en route to Palma de Mallorca but reversed course about an hour after takeoff.
Read Entire Article
Source link
Filling up your gas tank stings right now. I go to Costco specifically to save money on gas, and even there, I’m spending at least $15 more than usual to fill up my Volkswagen Jetta. If you drive a truck or an SUV, you already know a trip to the pump can run you well over $100.
Summer is right around the corner, which usually means lake days, beach runs, road trips, and generally getting out more often. That also means using gas to get there. With gas averaging over $4.50 per gallon nationally, every little trip adds up faster than it used to.
That got me thinking about the trips that probably don’t need a car at all, like quick grocery runs, going to a coffee shop down the road or going to the gym for a daily workout. For those kinds of trips, an electric scooter starts to make a lot of sense.
While an electric scooter won’t take you on a multi-hour summer road trip, it’s a solid alternative to those short drives we make without thinking about them. So I tested out the Gyroor C1 Pro and dug into the data to find out how much money it can actually save you on those short, everyday trips.
Before buying an electric scooter, it helps to know if it will actually save you money or if it just feels like it does. So, let’s run the numbers.
AAA puts the national average for regular unleaded gas at $4.52 per gallon. If your car gets 25 miles per gallon, which is typical for a sedan, you’re paying about 18 cents per mile.
The Gyroor C1 Pro gets you up to 25 miles on a single charge. You plug it into an outlet, like you would a laptop, and you can get back to a full battery in about 5 hours. The scooter runs on a 36-volt 10.4Ah battery with a total capacity of 374 watt-hours. Watt-hours measure how much energy a battery can store. The higher the number of watts, the farther you can go on one charge.
The national average electricity rate is 17.65 cents per kilowatt-hour, according to the Energy Information Administration. At this rate, a full charge costs about 66 cents. That works out to $0.003 per mile, which is 60 times cheaper per mile than driving a gas car.
Here’s the formula we used to get there:
Cost Per Mile = (Battery Capacity in Wh ➗1,000) X Electricity Rate per kWh / Range in miles
$0.003 per mile = (374 ➗ 1,000) X $0.1765 ➗25
If you swap out one short car trip per day at about 3 miles each way, that’s 168 miles a month that you no longer need gas for. In a realistic sense, this looks like your coffee run, a trip to the gym or another errand you’ve been driving to out of habit. Here’s how the cost compares for using the Gyroor C1 Pro electric scooter versus a gas car.
Comparing the costs of a gas car versus an e-scooter.
Costs of a gas car vs. an e-scooter
Vehicle type
Cost per mile
Cost to travel 10 miles
Average monthly cost (168 miles)
Gas car (25 MPG @ $4.52/gal)
$0.18
$1.81
$30.24
Gyroor C1 Pro
$0.00
$0.03
$0.50
Savings/Difference
$0.18
$1.78
$29.74
*AAA
**US Energy Information Administration
For the same 18 cents it costs to drive one mile in a gas car, you could ride the Gyroor C1 Pro about 60 miles. Replacing just one daily short trip with an electric scooter saves you nearly $30 per month. Make it a habit and swap a second trip, and you’re looking at closer to $60.
If you drive a truck or large SUV that gets lower gas mileage than a typical sedan, your savings will be even higher. Either way, charging an electric scooter costs significantly less than filling up at the gas pump.
Hidden savings add to that total. Car ownership comes with several costs that add up over the year, such as registration fees, oil changes and insurance. As someone who hates paying for parking, being able to lock up to a bike rack for free makes my day every time.
At $30 in monthly savings, the Gyroor C1 Pro pays for itself in about 15 months at its current sale price of $460 (regularly $600) before you add in savings from parking and other car-related fees. A well-maintained electric scooter can last three to five years, meaning years of savings after the scooter has paid for itself. If you’re filling up weekly, you’ll burn through that same $459 in less than two months at current gas prices.
The Gyroor C1 Pro’s sibling scooter, the C1S, made CNET’s best electric scooter list as the top pick for backpack-free errands. The C1 Pro takes everything CNET loves about the C1S and adds more range, power and a higher weight capacity (at a slightly higher price).
I had never ridden a seated scooter before, so the motion took a ride or two to click. But the whole experience of riding a scooter is so much fun, it makes my inner child come out in full force. The seat is comfortable, so you can actually enjoy the ride instead of wanting to get it over with. It comes in pink or green, so you can show a bit of personality with your ride.
The basket storage is a major part of what makes it great for everyday use. On a standing scooter, you’re stuck with a backpack and whatever fits inside it. The C1 Pro’s storage is spacious enough to carry groceries and a work bag without putting any of the stress on your shoulders and back.
The LED display shows your speed and battery level, so it’s easy to stay aware of your pace. There is also a headlight built in for nighttime rides, which is super helpful if you’re in an area with fewer street lights. The dual suspension keeps the ride smooth and easy to stay steady on two wheels. When I pushed the scooter to its highest speed (18.6 mph), it felt quick but not scary.
The 25-mile range comfortably covers most of my daily use cases. Plug it in overnight and you can start the morning with a full charge.
The C1 Pro’s display panel shows speed and battery life.
Where you live plays a big role in how much you get out of this scooter. If you’re in a city with mild weather, bike lanes and greenways, the C1 Pro fits into daily life pretty seamlessly. In areas with frequent heavy rains or harsh winters, the riding season will be shorter and slow the payback period. However, the battery compartment is sealed, and the IPX4 rating handles a light drizzle, but I wouldn’t recommend riding it through a storm.
Bike lanes and existing scooter culture make a big difference. In cities where sharing the road with scooters and bicycles is already normal, the transition is smoother. Areas without that infrastructure take more adjustment as you’re learning to ride as an exposed person next to traffic. I recommend finding the routes that fit you, including backroads, greenways and less hectic areas if possible. It’s not always the fastest route, but it can be a lot more enjoyable.
At 48 pounds, it’s not the lightest scooter on the market, but that’s due to the trade-off of having a seat, dual suspension and basket storage. It folds down easily enough to load into a trunk or to store away when you’re not riding. I wish it could fold small enough to fit in a suitcase and travel with me everywhere.
One other thing worth noting is that this scooter is best for a single rider. You could fit a small or medium-size dog in the basket, but it’s not the best scooter for shuttling kids around or running family errands together.
Off for another errand.
If your daily routine involves a handful of short trips you could easily make without a car, the Gyroor C1 Pro electric scooter will quickly pay for itself and then some. It’s a great fit for your office commute, or for a quick ride to a coffee shop, gym, grocery store or getting around a college campus. It allows you to skip the traffic, eliminates parking fees and costs a fraction of what a gas car does per mile.
It’s not the right fit for families moving more than one person or for those in a climate that would limit you to riding for only a few months of the year.
Replacing just one car trip a day with the Gyroor C1 Pro can save you around $30 a month, and you’ll have fun while doing it. If you go with the pink one, you’ll be twinning with me. At its current sale price of $460, it’s a fraction of what you could spend on gas this summer alone.
AI companies have grown into data-hungry entities as their models require ever-larger datasets to train on. To meet that need, many AI startups defy long-standing internet conventions — like respecting robots.txt files, which signal to automated crawlers which parts of a website are off-limits — and scrape data aggressively. This has forced websites to restrict access to their data and, in some cases, strike licensing deals with AI companies. Fitness and social running company Strava is making a move in this direction by restricting its website and introducing fees for developer access.
To stop scraping, the company is increasing security around its website and will now only allow authenticated users to view certain data. Earlier, users were able to see details like public profiles and fitness club listings without logging in. The company is putting all that data behind authentication to protect it from unauthorized AI scraping.
On the API front, developers could previously start building apps on Strava through a free, tiered access program — applying for basic access first, then requesting more as their app grew. Now the company is adding a flat $11.99 per month fee for all developers, though it noted the price may vary by geography.
Strava said its developer community has grown from 185,000 members last year to 241,000 this year, and the company plans to continue supporting them. As part of that, Strava also plans to add support for Model Context Protocol (MCP), an emerging standard that lets AI assistants and apps access external data in a structured way, giving Strava more control over exactly what gets shared and how.
The company is also planning to retire some API endpoints — discrete access points that let outside apps pull specific data, like club details — to protect user data. Strava had already tightened API rules in 2024, banning its use for AI training and limiting third-party apps from displaying other users’ data. Those changes drew backlash from developers who said their apps would be severely affected.
While some developers may accept paying a subscription fee, sunsetting certain API endpoints could still impact dependent apps. Strava is giving developers a 90-day grace period before making these changes.
In an interview with TechCrunch, Michael Martin, Strava’s CEO, said unchecked AI scraping could be the death knell of the public internet.
“AI companies are ruthlessly scraping public websites, given their endless need for training data, which is degrading site performance across the board,” Martin said. We’ve had multiple instances in the last several months where performance has been diminished and, in some cases, impaired. Beyond scraping the public sites, they’re also trying to use our API to get access to our data, ignoring API terms.”
He noted that Strava has refused overtures from leading AI labs seeking data licensing deals. He specifically singled out Perplexity, saying the AI search startup routed its scraping through aggregator services to obscure its origin despite being turned away. This is consistent with Perplexity having been accused of similar behavior elsewhere in the past.
Martin also flagged server overload caused by poorly built vibe-coded apps, whose API calls are often inefficiently structured and generate a disproportionate load on Strava’s systems. It’s a pattern: when Meta banned third-party chatbots from WhatsApp last year, it made a similar argument about system overhead.
The timing probably isn’t coincidental. Strava confidentially filed for an IPO earlier this year, and its move to protect its data may be intended to signal data discipline to prospective investors. The comparison to Reddit’s 2024 crackdown on API access is one Martin was quick to address. Unlike Reddit, which priced API access by the number of calls (making it unaffordable for many app developers), Strava is betting a flat fee keeps the developer ecosystem intact.
“We want the users to feel that they own their data and feel comfortable with how we are controlling and securing it. But we want the developers to continue to flourish and grow,” Martin said.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Israel says it has killed new Hamas military leader in Gaza City airstrikes
NASA taps Blue Origin to deliver lunar rovers for Moon Base initiative
2026 NBA Finals schedule, odds: Knicks await Thunder or Spurs after winning East
XRP *JUST* SUCCEEDED!!!! CLARITY ACT EXPOSED!!! (SHE EXPOSED IT)
This is BROKEN! INSANE 5x MONEY CAR WASH WEEK! The NEW GTA Online UPDATE Today! (GTA5 New Update)
Micron Crosses $1 Trillion Market Cap as AI Demand Reshapes Memory Sector
Hottest May day ever as London hits 34.8C in 2C leap from previous records
Selena Gomez Reportedly Upset Over Benny Blanco’s Comments on Her ‘Terrible’ Diet
China assigns ID codes to 28,000+ humanoid robots
Nikkei 225 Surges Past 65,000 for First Time as Iran Peace Hopes Fuel Record Rally
Waymo dominates autonomous vehicle registrations as Tesla trails behind
‘Breaking Bad’ Star’s Easy-to-Binge 6-Part Crime Series Spin-Off Is Finally Heading to Free Streaming
Crowds find riverside shade in York as temperatures soar
The Samsung pay deal is the moment Korean unions changed register
SpaceX just won a second Golden Dome contract. This one is $4.16 billion.
Days of our Lives 2-Week Spoilers May 25-June 5: Gwen Rages, Abe Confesses & 2 Tragic Anniversaries!
Taylor Swift Fans Label Travis Kelce’s Beer-Chugging A ‘Red Flag
FIRST NIGHT REVIEW: Take That bring the Circus back to life in spectacular sun-soaked style
Millions of AI agents imperiled by critical vulnerability in open source package
SHE IS KILLING XRP!!! WATCH URGENT AND ACT FAST
You must be logged in to post a comment Login