NVIDIA

Google CEO: If an AI bubble pops, no one is getting out clean

AI, AI energy, AI infrastructure, AI investment, AI regulation, alphabet, Biz & IT, climate change, deepmind, dotcom bubble, Ed Zitron, Google, Jamie Dimon, Jensen Huang, large language models, machine learning, NVIDIA, openai, Sundar Pichai / 9u50fv / November 18, 2025

Market concerns and Google’s position

Alphabet’s recent market performance has been driven by investor confidence in the company’s ability to compete with OpenAI’s ChatGPT, as well as its development of specialized chips for AI that can compete with Nvidia’s. Nvidia recently reached a world-first $5 trillion valuation due to making GPUs that can accelerate the matrix math at the heart of AI computations.

Despite acknowledging that no company would be immune to a potential AI bubble burst, Pichai argued that Google’s unique position gives it an advantage. He told the BBC that the company owns what he called a “full stack” of technologies, from chips to YouTube data to models and frontier science research. This integrated approach, he suggested, would help the company weather any market turbulence better than competitors.

Pichai also told the BBC that people should not “blindly trust” everything AI tools output. The company currently faces repeated accuracy concerns about some of its AI models. Pichai said that while AI tools are helpful “if you want to creatively write something,” people “have to learn to use these tools for what they’re good at and not blindly trust everything they say.”

In the BBC interview, the Google boss also addressed the “immense” energy needs of AI, acknowledging that the intensive energy requirements of expanding AI ventures have caused slippage on Alphabet’s climate targets. However, Pichai insisted that the company still wants to achieve net zero by 2030 through investments in new energy technologies. “The rate at which we were hoping to make progress will be impacted,” Pichai said, warning that constraining an economy based on energy “will have consequences.”

Even with the warnings about a potential AI bubble, Pichai did not miss his chance to promote the technology, albeit with a hint of danger regarding its widespread impact. Pichai described AI as “the most profound technology” humankind has worked on.

“We will have to work through societal disruptions,” he said, adding that the technology would “create new opportunities” and “evolve and transition certain jobs.” He said people who adapt to AI tools “will do better” in their professions, whatever field they work in.

Google CEO: If an AI bubble pops, no one is getting out clean Read More »

Review: New Framework Laptop 16 takes a fresh stab at the upgradeable laptop GPU

AMD, AMD Ryzen, Features, framework, framework laptop 13, framework laptop 16, gaming, NVIDIA, nvidia geforce, Reviews, right to repair, Tech / 9u50fv / November 12, 2025

framework laptop 16, take two

New components make it more useful and powerful but no less odd.

Credit: Andrew Cunningham

The original Framework Laptop 16 was trying to crack a problem that laptop makers have wrestled with on and off for years: Can you deliver a reasonably powerful, portable workstation and gaming laptop that supports graphics card upgrades just like a desktop PC?

Specs at a glance: Framework Laptop 16 (2025)
OS	Windows 11 25H2
CPU	AMD Ryzen AI 7 350 (4 Zen 5 cores, 4 Zen 5c cores)
RAM	32GB DDR5-5600 (upgradeable)
GPU	AMD Radeon 860M (integrated)/Nvidia GeForce RTX 5070 Mobile (dedicated)
SSD	1TB Western Digital Black SN770
Battery	85 WHr
Display	16-inch 2560×1600 165 Hz matte non-touchscreen
Connectivity	6x recessed USB-C ports (2x USB 4, 4x USB 3.2) with customizable “Expansion Card” dongles
Weight	4.63 pounds (2.1 kg) without GPU, 5.29 pounds (2.4 kg) with GPU
Price as tested	Roughly $2,649 for pre-built edition; $2,517 for DIY edition with no OS

Even in these days of mostly incremental, not-too-exciting GPU upgrades, the graphics card in a gaming PC or graphics-centric workstation will still feel its age faster than your CPU will. And the chance to upgrade that one component for hundreds of dollars instead of spending thousands replacing the entire machine is an appealing proposition.

Upgradeable, swappable GPUs would also make your laptop more flexible—you can pick and choose from various GPUs from multiple vendors based on what you want and need, whether that’s raw performance, power efficiency, Linux support, or CUDA capabilities.

Framework’s first upgrade to the Laptop 16—the company’s first upgrade to any of its products aside from the original Laptop 13—gets us pretty close to that reality. The laptop can now support two interchangeable motherboards: one with an older AMD Ryzen 7040-series CPU and one with a new Ryzen AI 300-series CPU. And both motherboards can be used either with just an integrated GPU or with dedicated GPUs from both AMD and Nvidia.

The Nvidia GeForce 5070 graphics module is the most exciting and significant part of this batch of updates, but there are plenty of other updates and revisions to the laptop’s external and internal components, too. These upgrades don’t address all of our problems with the initial version of the laptop, but they do help quite a bit. And a steady flow of updates like these would definitely make the Laptop 16 a platform worth investing in.

Re-meet the Framework Laptop 16

Framework’s Laptop 13 stacked on top of the 16. Credit: Andrew Cunningham

Framework treats each of its laptops as a platform to be modified and built upon rather than something to be wholly redesigned and replaced every time it’s updated. So these reviews necessarily re-cover ground we have already covered—I’ve also reused some of the photos from last time, since this is quite literally the same laptop in most respects. I’ll point you to the earlier review for detailed notes on the build process and how the laptop is put together.

To summarize our high-level notes about the look, feel, and design of the Framework Laptop 16: While the Framework Laptop 13 can plausibly claim to be in the same size and weight class as portables like the 13-inch MacBook Air, the Framework Laptop 16 is generally larger and heavier than the likes of the 16-inch MacBook Pro or portable PC workstations like the Lenovo ThinkPad P1 or Dell 16 Premium. That’s doubly true once you actually add a dedicated graphics module to the Laptop 16—these protrude a couple of inches from the back of the laptop and add around two-thirds of a pound to its weight.

	Frame-work 16 (no GPU)	Frame-work 16 (GPU)	Apple 16-inch MBP	Dell 16 Premium	Lenovo ThinkPad P1 Gen 8	HP ZBook X G1i	Lenovo Legion Pro 5i Gen 10	Razer Blade 16
Size (H x W x D inches)	0.71 x 14.04 x 10.63	0.82 x 14.04 x 11.43	0.66 x 14.01 x 9.77	0.75 x 14.1 x 9.4	0.39-0.62 x 13.95 x 9.49	0.9 x 14.02 x 9.88	0.85-1.01 x 14.34 x 10.55	0.59-0.69 x 13.98 x 9.86
Weight	4.63 lbs	5.29 lbs	4.7-4.8 lbs	4.65 pounds	4.06 lbs	4.5 lbs	5.56 lbs	4.71 lbs

You certainly can find laptops from the major PC OEMs that come close to or even exceed the size and weight of the Laptop 16. But in most cases, you’ll find that comparably specced and priced laptops are an inch or two less deep and at least half a pound lighter than the Laptop 16 with a dedicated GPU installed.

But if you’re buying from Framework, you’re probably at least notionally interested in customizing, upgrading, and repairing your laptop over time, all things that Framework continues to do better than any other company.

The Laptop 16’s customizable keyboard deck is still probably its coolest feature—it’s a magnetically attached series of panels that allows you to remove and replace components without worrying about the delicate and finicky ribbon cables the Laptop 13 uses. Practically, the most important aspect of this customizable keyboard area is that it lets you decide whether you want to install a dedicated number pad or not; this also allows you to choose whether you want the trackpad to be aligned with the center of the laptop or with wherever the middle of the keyboard is.

It might look a little rough, but the customizable keyboard deck is still probably the coolest thing about the Laptop 16 in day-to-day use. Andrew Cunningham

But Framework also sells an assortment of other functional and cosmetic panels and spacers to let users customize the laptop to their liking. The coolest, oddest accessories are still probably the LED matrix spacers and the clear, legend-less keyboard and number pad modules. We still think this assortment of panels gives the system a vaguely unfinished look, but Framework is clearly going for function over form here.

The Laptop 16 also continues to use Framework’s customizable, swappable Expansion Card modules. In theory, these let you pick the number and type of ports your laptop has, as well as customize your port setup on the fly based on what you need. But as with all AMD Ryzen-based Framework Laptops, there are some limits to what each port can do.

According to Framework’s support page, there’s no single Expansion Card slot that is truly universal:

Ports 1 and 4 support full 40Gbps USB 4 transfer speeds, display outputs, and up to 240 W charging, but if you use a USB-A Expansion Card in those slots, you’ll increase power use and reduce battery life.
Ports 2 and 4 support display outputs, up to 240 W charging, and lower power usage for USB-A ports, but they top out at 10Gbps USB 3.2 transfer speeds. Additionally, port 5 (the middle port on the right side of the laptop, if you’re looking at it head-on) supports the DisplayPort 1.4 standard where the others support DisplayPort 2.1.
Ports 3 and 4 are limited to 10Gbps USB 3.2 transfer speeds and don’t support display outputs or charging.

The Laptop 16 also doesn’t include a dedicated headphone jack, so users will need to burn one of their Expansion Card slots to get one.

Practically speaking, most users will be able to come up with a port arrangement that fits their needs, and it’s still handy to be able to add and remove things like Ethernet ports, HDMI ports, or SD card readers on an as-needed basis. But choosing the right Expansion Card slot for the job will still require some forethought, and customizable ports aren’t as much of a selling point for a 16-inch laptop as they are for a 13-inch laptop (the Framework Laptop 13 was partly a response to laptops like the MacBook Air and Dell XPS 13 that only came with a small number of USB-C ports; larger laptops have mostly kept their larger number and variety of ports).

What’s new in 2025’s Framework Laptop 16?

An upgraded motherboard and a new graphics module form the heart of this year’s Laptop 16 upgrade. The motherboard steps up from AMD Ryzen 7040-series processors to AMD Ryzen AI 7 350 and Ryzen AI 9 HX 370 chips. These are the same processors Framework put into the Laptop 13 earlier this year, though they ought to be able to run a bit faster in the Laptop 16 due to its larger heatsink and dual-fan cooling system.

Along with an upgrade from Zen 4-based CPU cores to Zen 5 cores, the Ryzen AI series includes an upgraded neural processing unit (NPU) that is fast enough to earn Microsoft’s Copilot+ PC label. These PCs have access to a handful of unique Windows 11 AI and machine-learning features (yes, Recall, but not just Recall) that are processed locally rather than in the cloud. If you don’t care about these features, you can mostly just ignore them, but if you do care, this is the first version of the Laptop 16 to support them.

Most of the new motherboard’s other specs and features are pretty similar to the first-generation version; there are two SO-DIMM slots for up to 96GB of DDR5-5600, one M.2 2280 slot for the system’s main SSD, and one M.2 2230 slot for a secondary SSD. Wi-Fi 7 and Bluetooth connectivity are provided by an AMD RZ717 Wi-Fi card that can at least theoretically also be replaced with something faster down the line if you want.

The more exciting upgrade, however, may be the GeForce RTX 5070 GPU. This is the first time Framework has offered an Nvidia product—its other GPUs have all come from either Intel or AMD—and it gives the new Laptop 16 access to Nvidia technologies like DLSS and CUDA, as well as much-improved performance for games with ray-traced lighting effects.

Those hoping for truly high-end graphics options for the Laptop 16 will need to keep waiting, though. The laptop version of the RTX 5070 is actually the same chip as the desktop version of the RTX 5060, a $300 graphics card with 8GB of RAM. As much as it adds to the Laptop 16, it still won’t let you come anywhere near 4K in most modern games, and for some, it may even struggle to take full advantage of the internal 165 Hz 1600p screen. Professional workloads (including AI workloads) that require more graphics RAM will also find the mobile 5070 lacking.

Old 180 W charger on top, new 240 W charger on bottom. Credit: Andrew Cunningham

Other components have gotten small updates as well. For those who upgrade an existing Laptop 16 with the new motherboard, Framework is selling 2nd-generation keyboard and number pad components. But their main update over the originals is new firmware that “includes a fix to prevent the system from waking while carried in a bag.” Owners of the original keyboard can install a firmware update to get the same functionality (and make their input modules compatible with the new board).

Upgraders should also note that the original system’s 180 W power adapter has been replaced with a 240 W model, the maximum amount of power that current USB-C and USB-PD standards are capable of delivering. You can charge the laptop with just about any USB-C power brick, but anything lower than 240 W risks reducing performance (or having the battery drain faster than it can charge).

Finally, the laptop uses a second-generation 16-inch, 2560×1600, 165 Hz LCD screen. It’s essentially identical in every way to the first-generation screen, but it formally supports G-Sync, Nvidia’s adaptive sync implementation. The original screen can still be used with the new motherboard, but it only supports AMD’s FreeSync, and Framework told us a few months ago that the panel supplier had no experience providing consumer-facing firmware updates that might add G-Sync to the old display. It’s probably not worth replacing the entire screen for, but it’s worth noting whether you’re upgrading the laptop or buying a new one.

Performance

Framework sent us the lower-end Ryzen AI 7 350 processor configuration for our new board, making it difficult to do straightforward apples-to-apples comparisons to the high-end Ryzen 9 7940HS in our first-generation Framework board. We did test the new chip, and you’ll see its results in our charts.

We’ve also provided numbers from the Ryzen AI 9 HX 370 in the Asus Zenbook S16 UM5606W to show approximately where you can expect the high-end Framework Laptop 16 configuration to land (Framework’s integrated graphics performance will be marginally worse since it’s using slower socketed RAM rather than LPDDR5X; other numbers may differ based on how each manufacturer has configured the chip’s power usage and thermal behavior). We’ve also included numbers from the same chip in the Framework Laptop 13, though Framework’s spec sheets indicate that the chips have different power limits and thus will perform differently.

We were able to test the new GeForce GPU in multiple configurations—both paired with the new Ryzen AI 7 350 processor and with the old Ryzen 9 7940HS chip. This should give anyone who bought the original Laptop 16 an idea of what kind of performance increase they can expect from the new GPU alone. In all, we’ve tested or re-tested:

The Ryzen 7 7940HS CPU from the first-generation Laptop 16 and its integrated Radeon 780M GPU
The Ryzen 7 7940HS and the original Radeon RX 7700S GPU module
The Ryzen 7 7940HS and the new GeForce RTX 5070 GPU module, for upgraders who only want to grab the new GPU
The Ryzen AI 7 350 CPU and the GeForce RTX 5070 GPU

We also did some light testing on the Radeon 860M integrated GPU included with the Ryzen AI 7 350.

All the Laptop 16 performance tests were run with Windows’ Best Performance power preset enabled, which will slightly boost performance at the expense of power efficiency.

Given all of those hardware combinations, we simply ran out of time to test the new motherboard with the old Radeon RX 7700S GPU—Framework is continuing to sell it, so it is a realistic combination of components. But our RTX 5070 testing suggests that these GPUs will perform pretty much the same regardless of which CPU you pair them with.

If you’re buying the cheaper Laptop 16 with the Ryzen AI 7 350, the good news is that it generally performs at least as well as—and usually a bit better than—the high-end Ryzen 9 7940HS from the last-generation model. Performance is also pretty similar to the Ryzen AI 9 HX 370 in smaller, thinner laptops—the extra power and cooling capacity in the Laptop 16 is paying off here. People choosing between a PC and a Mac should note that none of these Ryzen chips come anywhere near the M4 Pro used in comparably priced 16-inch MacBook Pros, but that’s just where the PC ecosystem is these days.

How big an upgrade the GeForce 5070 will be depends on the game you’re playing. In titles like Borderlands 3 that naturally run a bit better on AMD’s GPUs, there’s not much of a difference at all. In games like Cyberpunk 2077 with heavy ray-tracing effects enabled, the mobile RTX 5070 can be nearly twice as fast as the RX 7700S.

Most games will fall somewhere in between those two extremes; our tests show that the improvements hover between 20 and 30 percent most of the time, just a shade less than the 30 to 40 percent improvement that Framework claimed in its original announcement.

Beyond raw performance, the other thing you get with an Nvidia GPU is access to a bunch of important proprietary technologies like DLSS upscaling and CUDA—these technologies are often better and more widely supported than the equivalent technologies that AMD’s or Intel’s GPUs use, thanks in part to Nvidia’s overall dominance of the dedicated GPU market.

In the tests we’ve run on them, the Radeon 860M and 890M are both respectable integrated GPUs (the lower-end 860M typically falls just short of last generation’s top-end 780M, but it’s very close). They’re never able to provide more than a fraction of the Radeon RX 7700S’s performance, let alone the RTX 5070, but they’ll handle a lot of lighter games at 1080p. I would not buy a system this large or heavy just to use it with an integrated GPU.

Better to be unique than perfect

It’s expensive and quirky, but the Framework Laptop 16 is worth considering because it’s so different from what most other laptop makers are doing. Credit: Andrew Cunningham

Our original Framework Laptop 16 review called it “fascinating but flawed,” and the parts that made it flawed haven’t really changed much over the last two years. It’s still relatively large and heavy; the Expansion Card system still makes less sense in a larger laptop than it does in a thin-and-light; the puzzle-like grid of input modules and spacers looks kind of rough and unfinished.

But the upgrades do help to shift things in the Laptop 16’s favor. Its modular and upgradeable design was always a theoretical selling point, but the laptop now actually offers options that other laptops don’t.

The presence of both AMD and Nvidia GPUs is a big step up in flexibility for both gaming and professional applications. The GeForce module is a better all-around choice, with slightly to significantly faster game performance and proprietary technologies like DLSS and CUDA, while the Radeon GPU is a cheaper option with better support for Linux.

Given their cost, I still wish that these GPUs were more powerful—they’re between $350 or $449 for the Radeon RX 7700S and between $650 and $699 for the RTX 5070 (prices vary a bit and are cheaper when you’re buying them together with a new laptop rather than buying them separately). You’ll basically always spend more for a gaming laptop than you will for a gaming desktop with similar or better performance, but that does feel like an awful lot to spend for GPUs that are still limited to 8GB of RAM.

Cost is a major issue for the Laptop 16 in general. You may save money in the long run by buying a laptop that you can replace piece-by-piece as you need to rather than all at once. But it’s not even remotely difficult to find similar specs from the major PC makers for hundreds of dollars less. We can’t vouch for the build quality or longevity of any of those PCs, but it does mean that you have to be willing to pay an awful lot just for Framework’s modularity and upgradeability. That’s true to some degree of the Laptop 13 as well, but the price gap between the 13 and competing systems isn’t as large as it is for the 16.

Whatever its lingering issues, the Framework Laptop 16 is still worth considering because there’s nothing else quite like it, at least if you’re in the market for something semi-portable and semi-powerful. The MacBook Pro exists if you want something more appliance-like, and there’s a whole spectrum of gaming and workstation PCs in between with all kinds of specs, sizes, and prices. To stand out from those devices, it’s probably better to be unique than to be perfect, and the reformulated Laptop 16 certainly clears that bar.

The good

Modular, repairable, upgradeable design that’s made to last
Cool, customizable keyboard deck
Nvidia GeForce GPU option gives the Laptop 16 access to some gaming and GPU computing features that weren’t usable with AMD GPUs
GPU upgrade can be added to first-generation Framework Laptop 16
New processors are a decent performance improvement and are worth considering for new buyers
Old Ryzen 7040-series motherboard is sticking around as an entry-level option, knocking $100 off the former base price ($1,299 and up for a barebones DIY edition, $1,599 and up for the cheapest pre-built)
Framework’s software support has gotten better in the last year

The bad

Big and bulky for the specs you get
Mix-and-match input modules and spacers give it a rough, unfinished sort of look
Ryzen AI motherboards are more expensive than the originals were when they launched

The ugly

It’ll cost you—the absolute bare minimum price for Ryzen AI 7 350 and RTX 5070 combo is $2,149, and that’s without RAM, an SSD, or an operating system

Andrew is a Senior Technology Reporter at Ars Technica, with a focus on consumer tech including computer hardware and in-depth reviews of operating systems like Windows and macOS. Andrew lives in Philadelphia and co-hosts a weekly book podcast called Overdue.

Review: New Framework Laptop 16 takes a fresh stab at the upgradeable laptop GPU Read More »

OpenAI signs massive AI compute deal with Amazon

AI, AI chips, AI infrastructure, Amazon, AWS, Biz & IT, chatbots, datacenters, generative ai, Google, large language models, machine learning, microsoft, NVIDIA, openai, oracle, sam altman / Shannon Garcia / November 3, 2025

On Monday, OpenAI announced it has signed a seven-year, $38 billion deal to buy cloud services from Amazon Web Services to power products like ChatGPT and Sora. It’s the company’s first big computing deal after a fundamental restructuring last week that gave OpenAI more operational and financial freedom from Microsoft.

The agreement gives OpenAI access to hundreds of thousands of Nvidia graphics processors to train and run its AI models. “Scaling frontier AI requires massive, reliable compute,” OpenAI CEO Sam Altman said in a statement. “Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”

OpenAI will reportedly use Amazon Web Services immediately, with all planned capacity set to come online by the end of 2026 and room to expand further in 2027 and beyond. Amazon plans to roll out hundreds of thousands of chips, including Nvidia’s GB200 and GB300 AI accelerators, in data clusters built to power ChatGPT’s responses, generate AI videos, and train OpenAI’s next wave of models.

Wall Street apparently liked the deal, because Amazon shares hit an all-time high on Monday morning. Meanwhile, shares for long-time OpenAI investor and partner Microsoft briefly dipped following the announcement.

Massive AI compute requirements

It’s no secret that running generative AI models for hundreds of millions of people currently requires a lot of computing power. Amid chip shortages over the past few years, finding sources of that computing muscle has been tricky. OpenAI is reportedly working on its own GPU hardware to help alleviate the strain.

But for now, the company needs to find new sources of Nvidia chips, which accelerate AI computations. Altman has previously said that the company plans to spend $1.4 trillion to develop 30 gigawatts of computing resources, an amount that is enough to roughly power 25 million US homes, according to Reuters.

OpenAI signs massive AI compute deal with Amazon Read More »

Nvidia hits record $5 trillion mark as CEO dismisses AI bubble concerns

AI, AI chips, AI infrastructure, Apple, Biz & IT, Blackwell chips, data centers, Donald Trump, GPUs, Jensen Huang, machine learning, microsoft, Nokia, NVIDIA, openai, palantir, semiconductors, Uber / DJ Henderson / October 29, 2025

Partnerships and government contracts fuel optimism

At the GTC conference on Tuesday, Nvidia’s CEO went out of his way to repeatedly praise Donald Trump and his policies for accelerating domestic tech investment while warning that excluding China from Nvidia’s ecosystem could limit US access to half the world’s AI developers. The overall event stressed Nvidia’s role as an American company, with Huang even nodding to Trump’s signature slogan in his sign-off by thanking the audience for “making America great again.”

Trump’s cooperation is paramount for Nvidia because US export controls have effectively blocked Nvidia’s AI chips from China, costing the company billions of dollars in revenue. Bob O’Donnell of TECHnalysis Research told Reuters that “Nvidia clearly brought their story to DC to both educate and gain favor with the US government. They managed to hit most of the hottest and most influential topics in tech.”

Beyond the political messaging, Huang announced a series of partnerships and deals that apparently helped ease investor concerns about Nvidia’s future. The company announced collaborations with Uber Technologies, Palantir Technologies, and CrowdStrike Holdings, among others. Nvidia also revealed a $1 billion investment in Nokia to support the telecommunications company’s shift toward AI and 6G networking.

The agreement with Uber will power a fleet of 100,000 self-driving vehicles with Nvidia technology, with automaker Stellantis among the first to deliver the robotaxis. Palantir will pair Nvidia’s technology with its Ontology platform to use AI techniques for logistics insights, with Lowe’s as an early adopter. Eli Lilly plans to build what Nvidia described as the most powerful supercomputer owned and operated by a pharmaceutical company, relying on more than 1,000 Blackwell AI accelerator chips.

The $5 trillion valuation surpasses the total cryptocurrency market value and equals roughly half the size of the pan European Stoxx 600 equities index, Reuters notes. At current prices, Huang’s stake in Nvidia would be worth about $179.2 billion, making him the world’s eighth-richest person.

Nvidia hits record $5 trillion mark as CEO dismisses AI bubble concerns Read More »

New physical attacks are quickly diluting secure enclave defenses from Nvidia, AMD, and Intel

AMD, battering ram, Biz & IT, Features, intel, NVIDIA, secure enclaves, Security, tee.fail, trusted execution environments, wiretap / Shannon Garcia / October 29, 2025

On-chip TEEs withstand rooted OSes but fall instantly to cheap physical attacks.

Trusted execution environments, or TEEs, are everywhere—in blockchain architectures, virtually every cloud service, and computing involving AI, finance, and defense contractors. It’s hard to overstate the reliance that entire industries have on three TEEs in particular: Confidential Compute from Nvidia, SEV-SNP from AMD, and SGX and TDX from Intel. All three come with assurances that confidential data and sensitive computing can’t be viewed or altered, even if a server has suffered a complete compromise of the operating kernel.

A trio of novel physical attacks raises new questions about the true security offered by these TEES and the exaggerated promises and misconceptions coming from the big and small players using them.

The most recent attack, released Tuesday, is known as TEE.fail. It defeats the latest TEE protections from all three chipmakers. The low-cost, low-complexity attack works by placing a small piece of hardware between a single physical memory chip and the motherboard slot it plugs into. It also requires the attacker to compromise the operating system kernel. Once this three-minute attack is completed, Confidential Compute, SEV-SNP, and TDX/SDX can no longer be trusted. Unlike the Battering RAM and Wiretap attacks from last month—which worked only against CPUs using DDR4 memory—TEE.fail works against DDR5, allowing them to work against the latest TEEs.

Some terms apply

All three chipmakers exclude physical attacks from threat models for their TEEs, also known as secure enclaves. Instead, assurances are limited to protecting data and execution from viewing or tampering, even when the kernel OS running the processor has been compromised. None of the chipmakers make these carveouts prominent, and they sometimes provide confusing statements about the TEE protections offered.

Many users of these TEEs make public assertions about the protections that are flat-out wrong, misleading, or unclear. All three chipmakers and many TEE users focus on the suitability of the enclaves for protecting servers on a network edge, which are often located in remote locations, where physical access is a top threat.

“These features keep getting broken, but that doesn’t stop vendors from selling them for these use cases—and people keep believing them and spending time using them,” said HD Moore, a security researcher and the founder and CEO of runZero.

He continued:

Overall, it’s hard for a customer to know what they are getting when they buy confidential computing in the cloud. For on-premise deployments, it may not be obvious that physical attacks (including side channels) are specifically out of scope. This research shows that server-side TEEs are not effective against physical attacks, and even more surprising, Intel and AMD consider these out of scope. If you were expecting TEEs to provide private computing in untrusted data centers, these attacks should change your mind.

Those making these statements run the gamut from cloud providers to AI engines, blockchain platforms, and even the chipmakers themselves. Here are some examples:

Cloudflare says it’s using Secure Memory Encryption—the encryption engine driving SEV—to safeguard confidential data from being extracted from a server if it’s stolen.
In a post outlining the possibility of using the TEEs to secure confidential information discussed in chat sessions, Anthropic says the enclave “includes protections against physical attacks.”
Microsoft marketing (here and here) devotes plenty of ink to discussing TEE protections without ever noting the exclusion.
Meta, paraphrasing the Confidential Computing Consortium, says TEE security provides protections against malicious “system administrators, the infrastructure owner, or anyone else with physical access to the hardware.” SEV-SNP is a key pillar supporting the security of Meta’s WhatsApp Messenger.
Even Nvidia claims that its TEE security protects against “infrastructure owners such as cloud providers, or anyone with physical access to the servers.”
The maker of the Signal private messenger assures users that its use of SGX means that “keys associated with this encryption never leave the underlying CPU, so they’re not accessible to the server owners or anyone else with access to server infrastructure.” Signal has long relied on SGX to protect contact-discovery data.

I counted more than a dozen other organizations providing assurances that were similarly confusing, misleading, or false. Even Moore—a security veteran with more than three decades of experience—told me: “The surprising part to me is that Intel/AMD would blanket-state that physical access is somehow out of scope when it’s the entire point.”

In fairness, some TEE users build additional protections on top of the TEEs provided out of the box. Meta, for example, said in an email that the WhatsApp implementation of SEV-SNP uses protections that would block TEE.fail attackers from impersonating its servers. The company didn’t dispute that TEE.fail could nonetheless pull secrets from the AMD TEE.

The Cloudflare theft protection, meanwhile, relies on SME—the engine driving SEV-SNP encryption. The researchers didn’t directly test SME against TEE.fail. They did note that SME uses deterministic encryption, the cryptographic property that causes all three TEEs to fail. (More about the role of deterministic encryption later.)

Others who misstate the TEEs’ protections provide more accurate descriptions elsewhere. Given all the conflicting information, it’s no wonder there’s confusion.

How do you know where the server is? You don’t.

Many TEE users run their infrastructure inside cloud providers such as AWS, Azure, or Google, where protections against supply-chain and physical attacks are extremely robust. That raises the bar for a TEE.fail-style attack significantly. (Whether the services could be compelled by governments with valid subpoenas to attack their own TEE is not clear.)

All these caveats notwithstanding, there’s often (1) little discussion of the growing viability of cheap, physical attacks, (2) no evidence (yet) that implementations not vulnerable to the three attacks won’t fall to follow-on research, or (3) no way for parties relying on TEEs to know where the servers are running and whether they’re free from physical compromise.

“We don’t know where the hardware is,” Daniel Genkin, one of the researchers behind both TEE.fail and Wiretap, said in an interview. “From a user perspective, I don’t even have a way to verify where the server is. Therefore, I have no way to verify if it’s in a reputable facility or an attacker’s basement.”

In other words, parties relying on attestations from servers in the cloud are once again reduced to simply trusting other people’s computers. As Moore observed, solving that problem is precisely the reason TEEs exist.

In at least two cases, involving the blockchain services Secret Network and Crust, the loss of TEE protections made it possible for any untrusted user to present cryptographic attestations. Both platforms used the attestations to verify that a blockchain node operated by one user couldn’t tamper with the execution or data passing to another user’s nodes. The Wiretap hack on SGX made it possible for users to run the sensitive data and executions outside of the TEE altogether while still providing attestations to the contrary. In the AMD attack, the attacker could decrypt the traffic passing through the TEE.

Both Secret Network and Crust added mitigations after learning of the possible physical attacks with Wiretap and Battering RAM. Given the lack of clear messaging, other TEE users are likely making similar mistakes.

A predetermined weakness

The root cause of all three physical attacks is the choice of deterministic encryption. This form of encryption produces the same ciphertext each time the same plaintext is encrypted with the same key. A TEE.fail attacker can copy ciphertext strings and use them in replay attacks. (Probabilistic encryption, by contrast, resists such attacks because the same plaintext can encrypt to a wide range of ciphertexts that are randomly chosen during the encryption process.)

TEE.fail works not only against SGX but also a more advanced Intel TEE known as TDX. The attack also defeats the protections provided by the latest Nvidia Confidential Compute and AMD SEV-SNP TEEs. Attacks against TDX and SGX can extract the Attestation Key, an ECDSA secret that certifies to a remote party that it’s running up-to-date software and can’t expose data or execution running inside the enclave. This Attestation Key is in turn signed by an Intel X.509 digital certificate providing cryptographic assurances that the ECDSA key can be trusted. TEE.fail works against all Intel CPUs currently supporting TDX and SDX.

With possession of the key, the attacker can use the compromised server to peer into data or tamper with the code flowing through the enclave and send the relying party an assurance that the device is secure. With this key, even CPUs built by other chipmakers can send an attestation that the hardware is protected by the Intel TEEs.

GPUs equipped with Nvidia Confidential Compute don’t bind attestation reports to the specific virtual machine protected by a specific GPU. TEE.fail exploits this weakness by “borrowing” a valid attestation report from a GPU run by the attacker and using it to impersonate the GPU running Confidential Compute. The protection is available on Nvidia’s H100/200 and B100/200 server GPUs.

“This means that we can convince users that their applications (think private chats with LLMs or Large Language Models) are being protected inside the GPU’s TEE while in fact it is running in the clear,” the researchers wrote on a website detailing the attack. “As the attestation report is ‘borrowed,’ we don’t even own a GPU to begin with.”

SEV-SNP (Secure Encrypted Virtualization-Secure Nested Paging) uses ciphertext hiding in AMD’s EPYC CPUs based on the Zen 5 architecture. AMD added it to prevent a previous attack known as Cipherleaks, which allowed malicious hypervisors to extract cryptographic keys stored in the enclaves of a virtual machine. Ciphertext, however, doesn’t stop physical attacks. With the ability to reopen the side channel that Cipherleaks relies on, TEE.fail can steal OpenSSL credentials and other key material based on constant-time encryption.

Cheap, quick, and the size of a briefcase

“Now that we have interpositioned DDR5 traffic, our work shows that even the most modern of TEEs across all vendors with available hardware is vulnerable to cheap physical attacks,” Genkin said.

The equipment required by TEE.fail runs off-the-shelf gear that costs less than $1,000. One of the devices the researchers built fits into a 17-inch briefcase, so it can be smuggled into a facility housing a TEE-protected server. Once the physical attack is performed, the device does not need to be connected again. Attackers breaking TEEs on servers they operate have no need for stealth, allowing them to use a larger device, which the researchers also built.

A logic analyzer attached to an interposer.

The researchers demonstrated attacks against an array of services that rely on the chipmakers’ TEE protections. (For ethical reasons, the attacks were carried out against infrastructure that was identical to but separate from the targets’ networks.) Some of the attacks included BuilderNet, dstack, and Secret Network.

BuilderNet is a network of Ethereum block builders that uses TDX to prevent parties from snooping on others’ data and to ensure fairness and that proof currency is redistributed honestly. The network builds blocks valued at millions of dollars each month.

“We demonstrated that a malicious operator with an attestation key could join BuilderNet and obtain configuration secrets, including the ability to decrypt confidential orderflow and access the Ethereum wallet for paying validators,” the TEE.fail website explained. “Additionally, a malicious operator could build arbitrary blocks or frontrun (i.e., construct a new transaction with higher fees to ensure theirs is executed first) the confidential transactions for profit while still providing deniability.”

To date, the researchers said, BuilderNet hasn’t provided mitigations. Attempts to reach BuilderNet officials were unsuccessful.

dstack is a tool for building confidential applications that run on top of virtual machines protected by Nvidia Confidential Compute. The researchers used TEE.fail to forge attestations certifying that a workload was performed by the TDX using the Nvidia protection. It also used the “borrowed” attestations to fake ownership of GPUs that a relying party trusts.

Secret Network is a platform billing itself as the “first mainnet blockchain with privacy-preserving smart contracts,” in part by encrypting on-chain data and execution with SGX. The researchers showed that TEE.fail could extract the “Concensus Seed,” the primary network-side private key encrypting confidential transactions on the Secret Network. As noted, after learning of Wiretap, the Secret Network eliminated this possibility by establishing a “curated” allowlist of known, trusted nodes allowed on the network and suspended the acceptance of new nodes. Academic or not, the ability to replicate the attack using TEE.fail shows that Wiretap wasn’t a one-off success.

A tough nut to crack

As explained earlier, the root cause of all the TEE.fail attacks is deterministic encryption, which forms the basis for protections in all three chipmakers’ TEEs. This weaker form of encryption wasn’t always used in TEEs. When Intel initially rolled out SGX, the feature was put in client CPUs, not server ones, to prevent users from building devices that could extract copyrighted content such as high-definition video.

Those early versions encrypted no more than 256MB of RAM, a small enough space to use the much stronger probabilistic form of encryption. The TEEs built into server chips, by contrast, must often encrypt terabytes of RAM. Probabilistic encryption doesn’t scale to that size without serious performance penalties. Finding a solution that accommodates this overhead won’t be easy.

One mitigation over the short term is to ensure that each 128-bit block of ciphertext has sufficient entropy. Adding random plaintext to the blocks prevents ciphertext repetition. The researchers say the entropy can be added by building a custom memory layout that inserts a 64-bit counter with a random initial value to each 64-bit block before encrypting it.

The last countermeasure the researchers proposed is adding location verification to the attestation mechanism. While insider and supply chain attacks remain a possibility inside even the most reputable cloud services, strict policies make them much less feasible. Even those mitigations, however, don’t foreclose the threat of a government agency with a valid subpoena ordering an organization to run such an attack inside their network.

In a statement, Nvidia said:

NVIDIA is aware of this research. Physical controls in addition to trust controls such as those provided by Intel TDX reduce the risk to GPUs for this style of attack, based on our discussions with the researchers. We will provide further details once the research is published.

Intel spokesman Jerry Bryant said:

Fully addressing physical attacks on memory by adding more comprehensive confidentiality, integrity and anti-replay protection results in significant trade-offs to Total Cost of Ownership. Intel continues to innovate in this area to find acceptable solutions that offer better balance between protections and TCO trade-offs.

The company has published responses here and here reiterating that physical attacks are out of scope for both TDX and SGX

AMD didn’t respond to a request for comment.

Stuck on Band-Aids

For now, TEE.fail, Wiretap, and Battering RAM remain a persistent threat that isn’t solved with the use of default implementations of the chipmakers’ secure enclaves. The most effective mitigation for the time being is for TEE users to understand the limitations and curb uses that the chipmakers say aren’t a part of the TEE threat model. Secret Network tightening requirements for operators joining the network is an example of such a mitigation.

Moore, the founder and CEO of RunZero, said that companies with big budgets can rely on custom solutions built by larger cloud services. AWS, for example, makes use of the Nitro Card, which is built using ASIC chips that accelerate processing using TEEs. Google’s proprietary answer is Titanium.

“It’s a really hard problem,” Moore said. “I’m not sure what the current state of the art is, but if you can’t afford custom hardware, the best you can do is rely on the CPU provider’s TEE, and this research shows how weak this is from the perspective of an attacker with physical access. The enclave is really a Band-Aid or hardening mechanism over a really difficult problem, and it’s both imperfect and dangerous if compromised, for all sorts of reasons.”

Dan Goodin is Senior Security Editor at Ars Technica, where he oversees coverage of malware, computer espionage, botnets, hardware hacking, encryption, and passwords. In his spare time, he enjoys gardening, cooking, and following the independent music scene. Dan is based in San Francisco. Follow him at here on Mastodon and here on Bluesky. Contact him on Signal at DanArs.82.

New physical attacks are quickly diluting secure enclave defenses from Nvidia, AMD, and Intel Read More »

Ars Live recap: Is the AI bubble about to pop? Ed Zitron weighs in.

AI, AI agents, AI chips, AI criticism, AI economics, AI infrastructure, Benj Edwards, Biz & IT, bubble, Cerebras, datacenters, Ed Zitron, financial losses, generative ai, GPU, grok, large language models, machine learning, NVIDIA, openai, oracle, Power consumption, sam altman, Stargate, venture capital / DJ Henderson / October 18, 2025

Despite connection hiccups, we covered OpenAI’s finances, nuclear power, and Sam Altman.

On Tuesday of last week, Ars Technica hosted a live conversation with Ed Zitron, host of the Better Offline podcast and one of tech’s most vocal AI critics, to discuss whether the generative AI industry is experiencing a bubble and when it might burst. My Internet connection had other plans, though, dropping out multiple times and forcing Ars Technica’s Lee Hutchinson to jump in as an excellent emergency backup host.

During the times my connection cooperated, Zitron and I covered OpenAI’s financial issues, lofty infrastructure promises, and why the AI hype machine keeps rolling despite some arguably shaky economics underneath. Lee’s probing questions about per-user costs revealed a potential flaw in AI subscription models: Companies can’t predict whether a user will cost them $2 or $10,000 per month.

You can watch a recording of the event on YouTube or in the window below.

Our discussion with Ed Zitron. Click here for transcript.

“A 50 billion-dollar industry pretending to be a trillion-dollar one”

I started by asking Zitron the most direct question I could: “Why are you so mad about AI?” His answer got right to the heart of his critique: the disconnect between AI’s actual capabilities and how it’s being sold. “Because everybody’s acting like it’s something it isn’t,” Zitron said. “They’re acting like it’s this panacea that will be the future of software growth, the future of hardware growth, the future of compute.”

In one of his newsletters, Zitron describes the generative AI market as “a 50 billion dollar revenue industry masquerading as a one trillion-dollar one.” He pointed to OpenAI’s financial burn rate (losing an estimated $9.7 billion in the first half of 2025 alone) as evidence that the economics don’t work, coupled with a heavy dose of pessimism about AI in general.

Donald Trump listens as Nvidia CEO Jensen Huang speaks at the White House during an event on “Investing in America” on April 30, 2025, in Washington, DC. Credit: Andrew Harnik / Staff | Getty Images News

“The models just do not have the efficacy,” Zitron said during our conversation. “AI agents is one of the most egregious lies the tech industry has ever told. Autonomous agents don’t exist.”

He contrasted the relatively small revenue generated by AI companies with the massive capital expenditures flowing into the sector. Even major cloud providers and chip makers are showing strain. Oracle reportedly lost $100 million in three months after installing Nvidia’s new Blackwell GPUs, which Zitron noted are “extremely power-hungry and expensive to run.”

Finding utility despite the hype

I pushed back against some of Zitron’s broader dismissals of AI by sharing my own experience. I use AI chatbots frequently for brainstorming useful ideas and helping me see them from different angles. “I find I use AI models as sort of knowledge translators and framework translators,” I explained.

After experiencing brain fog from repeated bouts of COVID over the years, I’ve also found tools like ChatGPT and Claude especially helpful for memory augmentation that pierces through brain fog: describing something in a roundabout, fuzzy way and quickly getting an answer I can then verify. Along these lines, I’ve previously written about how people in a UK study found AI assistants useful accessibility tools.

Zitron acknowledged this could be useful for me personally but declined to draw any larger conclusions from my one data point. “I understand how that might be helpful; that’s cool,” he said. “I’m glad that that helps you in that way; it’s not a trillion-dollar use case.”

He also shared his own attempts at using AI tools, including experimenting with Claude Code despite not being a coder himself.

“If I liked [AI] somehow, it would be actually a more interesting story because I’d be talking about something I liked that was also onerously expensive,” Zitron explained. “But it doesn’t even do that, and it’s actually one of my core frustrations, it’s like this massive over-promise thing. I’m an early adopter guy. I will buy early crap all the time. I bought an Apple Vision Pro, like, what more do you say there? I’m ready to accept issues, but AI is all issues, it’s all filler, no killer; it’s very strange.”

Zitron and I agree that current AI assistants are being marketed beyond their actual capabilities. As I often say, AI models are not people, and they are not good factual references. As such, they cannot replace human decision-making and cannot wholesale replace human intellectual labor (at the moment). Instead, I see AI models as augmentations of human capability: as tools rather than autonomous entities.

Computing costs: History versus reality

Even though Zitron and I found some common ground about AI hype, I expressed a belief that criticism over the cost and power requirements of operating AI models will eventually not become an issue.

I attempted to make that case by noting that computing costs historically trend downward over time, referencing the Air Force’s SAGE computer system from the 1950s: a four-story building that performed 75,000 operations per second while consuming two megawatts of power. Today, pocket-sized phones deliver millions of times more computing power in a way that would be impossible, power consumption-wise, in the 1950s.

The blockhouse for the Semi-Automatic Ground Environment at Stewart Air Force Base, Newburgh, New York. Credit: Denver Post via Getty Images

“I think it will eventually work that way,” I said, suggesting that AI inference costs might follow similar patterns of improvement over years and that AI tools will eventually become commodity components of computer operating systems. Basically, even if AI models stay inefficient, AI models of a certain baseline usefulness and capability will still be cheaper to train and run in the future because the computing systems they run on will be faster, cheaper, and less power-hungry as well.

Zitron pushed back on this optimism, saying that AI costs are currently moving in the wrong direction. “The costs are going up, unilaterally across the board,” he said. Even newer systems like Cerebras and Grok can generate results faster but not cheaper. He also questioned whether integrating AI into operating systems would prove useful even if the technology became profitable, since AI models struggle with deterministic commands and consistent behavior.

The power problem and circular investments

One of Zitron’s most pointed criticisms during the discussion centered on OpenAI’s infrastructure promises. The company has pledged to build data centers requiring 10 gigawatts of power capacity (equivalent to 10 nuclear power plants, I once pointed out) for its Stargate project in Abilene, Texas. According to Zitron’s research, the town currently has only 350 megawatts of generating capacity and a 200-megawatt substation.

“A gigawatt of power is a lot, and it’s not like Red Alert 2,” Zitron said, referencing the real-time strategy game. “You don’t just build a power station and it happens. There are months of actual physics to make sure that it doesn’t kill everyone.”

He believes many announced data centers will never be completed, calling the infrastructure promises “castles on sand” that nobody in the financial press seems willing to question directly.

An orange, cloudy sky backlights a set of electrical wires on large pylons, leading away from the cooling towers of a nuclear power plant.

After another technical blackout on my end, I came back online and asked Zitron to define the scope of the AI bubble. He says it has evolved from one bubble (foundation models) into two or three, now including AI compute companies like CoreWeave and the market’s obsession with Nvidia.

Zitron highlighted what he sees as essentially circular investment schemes propping up the industry. He pointed to OpenAI’s $300 billion deal with Oracle and Nvidia’s relationship with CoreWeave as examples. “CoreWeave, they literally… They funded CoreWeave, became their biggest customer, then CoreWeave took that contract and those GPUs and used them as collateral to raise debt to buy more GPUs,” Zitron explained.

When will the bubble pop?

Zitron predicted the bubble would burst within the next year and a half, though he acknowledged it could happen sooner. He expects a cascade of events rather than a single dramatic collapse: An AI startup will run out of money, triggering panic among other startups and their venture capital backers, creating a fire-sale environment that makes future fundraising impossible.

“It’s not gonna be one Bear Stearns moment,” Zitron explained. “It’s gonna be a succession of events until the markets freak out.”

The crux of the problem, according to Zitron, is Nvidia. The chip maker’s stock represents 7 to 8 percent of the S&P 500’s value, and the broader market has become dependent on Nvidia’s continued hyper growth. When Nvidia posted “only” 55 percent year-over-year growth in January, the market wobbled.

“Nvidia’s growth is why the bubble is inflated,” Zitron said. “If their growth goes down, the bubble will burst.”

He also warned of broader consequences: “I think there’s a depression coming. I think once the markets work out that tech doesn’t grow forever, they’re gonna flush the toilet aggressively on Silicon Valley.” This connects to his larger thesis: that the tech industry has run out of genuine hyper-growth opportunities and is trying to manufacture one with AI.

“Is there anything that would falsify your premise of this bubble and crash happening?” I asked. “What if you’re wrong?”

“I’ve been answering ‘What if you’re wrong?’ for a year-and-a-half to two years, so I’m not bothered by that question, so the thing that would have to prove me right would’ve already needed to happen,” he said. Amid a longer exposition about Sam Altman, Zitron said, “The thing that would’ve had to happen with inference would’ve had to be… it would have to be hundredths of a cent per million tokens, they would have to be printing money, and then, it would have to be way more useful. It would have to have efficacy that it does not have, the hallucination problems… would have to be fixable, and on top of this, someone would have to fix agents.”

A positivity challenge

Near the end of our conversation, I wondered if I could flip the script, so to speak, and see if he could say something positive or optimistic, although I chose the most challenging subject possible for him. “What’s the best thing about Sam Altman,” I asked. “Can you say anything nice about him at all?”

“I understand why you’re asking this,” Zitron started, “but I wanna be clear: Sam Altman is going to be the reason the markets take a crap. Sam Altman has lied to everyone. Sam Altman has been lying forever.” He continued, “Like the Pied Piper, he’s led the markets into an abyss, and yes, people should have known better, but I hope at the end of this, Sam Altman is seen for what he is, which is a con artist and a very successful one.”

Then he added, “You know what? I’ll say something nice about him, he’s really good at making people say, ‘Yes.’”

Benj Edwards is Ars Technica’s Senior AI Reporter and founder of the site’s dedicated AI beat in 2022. He’s also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC.

Ars Live recap: Is the AI bubble about to pop? Ed Zitron weighs in. Read More »

Nvidia sells tiny new computer that puts big AI on your desktop

Acer, AI, AI chips, AI development tools, AI GPU, AI infrastructure, ai supercomputer, ASUS, Biz & IT, desktop computing, DGX Spark, elon musk, Google, Grace Blackwell, HP, Jensen Huang, machine learning, msi, NVIDIA, openai, Project DIGITS, spacex, Tech / 9u50fv / October 14, 2025

For the OS, the Spark is an ARM-based system that runs Nvidia’s DGX OS, an Ubuntu Linux-based operating system built specifically for GPU processing. It comes with Nvidia’s AI software stack preinstalled, including CUDA libraries and the company’s NIM microservices.

Prices for the DGX Spark start at US $3,999. That may seem like a lot, but given the cost of high-end GPUs with ample video RAM like the RTX Pro 6000 (about $9,000) or AI server GPUs (like $25,000 for a base-level H100), the DGX Spark may represent a far less expensive option overall, though it’s not nearly as powerful.

In fact, according to The Register, the GPU computing performance of the GB10 chip is roughly equivalent to an RTX 5070. However, the 5070 is limited to 12GB of video memory, which limits the size of AI models that can be run on such a system. With 128GB of unified memory, the DGX Spark can run far larger models, albeit at a slower speed than, say, an RTX 5090 (which typically ships with 24 GB of RAM). For example, to run the 120 billion-parameter larger version of OpenAI’s recent gpt-oss language model, you’d need about 80GB of memory, which is far more than you can get in a consumer GPU.

A callback to 2016

Nvidia founder and CEO Jensen Huang marked the occasion of the DGX Spark launch by personally delivering one of the first units to Elon Musk at SpaceX’s Starbase facility in Texas, echoing a similar delivery Huang made to Musk at OpenAI in 2016.

“In 2016, we built DGX-1 to give AI researchers their own supercomputer. I hand-delivered the first system to Elon at a small startup called OpenAI, and from it came ChatGPT,” Huang said in a statement. “DGX-1 launched the era of AI supercomputers and unlocked the scaling laws that drive modern AI. With DGX Spark, we return to that mission.”

Nvidia sells tiny new computer that puts big AI on your desktop Read More »

Bank of England warns AI stock bubble rivals 2000 dotcom peak

AI, AI stocks, alphabet, Amazon, andrew bailey, Apple, bank of england, Biz & IT, Donald Trump, dotcom bubble, Federal Reserve, financial stability, Google, Lisa Cook, machine learning, mark zuckerberg, market correction, Meta, microsoft, NVIDIA, UK economy / DJ Henderson / October 9, 2025

Share valuations based on past earnings have also reached their highest levels since the dotcom bubble 25 years ago, though the BoE noted they appear less extreme when based on investors’ expectations for future profits. “This, when combined with increasing concentration within market indices, leaves equity markets particularly exposed should expectations around the impact of AI become less optimistic,” the central bank said.

Toil and trouble?

The dotcom bubble offers a potentially instructive parallel to our current era. In the late 1990s, investors poured money into Internet companies based on the promise of a transformed economy, seemingly ignoring whether individual businesses had viable paths to profitability. Between 1995 and March 2000, the Nasdaq index rose 600 percent. When sentiment shifted, the correction was severe: the Nasdaq fell 78 percent from its peak, reaching a low point in October 2002.

Whether we’ll see the same thing or worse if an AI bubble pops is mere speculation at this point. But similar to the early 2000s, the question about today’s market isn’t necessarily about the utility of AI tools themselves (the Internet was useful, afterall, despite the bubble), but whether the amount of money being poured into the companies that sell them is out of proportion with the potential profits those improvements might bring.

We don’t have a crystal ball to determine when such a bubble might pop, or even if it is guaranteed to do so, but we’ll likely continue to see more warning signs ahead if AI-related deals continue to grow larger and larger over time.

Bank of England warns AI stock bubble rivals 2000 dotcom peak Read More »

AMD wins massive AI chip deal from OpenAI with stock sweetener

AI, AI chips, AI development tools, AI GPU, AI infrastructure, AMD, Biz & IT, Broadcom, GPUs, machine learning, microsoft, NVIDIA, openai / Kelly Newman / October 6, 2025

As part of the arrangement, AMD will allow OpenAI to purchase up to 160 million AMD shares at 1 cent each throughout the chips deal.

OpenAI diversifies its chip supply

With demand for AI compute growing rapidly, companies like OpenAI have been looking for secondary supply lines and sources of additional computing capacity, and the AMD partnership is part the company’s wider effort to secure sufficient computing power for its AI operations. In September, Nvidia announced an investment of up to $100 billion in OpenAI that included supplying at least 10 gigawatts of Nvidia systems. OpenAI plans to deploy a gigawatt of Nvidia’s next-generation Vera Rubin chips in late 2026.

OpenAI has worked with AMD for years, according to Reuters, providing input on the design of older generations of AI chips such as the MI300X. The new agreement calls for deploying the equivalent of 6 gigawatts of computing power using AMD chips over multiple years.

Beyond working with chip suppliers, OpenAI is widely reported to be developing its own silicon for AI applications and has partnered with Broadcom, as we reported in February. A person familiar with the matter told Reuters the AMD deal does not change OpenAI’s ongoing compute plans, including its chip development effort or its partnership with Microsoft.

AMD wins massive AI chip deal from OpenAI with stock sweetener Read More »

Big AI firms pump money into world models as LLM advances slow

AI, Google, Meta, NVIDIA, openai, superintelligence, syndication, world models / Mike M. / September 29, 2025

Runway, a video generation start-up that has deals with Hollywood studios, including Lionsgate, launched a product last month that uses world models to create gaming settings, with personalized stories and characters generated in real time.

“Traditional video methods [are a] brute-force approach to pixel generation, where you’re trying to squeeze motion in a couple of frames to create the illusion of movement, but the model actually doesn’t really know or reason about what’s going on in that scene,” said Cristóbal Valenzuela, chief executive officer at Runway.

Previous video-generation models had physics that were unlike the real world, he added, which general-purpose world model systems help to address.

To build these models, companies need to collect a huge amount of physical data about the world.

San Francisco-based Niantic has mapped 10 million locations, gathering information through games including Pokémon Go, which has 30 million monthly players interacting with a global map.

Niantic ran Pokémon Go for nine years and, even after the game was sold to US-based Scopely in June, its players still contribute anonymized data through scans of public landmarks to help build its world model.

“We have a running start at the problem,” said John Hanke, chief executive of Niantic Spatial, as the company is now called following the Scopely deal.

Both Niantic and Nvidia are working on filling gaps by getting their world models to generate or predict environments. Nvidia’s Omniverse platform creates and runs such simulations, assisting the $4.3 trillion tech giant’s push toward robotics and building on its long history of simulating real-world environments in video games.

Nvidia Chief Executive Jensen Huang has asserted that the next major growth phase for the company will come with “physical AI,” with the new models revolutionizing the field of robotics.

Some such as Meta’s LeCun have said this vision of a new generation of AI systems powering machines with human-level intelligence could take 10 years to achieve.

But the potential scope of the cutting-edge technology is extensive, according to AI experts. World models “open up the opportunity to service all of these other industries and amplify the same thing that computers did for knowledge work,” said Nvidia’s Lebaredian.

Additional reporting by Melissa Heikkilä in London and Michael Acton in San Francisco.

Big AI firms pump money into world models as LLM advances slow Read More »

Why does OpenAI need six giant data centers?

AI, AI chips, AI development tools, AI infrastructure, Biz & IT, chatgpt, Clay Magouyrk, CoreWeave, datacenters, machine learning, Masayoshi Son, NVIDIA, openai, oracle, sam altman, SB Energy, SoftBank / 9u50fv / September 24, 2025

Training next-generation AI models compounds the problem. On top of running existing AI models like those that power ChatGPT, OpenAI is constantly working on new technology in the background. It’s a process that requires thousands of specialized chips running continuously for months.

The circular investment question

The financial structure of these deals between OpenAI, Oracle, and Nvidia has drawn scrutiny from industry observers. Earlier this week, Nvidia announced it would invest up to $100 billion as OpenAI deploys Nvidia systems. As Bryn Talkington of Requisite Capital Management told CNBC: “Nvidia invests $100 billion in OpenAI, which then OpenAI turns back and gives it back to Nvidia.”

Oracle’s arrangement follows a similar pattern, with a reported $30 billion-per-year deal where Oracle builds facilities that OpenAI pays to use. This circular flow, which involves infrastructure providers investing in AI companies that become their biggest customers, has raised eyebrows about whether these represent genuine economic investments or elaborate accounting maneuvers.

The arrangements are becoming even more convoluted. The Information reported this week that Nvidia is discussing leasing its chips to OpenAI rather than selling them outright. Under this structure, Nvidia would create a separate entity to purchase its own GPUs, then lease them to OpenAI, which adds yet another layer of circular financial engineering to this complicated relationship.

“NVIDIA seeds companies and gives them the guaranteed contracts necessary to raise debt to buy GPUs from NVIDIA, even though these companies are horribly unprofitable and will eventually die from a lack of any real demand,” wrote tech critic Ed Zitron on Bluesky last week about the unusual flow of AI infrastructure investments. Zitron was referring to companies like CoreWeave and Lambda Labs, which have raised billions in debt to buy Nvidia GPUs based partly on contracts from Nvidia itself. It’s a pattern that mirrors OpenAI’s arrangements with Oracle and Nvidia.

So what happens if the bubble pops? Even Altman himself warned last month that “someone will lose a phenomenal amount of money” in what he called an AI bubble. If AI demand fails to meet these astronomical projections, the massive data centers built on physical soil won’t simply vanish. When the dot-com bubble burst in 2001, fiber optic cable laid during the boom years eventually found use as Internet demand caught up. Similarly, these facilities could potentially pivot to cloud services, scientific computing, or other workloads, but at what might be massive losses for investors who paid AI-boom prices.

Why does OpenAI need six giant data centers? Read More »

If you own a Volvo EX90, you’re getting a free computer upgrade

Cars, NVIDIA, Volvo EX90 / Shannon Garcia / September 19, 2025

If you own a 2025 Volvo EX90, here’s some good news: You’re getting a car computer upgrade. Even better news? It’s free.

The Swedish automaker says that owners of model year 2025 EX90s—like the one we tested earlier this summer—are eligible for an upgrade to the electric vehicle’s core computer. Specifically, the cars will get a new dual Nvidia DRIVE AGX Orin setup, which Volvo says will improve performance and reduce battery drainage, as well as enabling some features that have been TBD so far.

That will presumably be welcome news—the EX90 is a shining example of how the “minimal viable product” idea has infiltrated the auto industry from the tech sphere. That’s because Volvo has had a heck of a time with the EX90 development, having to delay the EV not once but twice in order to get a handle on the car’s software.

When we got our first drive in the electric SUV this time last year, that London Taxi-like hump on the roof contained a functional lidar that wasn’t actually integrated into the car’s advanced driver-assistance systems. In fact, a whole load of features weren’t ready yet, not just ADAS features.

The EX90 was specced with a single Orin chip, together with a less-powerful Xavier chip, also from Nvidia. But that combo isn’t up to the job, and for the ES90 electric sedan, the automaker went with a pair of Orins. And that’s what it’s going to retrofit to existing MY25 EX90s, gratis.

If you own a Volvo EX90, you’re getting a free computer upgrade Read More »