Q.ANT, a startup out of Stuttgart, Germany, builds processors that use light instead of electricity to do some of the math behind AI. The company pitches them as a way to run AI on a fraction of the power today’s chips need.
Now developers can start writing software for those chips without owning one. Q.ANT pushed a free, open-source software kit to GitHub this week that lets developers build and test programs on a normal computer, then run them on the real chips once they get access.
This is a move out of Nvidia’s playbook. Nvidia owes its lead in AI as much to CUDA, the software developers use to program its GPUs, as it does to the chips themselves.
But with Q.ANT, the catch is the hardware. Q.ANT’s chips are running at a few research computing centers, and everyone else has to wait “the coming months” for cloud access through German provider IONOS or an on-site server from Q.ANT.
The kit, called the Q.ANT Native Computing Toolkit, is free on GitHub under a license that allows commercial use. Developers can work in Python or C. The key piece is a simulator that mimics the chip on a regular computer, with no Q.ANT drivers required.
What can it do today? The AI tools in this first version focus on running models that have already been trained. The examples read handwritten numbers, identify objects in photos and outline shapes in images. Training still happens on regular CPUs and GPUs.
The pitch for photonic computing is power. AI chips burn a lot of energy moving data back and forth between memory and the processor. Q.ANT’s chips do part of the math with light, specifically wave-shaped functions similar to a cosine, which regular chips calculate digitally. Q.ANT says AI models built around those functions get better results with fewer parameters, the settings a model learns during training. Fewer parameters means a smaller model, less data to move and less power. The kit includes examples comparing a standard model with one built Q.ANT’s way. Those comparisons are the company’s own.
“An ecosystem isn’t created by hardware alone. It emerges when the software layer is open and others can build on it,” said Michael Förtsch, Q.ANT’s founder and CEO. He calls the release the “Linux moment” of photonic computing.
Q.ANT is betting light can do the math itself. Lightmatter, one of the best-known companies in the field, now puts its focus on Passage, which uses light to move data between chips. The idea of light-based AI isn’t new, either. TNS covered MIT’s photonic processor for building optical neural networks back in 2017.
Q.ANT raised €62 million in July 2025 in a round led by Cherry Ventures, UVC Partners and imec.xpand. In March, it said its second-generation chips were running at the Leibniz Supercomputing Centre near Munich. The results it published from there compare the new chip with its old one: more than 50 times faster at the kind of math that does most of the work in AI models, and six times less energy on typical jobs, by the company’s numbers. Its bigger claims, like up to 30 times better energy efficiency, don’t say what they’re measured against.
Good software alone won’t carry a new chip. Nvidia has been building CUDA for nearly 20 years and is still adding to it, including deeper native Python support last year. Graphcore, the British AI chip startup, had its own software kit and still ended up being sold to SoftBank in 2024.
Q.ANT calls this the first openly available software kit for programming a photonic processor. That depends on how you count. Xanadu has offered free, open software for its light-based quantum computers since 2018. For now, developers can play with the simulator. What they can’t do yet is test Q.ANT’s power-saving claims on their own models. That has to wait until the chips open up.
The post Q.ANT gives away the software for its light-powered AI chips in a CUDA-style bet on developers appeared first on The New Stack.
The 1.58 in a 1.58-bit language model sounds like a hard limit, but Intel researchers pushed a ternary model below it by changing how its weights are stored rather than changing the model itself.
Their new BITCOS format compressed one checkpoint to 1.485 bits per weight and improved decoding throughput by as much as 18% on CPUs and 27% on GPUs.
The key is that the familiar 1.58-bit figure assumes a model uses its three possible weight values equally, while real ternary models contain far more zeros than that calculation accounts for. BITCOS stores the location and sign of each nonzero weight separately, allowing zeros to take up less space without retraining the model or altering its output — the equivalent of packing the same contents into a smaller box.
Ternary models use only three weight values — -1, 0, and +1 — and 1.58 bits is the theoretical minimum needed to represent three equally likely options. That number is cleaner than the reality of storing the weights, where the standard approach fits five ternary values into an eight-bit byte for an average of 1.6 bits each. Models commonly store weights in blocks of 128; however, this leaves the final byte partly unused and pushes the actual rate to 1.625 bits per weight.
When Intel’s researchers measured the distribution of weights across 29 checkpoints from seven ternary model families, they found that zeros accounted for between 29.7% and 51.5% of the weights. In 26 of those checkpoints, there were enough zeros for BITCOS to beat five-trit packing.
The sparsest was a ternary version of Qwen3-1.7B produced with CAT-Q post-training quantization, where 51.48% of the weights were zero, and BITCOS brought the storage cost down to 1.485 bits per weight.
Models commonly store weights in blocks of 128; however, this leaves the final byte partly unused and pushes the actual rate to 1.625 bits per weight.
BITCOS stands for “BITmap and COmpacted Signs” and divides a model’s weights into two streams. The first assigns one bit to every weight to record whether it is zero or nonzero, while the second assigns a sign bit only to nonzero weights.
A positive or negative weight therefore consumes two bits, but a zero needs only the presence bit because it has no sign to record.
If z is the proportion of zero weights, BITCOS uses 2 − z bits per weight, dropping from 1.6 bits at 40% zeros to 1.485 bits at 51.5%. Because it changes only the storage format, unpacking restores the original -1, 0 and +1 values without affecting accuracy.
BITCOS becomes smaller than five-trit packing once more than 37.5% of a model’s weights are zero, a threshold reached by 26 of the 29 checkpoints Intel examined.
BITCOS becomes smaller than five-trit packing once more than 37.5% of a model’s weights are zero, a threshold reached by 26 of the 29 checkpoints Intel examined.
Built for token-by-token decoding with small batch sizes, the format reduces the weight data moving through memory. Intel developed separate unpacking kernels for AVX-512 and AVX2 CPUs as well as Xe2 GPUs, joining other efforts to fit compressed models into faster inference pipelines for AI agents.
On AVX-512 hardware, the kernel uses the presence bitmap as a mask and pdep to scatter the compacted sign bits across the nonzero weight positions. Because Xe2 GPUs lack an equivalent instruction, Intel implemented the same operation with a 2KB lookup table.
Compared with the 2-bit kernels, BITCOS ran 10% to 18% faster on the 64-core Xeon server and 2% to 15% faster on the 24-core Core Ultra 9. Performance improved by 9% to 22% on the integrated Arc 140V and by 2% to 27% on the discrete Arc Pro B70. These results measure decoding after the model has loaded, separate from efforts to cut GPU inference cold starts from minutes to seconds.
On the eight-core Lunar Lake CPU, Intel’s fixed 2-bit kernel beat BITCOS on every model because the system had enough bandwidth to make unpacking the bottleneck. BITCOS remained faster on the GPUs, although decoding overhead limited the gains. Computer scientist and AI infrastructure author Chip Huyen has made the same point about inference more generally, arguing that the right optimization depends on whether compute, memory or bandwidth is holding back the workload.
On the eight-core Lunar Lake CPU, Intel’s fixed 2-bit kernel beat BITCOS on every model because the system had enough bandwidth to make unpacking the bottleneck.
The paper has not been peer-reviewed; all five test systems used Intel hardware, and the end-to-end benchmarks covered seven models at batch size one. Intel has yet to test the format on Nvidia, AMD, or Arm hardware.
The post Intel squeezed a 1.58-bit LLM down to 1.485 bits without changing a single weight appeared first on The New Stack.
Running an LLM on your PC is easy enough, but putting an agent to work there is a different story. Portable Computer, the local version of Perplexity’s Computer agent, is now available inside the Perplexity app for Windows on compatible Nvidia GeForce RTX and RTX PRO GPUs.
That’s the good news; the catch is, you’ll need an Nvidia GPU with at least 24GB of VRAM.
The Windows launch gives Perplexity three platforms in less than three weeks. Portable Computer debuted on Linux and Nvidia DGX Spark on August 25, followed a week later by hybrid compute for Apple silicon, which splits tasks between local and cloud models on Macs. Now Windows joins the mix, but bringing Portable Computer over took more than simply porting the app. Perplexity had to adapt the model runtime, orchestration, security, and hardware integration for each platform while keeping the user experience the same.
you’ll need an Nvidia GPU with at least 24GB of VRAM to use it.
Portable Computer bundles those pieces together. On Windows, it supports PPLX 27B — Perplexity’s post-trained model — and Qwen 3.8 27B, both optimized for RTX GPUs, alongside a built-in browser, tool calling, and Perplexity’s proprietary SPACE sandbox.
It’s a different lane from LM Studio or Ollama, which make running models locally as painless as possible but stop well short of giving a model autonomy over multistep work. DeepSeek’s recent hiring spree of roughly 150 new roles, nearly all of them focused on agent infrastructure rather than the model, hints at how much engineering sits between a capable model and a capable agent.
DeepSeek’s recent hiring spree of roughly 150 new roles, nearly all of them focused on agent infrastructure rather than the model, hints at how much engineering sits between a capable model and a capable agent.
Perplexity ships connectors for Microsoft Outlook, OneDrive, and Word, plus Google Drive, Gmail, Slack, and GitHub — which tells you something about what “local” actually means here.
The agent can reach external services because it’s not air-gapped. Locally completed tasks can process files without sending documents to a cloud model. Once an agent has access to both local files and remote APIs on the same machine, figuring out which resources it actually needs and where to find them gets harder.
Perplexity isn’t pretending that a 27-billion-parameter model running on a desktop GPU can handle everything, which explains the hybrid architecture. When the agent determines that a task needs more reasoning power than the local model can deliver, it can escalate to Perplexity’s cloud models.
According to Nvidia, the agent identifies when cloud support would help and asks the user for permission before sending any data off the machine.
For organizations handling sensitive or regulated data, that split can make all the difference. A local agent can grind through source code or financial records without uploading them to a hosted model for basic processing. There’s a cost angle too, since tasks completed locally don’t burn Perplexity Computer credits.
Portable Computer is available with Perplexity Pro ($20/month) and Max ($200/month), across individual and enterprise plans, with Nvidia DGX Station support coming later. The real challenge is taking local agents from developer passion projects to enterprise-ready tools. By baking this into Windows, it immediately gets in front of the scale of users needed to make that happen.
The real challenge is taking local agents from developer passion projects to enterprise-ready tools.
The post Perplexity’s new agent runs entirely on your GPU — with one expensive catch appeared first on The New Stack.
Perplexity wants its AI agents to use more of the computing power already sitting inside a Mac. The company launched Hybrid Compute on Tuesday, a new feature that lets Perplexity Computer move parts of the same task between powerful cloud models and smaller models running directly on Apple silicon.
The timing is significant. Hybrid Compute arrives on the first day of John Ternus’ tenure as Apple’s CEO. Ternus, who previously led Apple’s hardware engineering organization and played a key role in the company’s transition from Intel processors to its own Apple silicon, succeeds Tim Cook after 15 years as CEO. Perplexity is now betting that same hardware can become part of the infrastructure behind autonomous AI agents.
The task starts in the cloud. When a step involves sensitive information, Computer can move that part of the job to a model running on the Mac without having to start over.
When a step involves sensitive information, Computer can move that part of the job to a model running on the Mac without starting over.
A Perplexity-trained Privacy Gate runs locally on the Mac and looks for sensitive information such as names, addresses, account numbers, and secrets. When it flags something, the user can decide whether that part of the task should stay on the device.
Users can review what the system wants to keep local before work begins and ensure it hasn’t missed anything they don’t want sent to the cloud. They can also choose which model handles the local work.
At launch, users can choose between Gemma E4B, Qwen3.6 35B-A3B, and a version of Qwen3.6 35B that Perplexity post-trained itself, with more models planned for later. Perplexity also handles the installation through the desktop app, so users don’t have to open a terminal or set up the model themselves.
Once a task is running, the app displays local CPU, GPU, and memory utilization, as well as the number of tokens consumed. Users aren’t charged for tokens generated by models running locally on their Mac. That gives Perplexity another reason to move work onto the device beyond privacy, since running models locally can also cut inference costs. (The question of who bears inference costs and when is becoming a competitive issue across the industry.)
There is a compromise. Perplexity uses its most capable models in the cloud, while smaller models handle local work, so keeping more of a task on the Mac can improve privacy and cut costs at the expense of some capability. Hybrid Compute leaves that choice to the user.
Perplexity uses its most capable models in the cloud, while smaller models handle local work, so keeping more of a task on the Mac can improve privacy and cut costs at the expense of some capability.
The trickier bit is what happens to context when part of a task moves onto the Mac. The cloud model still needs to know enough about what happened locally to continue the job, without getting access to the private information that was supposed to stay there.
Perplexity says Computer can move a step from the cloud to a local model without restarting the task or losing context, and ultimately combine the cloud and local work into a single result. It does not detail in its announcement exactly what context is passed between those environments or how information produced by the local subagent is filtered before it returns to the broader workflow.
The local subagent can work with private files and data and take actions on the Mac. Users can also start a task on an iPhone and hand off local work to their Mac without having to start over.
For enterprise customers, Perplexity adds company-wide rules for what stays local and a record of what leaves each device, bringing the same governance questions facing AI agents down to the device level.
Hybrid Compute reverses the approach Perplexity introduced for Nvidia’s DGX Spark last week. DGX Spark starts locally and reaches out to frontier cloud models only with permission, while the Mac version starts in the cloud and moves work onto the device when needed.
In both cases, the agent harness decides where each part of the job runs. That orchestration — deciding which tools and context an agent actually needs — is becoming an increasingly difficult engineering problem as agents gain access to more systems and data.
Running this much of an agent locally still requires a fairly powerful Mac. Perplexity recommends at least 32GB of unified memory, Apple silicon, and macOS 15. The feature is available to Pro and Max subscribers as well as enterprise customers.
Those requirements show the limits of local AI at the moment. Smaller models can run on plenty of Macs, but giving an agent enough compute to handle meaningful work still requires relatively high-end hardware.
That will likely change as Macs get better at running larger models. For Ternus, the bigger question is how much of the AI work now happening in the cloud will eventually move onto the machines Apple sells.
For Ternus, the bigger question is how much of the AI work now happening in the cloud will eventually move onto the machines Apple sells.
The post Your Mac is now part of Perplexity’s AI infrastructure appeared first on The New Stack.
SpaceX and Nvidia say they are adapting the Vera Rubin NVL72 rack-scale AI platform for orbital use, with SpaceX targeting a first launch in the fourth quarter of 2027.
The dream of an AI data center in space lives on in SpaceX and Nvidia’s August 24 announcements that the platform for Low Earth Orbit (LEO) Starmind AI satellites will be based on the Vera Rubin NVL72 chip family and architecture.
This proposed system would form the computing core of SpaceXAI’s first-generation Starmind AI satellite and extend Nvidia’s architecture from terrestrial AI data centers into space.
SpaceX CEO Elon Musk posted on X the same day, “SpaceX, in partnership with Nvidia, has designed a space-optimized Vera Rubin NVL72 system for launch to orbit in Q4 next year, with significant scale in 2028.”
SpaceX, in partnership with Nvidia, has designed a space-optimized Vera Rubin NVL72 system for launch to orbit in Q4 next year, with significant scale in 2028 https://t.co/qdDq8YBkzl
— Elon Musk (@elonmusk) August 24, 2026
Musk’s post came after he said during SpaceX’s Q2 earnings call, “Going forward, we’ve decided to build exclusively on Nvidia because we think the Vera Rubin architecture is the best architecture.” Musk continued, “This is not some sort of far-future, distant thing; we expect to start launching these next year. We think the design of the NVL72 VR computer is a much better design than, say, having a standard rack -style design. So we expect to deploy this on the ground as well as in orbit, because we think it’s going to be a radical simplification of the standard NVL72 rack. It will cost less. It will be more effective. If we’re going to put it in space, why not want to put it on the ground? I think that’s going to be pretty cool.”
On Earth, the Vera Rubin NVL72 is Nvidia’s rack-scale AI design that combines 72 Rubin GPUs and 36 Vera CPUs, alongside high-speed networking components such as ConnectX-9 SuperNICs. Nvidia says SpaceXAI’s planned Starmind satellite will be based on an optimized version of that system.
A conventional NVL72 rack assumes gravity, technicians, stable grid power, a building-scale liquid loop and frequent replacement of failed parts. Orbit removes each of these assumptions.
The idea is more ambitious than putting a conventional edge-AI accelerator aboard a spacecraft. Nvidia and SpaceXAI are proposing to bring a modified architecture used in AI data centers into orbit, while altering it for orbital operational requirements.
Getting that working in orbit, though, is easier said than done.
As Curtis Pyke, founder of Kingy AI, writes, “A conventional NVL72 rack assumes gravity, technicians, stable grid power, a building-scale liquid loop and frequent replacement of failed parts. Orbit removes each of these assumptions.“
“Cooling is unforgiving. Space is cold, but vacuum does not carry heat away through convection.”
In particular, Pyke continues, “Cooling is unforgiving. Space is cold, but vacuum does not carry heat away through convection. Heat must travel from the chips to the radiator surfaces and then leave as infrared radiation. SpaceX says AI1 can avoid chillers, cooling towers and fans and reduce cooling overhead by an order of magnitude.”
SpaceX explains that AI1 would instead use closed-loop liquid cooling inside the spacecraft and large deployable radiators to send heat directly to space as infrared radiation. While the claimed reduction is physically plausible in principle, there’s no proof yet that these AI satellites’ cooling systems can deliver.
Another major problem that remains unaddressed is how to make the orbital rack radiation-tolerant. Making Vera Rubin NVL72 radiation-tolerant means far more than putting an ordinary NVL72 rack in a shielded satellite enclosure. It would require a system-level redesign of its GPUs, CPUs, memory, networking, power, cooling, firmware, and operations around a specified orbit and mission life.
LEO orbit is not benign. NASA cites typical trapped-particle dose rates of 100 to 1,000 rad(Si) per year for low-inclination LEO spacecraft below 500 km. That level of radiation is not an immediate death sentence for electronics, but over a multiyear mission it will cause cumulative degradation. Radiation-qualified space hardware can deal with that. Commercial Off-The-Shelf (COTS) electronics are another matter. A true radiation-hardened Rubin GPU would also require design changes at the transistor and circuit levels.
Even were Nvidia to make such a chip, for a high-density AI system such as the SpaceX design, the concern isn’t simply whether one processor survives a 5- or 10-year dose. The satellite contains numerous radiation-sensitive elements, such as GPU logic, SRAM caches, register files, system memory, and memory controllers. With thousands of cores and billions of memory storage cells, the aggregate fault rate — not the behavior of an individual component — drives the design.
The most realistic near-term answer would be a radiation-tolerant, fault-managed Rubin-derived orbital system, not a fully radiation-hardened NVL72 in the traditional military-space sense. It could use selected commercial Nvidia parts, substantial shielding, ECC and data integrity mechanisms, redundant controllers and power paths, aggressive fault detection, software recovery, and reduced-performance operating modes.
The post SpaceX designed an orbital Vera Rubin. Radiation comes next. appeared first on The New Stack.
Nvidia has reportedly agreed to buy Hugging Face for $12.9 billion, putting one of the biggest names in AI hardware in charge of a platform developers rely on to find and run open models.
The Information first reported the deal Wednesday, citing a person familiar with the agreement. Nvidia and Hugging Face had not publicly confirmed it as of publication.
Hugging Face doesn’t push developers toward one chipmaker, which is what makes the acquisition interesting. Its Optimum libraries work with Nvidia’s TensorRT-LLM and also support hardware from AMD, Intel, and AWS. Projects such as Optimum AMD and Optimum Intel let developers run Transformers and Diffusers models on non-Nvidia hardware.
The company also plays a role in what happens after a developer chooses a model, including how easily they can get it running on the hardware they want to use. The problem for Nvidia might be this: It is buying a platform whose value depends on openness and hardware neutrality, but if the purchase means Nvidia hardware is favored, that value might diminish.
The company also plays a role in what happens after a developer chooses a model, including how easily they can get it running on the hardware they want to use.
Hugging Face has expanded well beyond file hosting. With Inference Endpoints, developers can deploy a model from the Hub while Hugging Face handles the underlying infrastructure.
Those hosted deployments can run on AWS, Microsoft Azure or Google Cloud, but most of the GPU options Hugging Face lists are Nvidia chips, including the T4, L4 and A100. That gives developers a wider choice of hardware through Hugging Face’s open-source libraries than through its hosted services.
If the deal goes through, Nvidia would own both sides of that experience.
NIM (Nvidia Inference Microservices) already works with models hosted on Hugging Face. Developers can point NIM to an hf:// repository path and pull the model directly from the Hub.
Owning Hugging Face would give Nvidia more room to bring NIM and CUDA-optimized containers directly into the deployment experience. Nvidia hasn’t announced plans to make NIM the default, and support for AMD and Intel could remain exactly where it is.
Owning Hugging Face would give Nvidia more room to bring NIM and CUDA-optimized containers directly into the deployment experience.
The bigger question is what happens over time. Nvidia could provide earlier support for new models on its own hardware or make deployment easier. At the same time, AMD, Intel, and AWS may have to reconsider how much engineering work they want to contribute to integrations maintained within a competitor-owned platform. Some of that work could eventually move elsewhere.
For developers, the difference may come down to which path requires less work. A competing chip doesn’t have to disappear from Hugging Face to become less appealing if an Nvidia model deployment takes fewer steps. We’ve seen a similar fight over the layers between AI models and the developers using them, with Cloudflare building more of that infrastructure itself.
That tension also helps explain why Hugging Face could be worth considerably more to Nvidia than its revenue alone would suggest.
Nvidia has been expanding its own Nemotron family of open models while investing heavily across the AI ecosystem. At the same time, some of its biggest customers are working to reduce their dependence on Nvidia hardware.
Google has its TPUs, AWS has Trainium, and Microsoft has Maia. OpenAI and Anthropic are also developing their own AI server chips. The push extends beyond the hyperscalers. Earlier this month, five European companies committed to purchasing AI compute built around non-Nvidia hardware that hasn’t been manufactured yet — a sign that the appetite for alternative accelerators is strong enough to attract forward contracts.
OpenAI this week published results from its new Jalapeño accelerator, which showed 1.5 to 1.9 times more work per watt while cutting end-to-end latency by up to 3.6 times on large open-weight models — although the chip has not yet been deployed at anything approaching Nvidia’s scale. A strong open-model ecosystem gives Nvidia a counterweight to that trend.
Open models are often expected to run in very different environments, and Hugging Face helps developers make that possible. A model found on the Hub might end up running on Nvidia hardware, an AMD GPU, or a cloud accelerator.
That flexibility is part of what Nvidia would be buying. Pushing Hugging Face too heavily toward its own hardware could make the platform less useful to developers who rely on it to work across different systems.
Interest in those models is also growing. Models from companies including DeepSeek, Moonshot AI and Z.ai have narrowed the gap with proprietary systems. At the same time, Hugging Face CEO Clément Delangue told The Information in June that the company had doubled its number of paying subscribers during the first six months of 2026. Delangue later said the company was “close to profitability.”
The Information puts Hugging Face’s annualized revenue at about $150 million. Against a $12.9 billion price tag, that’s a multiple of roughly 86.
So Nvidia would be paying for much more than Hugging Face’s current business. It would be buying a place developers already turn to when they want to work with open models, including models that don’t have to run on Nvidia hardware.
It would be buying a place developers already turn to when they want to work with open models, including models that don’t have to run on Nvidia hardware.
Buying Hugging Face wouldn’t give Nvidia control over everything developers find there. Libraries such as transformers and diffusers are open source, and models on the Hub remain subject to their own licenses. Openly licensed models can still be hosted elsewhere, while the underlying libraries can be forked.
Much harder to recreate is the community Hugging Face has built around them. Developers already know where to look for models and have built workflows around the Hub and its integrations.
The post Nvidia’s $12.9B Hugging Face deal has an open-source problem appeared first on The New Stack.
When OpenAI unveiled Jalapeño, its first custom inference chip, in June, the company made some big promises. The chip, developed with Broadcom, was built from scratch for large language model inference, with OpenAI saying early testing showed substantially better performance per watt than existing accelerators. At the time, though, OpenAI didn’t release the detailed performance results to back that up.
On Tuesday, OpenAI published its first results from working Jalapeño silicon across GPT-OSS 120B, DeepSeek R1 and Kimi K2.5. The results show what OpenAI was aiming for with Jalapeño: higher throughput without the longer response times that can come with it.
“Agents need to complete many steps in sequence, so delays can compound across an entire task.”
An agent may call a model over and over as it works through a task, using tools and deciding what to do next based on the results, which means a delay that barely registers during a single inference can become much more noticeable when it happens repeatedly over the course of a longer task.
“Agents need to complete many steps in sequence, so delays can compound across an entire task,” OpenAI said.
Jalapeño was designed with those delays in mind. Different parts of running a large language model place different demands on the hardware, with the initial prompt requiring more compute and the response generation putting more pressure on memory bandwidth. Every time data has to move between cores and chips, that can add even more waiting.
Jalapeño takes a different approach, cutting down on that waiting without optimizing one part of the process at the expense of another.
“Agents need to complete many steps in sequence so that delays can compound across an entire task,” OpenAI said.
That helps explain some of the choices OpenAI made with Jalapeño. Running a large language model puts different demands on the hardware at different points: processing the initial prompt requires a lot of compute, while generating the response token by token relies more heavily on memory bandwidth. There’s also time lost whenever data has to move between cores and chips, leaving parts of the system waiting for what they need.
The idea is to reduce that waiting without optimizing one part of the process at the expense of another. Model state, including the KV cache used while generating a response, can be kept local, while Jalapeño’s networking allows more of the workload to stay within the same connected system. That means less time spent moving data around as the workload shifts between compute and memory.
OpenAI put Jalapeño through InferenceX, SemiAnalysis’ public benchmark for AI inference, using GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Across the three models, Jalapeño handled 1.5 to 1.9 times more work per watt while cutting end-to-end latency by 1.7 to 3.6 times. On highly interactive workloads, OpenAI says it was 2.1 to 4.1 times faster than the systems it compared against.
The differences become particularly large when Jalapeño is compared at the previous best time-between-tokens operating point. OpenAI reported between 8.6 and 104.3 times more work per watt, depending on the model.
OpenAI based the power-efficiency comparisons on each accelerator’s published power rating. Jalapeño is rated at 700 watts, although the company says it never drew more than 550 watts during these tests. The bigger point is that OpenAI isn’t trying to improve throughput at the expense of response time, which is often the tradeoff with inference.
Batching more work can make infrastructure more efficient, but it can also mean making an individual user wait longer. OpenAI’s argument with Jalapeño is that an inference system increasingly needs to be good at both — particularly as the company continues cutting the cost of API access while also needing to keep interactive workloads responsive.
The bigger point is that OpenAI isn’t trying to improve throughput at the expense of response time, which is often the tradeoff with inference.
OpenAI used its own models throughout Jalapeño’s development, helping the hardware team move from initial design to tapeout in nine months by exploring implementations and shortening design, measurement, and verification cycles. The work didn’t stop once the chip was built.
The company says AI-generated implementations of selected GPT-OSS attention and mixture-of-experts blocks ran 1.5 to 1.8 times faster than versions written by its own experts. That doesn’t mean the entire model ran that much faster, but it does show what OpenAI is trying to do with Jalapeño: make the chip straightforward enough for AI, not just humans, to program and optimize.
Engineers describe work using local tensors, explicit communication and predictable synchronization, giving AI a way to help determine how that work should be mapped, placed and scheduled across the system. That could make it faster to adapt the chip as new models come along, although OpenAI says each new model family still requires its own kernels and optimizations.
Using Codex with GPT-Astra and earlier OpenAI models, the hardware team brought three open-weight models that weren’t part of Jalapeño’s original production plan to high performance within two months. That fits with OpenAI’s broader plans for Codex, which the company has said is still early in its development, and shows how it could eventually play a role well beyond writing code.
OpenAI plans to start using Jalapeño in its own infrastructure by the end of the year, and it’s already working on the next two generations. The company will continue to use accelerators from Nvidia and other partners, but building its own chips gives OpenAI more control over how the hardware evolves alongside its models.
OpenAI plans to start using Jalapeño in its own infrastructure by the end of the year, and it’s already working on the next two generations.
The post OpenAI’s Jalapeño chip tackles a problem AI agents make worse appeared first on The New Stack.
AI data centers in space sound great, but practically speaking, they may be next to impossible.
For tech bros, it sounds great. Two of the buzziest tech giants, SpaceX and NVIDIA, are partnering together to bring AI data centers into space using the just-announced Starmind AI1 satellite.
These 30-meter-tall satellites with a 75-meter solar-array wingspan will contain the latest NVIDIA Vera CPUs and Rubin GPUs. These will live in a Low Earth Orbit (LEO) of about 600 kilometers. For networking, it will use Starlink’s laser links. SpaceX says the first AI1 spacecraft will perform localized AI computing in orbit and relay results to Earth via Starlink.
According to SpaceX, AI1 is designed around a compute payload drawing up to 250 kW at peak and 175 kW on average. It will be solar-powered, unlike its Earth-bound competitors, which frequently require the construction of new power plants.
Starmind is not simply a conventional NVIDIA AI cluster launched into orbit. The effort hinges on integrating high-density accelerator hardware with a spacecraft platform capable of generating power, rejecting waste heat, surviving radiation, maintaining laser communications, and being produced in large quantities. None of that is easy.
Once in orbit, which will require SpaceX’s still-not-ready-for-prime-time Starship rockets to launch the estimated 2.3-metric-ton satellites, the satellites will work together.
Eventually, to reach SpaceX’s goal of a million (that’s not a typo, that’s a million) Starmind satellites, the two companies will need to design a standard model spacecraft. These will be built in SpaceX’s 11-million-square-foot manufacturing campus, Gigasat Factory, which is still under construction in Bastrop County, Texas.
This AI-in-space proposal is the most ambitious yet of SpaceX CEO Elon Musk’s dream of placing energy-intensive AI infrastructure in orbit. There, these satellites won’t need to compete for land, electrical-grid capacity, or water with increasingly contentious terrestrial data center buildouts.
However, SpaceX glosses over the technical issues of turning this vision into reality.
Let’s start with the biggest headache: Cooling.
Contrary to what you may think from bad science-fiction movies, the vacuum of space is not cold per se. Whether the surface of an object is hot or cold depends entirely on whether it’s facing the sun. Those on the sun side will heat up, while those away from the sun will eventually cool down toward the 3 Kelvin background of deep space.
The keyword is “eventually.” You can’t simply use convection, cooling towers, or evaporative cooling to carry away heat. The heat must radiate away as infrared radiation, and that’s a very slow process.
The physics creates a direct trade-off between computing power, radiator area, spacecraft mass, and operating temperature. A system running hundreds of kilowatts of AI hardware must reject nearly all of that power as waste heat. Liquid cooling can carry heat away from chips, but it does not eliminate the requirement for extensive radiator surfaces.
As NASA has found, “satellites experience harsh environments in orbit,” ranging from about 393 Kelvin in full sun (248 degrees Fahrenheit) all the way down to ~3 Kelvin (-454 degrees Fahrenheit).
To cool down the Starmind satellites, each will have a deployable liquid radiator system measuring 160 square meters. What liquid? We don’t know yet. Hugh Lewis, a professor of astronautics at the University of Birmingham, expects it to use ammonia, which is already used on the International Space Station (ISS). Whether this will reliably scale to data-center-class AI deployments with their enormous heat remains to be seen.
Another issue is its networking. The architecture depends heavily on Starlink’s optical inter-satellite links. SpaceX says AI1 satellites will use high-speed laser links to communicate with other spacecraft and send AI results to Earth via the Starlink network.
Starlink’s published technology specifications describe mini laser terminals operating at up to 25 Gbps across distances as long as 4,000 kilometers, while SpaceX cites roughly 25-millisecond latency for its customer service.
Those figures suggest a potentially useful network for distributing inference results, transmitting model updates, connecting orbital sensors to compute nodes, and avoiding some reliance on ground-station passes. But they do not establish that a satellite constellation can function like the tightly coupled networking fabric of a terrestrial AI supercomputer.
We won’t be seeing large-scale machine learning and training in space. This requires huge, predictable bandwidth and very low latency for GPU-to-GPU communications. An orbital network would also face physical propagation delays, laser-link acquisition and handoffs, routing across a moving constellation, and limits on available capacity per spacecraft.
Another issue, according to Doug Mohney, a long-time space influencer, is debris. “One bad day, a piece of random junk hits one satellite, which fragments into multiple pieces of shrapnel, which hits another satellite and so on and so on until you get a Kessler event that turns the selective orbit into a roaming cloud of debris.”
A Kessler event is when one satellite breaks up, and its fragments hit another, and so on until an area of LEO is filled with wreckage rather than viable satellites.
Adding insult to injury, a Kessler event may not happen by accident. Mohney also observes that space warfare is a real threat: “A bad actor such as Russia, China, Iran, or North Korea could use kinetic (unrandom junk!) means to target one or more satellites, resulting in space debris.” Or, “One good nuclear weapon uses an electromagnetic pulse to get rid of all of them at once. Both Russia and China (and the US) already have anti-satellite weapons (ASAT) programs. North Korea could have ASAT, but a nuke would ensure mass destruction of orbital capability.”
If that sounds crazy, keep in mind that Starlink satellites are already being used by Ukraine, and Russia has been trying to block their transmissions. There have also been credible reports of Russia developing ASAT weapons specifically designed to knock Starlink satellites out of the sky. Larger and more fragile Starmind satellites would be far more vulnerable.
Mohney also worries about the “known unknown” of space weather.
“A Solar flare that hit the Earth along the lines of the 1859 Carrington Event, the largest recorded solar storm, would take out orbital electronics of all satellites.” This, in turn, as uncontrolled satellites drift from their orbit, might cause a Kessler event. Lesser events have already pushed LEO satellites out of space. For example, a February 2022 geomagnetic storm forced thirty-eight newly launched Starlink satellites out of orbit.
There are also business concerns. For all the obstacles that new and expanded ground-based AI data centers face, the energy analytics firm Wood Mackenzie believes “A hypothetical 1 GW orbital data center would cost an estimated $170 billion, more than three times the equivalent terrestrial facility, with launch and satellite costs accounting for approximately 60% of that total. To bring orbital costs to parity with terrestrial alternatives would require a 70% reduction.”
The company thinks that might be possible, but Robert Liew, Wood Mackenzie Research Director, observes, “That gap does not close without sustained and dramatic progress on launch costs. We forecast US$ 9 trillion of terrestrial data center investment between now and 2040. That is where capital goes first. Orbital data centers are a serious long-term proposition, but right now they remain a bet on the cost curve.”
For now, SpaceX has offered a broad technical vision and a hardware partnership with NVIDIA, but few of the operational metrics that would establish commercial viability. The real test will be whether SpaceX Starship becomes a practical launch vehicle and can overcome its cooling and safety issues. Then, the AI1 must also show enough usable compute per kilogram, kilowatt, square meter of radiator, and dollar of launch cost to outperform or complement ground-based AI infrastructure. I don’t see this happening anytime soon.
The post Why space is actually a terrible place to cool a data center appeared first on The New Stack.
This week during an interview with Bloomberg, Jensen Huang made quite the prediction.
The Nvidia CEO said the semiconductor industry will need to grow roughly five to tenfold over the next decade to support AI agents and robots to support what he believes is the next wave of computing. Huang believes that future demand will come from autonomous software agents and physical robots consuming compute around the clock.
“In the future, we have AI agents and robots, and they will be using computers,” Huang said. “Instead of a billion people using computers, we will have 100 billion agents and billions of robots all using computers. The computer industry built on top of the chip industry is certainly not big enough. Computers are being built not just for people to use, but computers are being built for computers to use.”
“Instead of a billion people using computers, we will have 100 billion agents and billions of robots all using computers. The computer industry built on top of the chip industry is certainly not big enough.”
The 5-10x forecast — which Huang framed as his personal estimate, not a certainty — builds on a message he has been repeating for months, including a recent appearance where he declared traditional coding dead in favor of engineers who build AI agents. Still, it reflects the need to build backend systems for AI agents and machines, something infrastructure teams are already contending with.
On Nvidia’s fiscal Q1 2027 earnings call in May, Huang described the move from generative AI to agentic AI — systems “capable of perceiving, reasoning, planning, and acting” — as the next major phase of the industry.
As API requests come from AI agents more often, standard assumptions around rate limiting, session memory, sub-millisecond inference routing, and API gateway concurrency are starting to break down. An environment in which most traffic originates from autonomous background loops rather than human thumbs changes how backend infrastructure must be built from the ground up.
To support an endpoint explosion of this scale, the physical supply chain must scale dramatically at the memory and data center layers. Speaking at the AI Summit in San Francisco on July 24, Huang pointed to South Korea as an important linchpin of the global AI buildout. “This is truly the beginning of a golden age for Korea,” he said, noting that the country’s semiconductor and industrial capabilities position it to help the world build out AI infrastructure.
“This is truly the beginning of a golden age for Korea.”
To back that vision, Nvidia announced a comprehensive partnership with SK Group valued at over $500 billion. The initiative spans massive purchasing of next-generation High-Bandwidth Memory (HBM) from SK Hynix, jointly co-developing custom HBM4 roadmaps designed specifically for agentic and physical AI workloads, and deploying Nvidia supercomputers.
The announcement also included major infrastructure investments across South Korea. SK Telecom said it plans to build a 2-gigawatt AI data center using Nvidia’s Vera Rubin architecture and SK Hynix’s HBM4 memory, with the first facility expected to come online in 2027. At the same time, Nvidia will invest $1 billion in Naver Corp, with Brookfield funding up to $9 billion as the project’s infrastructure capital partner, to help expand the company’s AI data center capacity from 55 megawatts to 200 megawatts by 2028.
Huang’s prediction also helps explain why Nvidia and other infrastructure companies are locking up supply years in advance. The company recently disclosed $119 billion in supply-related commitments as it works to secure everything from advanced packaging capacity to power, land, and high-bandwidth memory.
“Computers are being built not just for people to use, but computers are being built for computers to use.”
Huang believes the industry needs to stop thinking about a world where computers primarily serve people and start planning for one where AI agents and robots generate much of the demand. In his view, the ultimate limiting factor will be whether the industry can build enough physical infrastructure to keep up — a constraint already reshaping how companies like Nvidia and Palantir approach sovereign AI deployments.
The post Jensen Huang says AI agents could drive a 5-10x computing boom: “100 billion agents and billions of robots” appeared first on The New Stack.
The race to make AI inference cheaper is pushing chip design beyond general-purpose accelerators. We’re now moving toward silicon that is tailored to individual models. And Google’s reported “Frozen v2” project suggests the Gemini AI model is part of the movement.
First reported by The Information, the unannounced chip would reportedly hardwire parts of Gemini’s architecture while leaving its weights updatable. That compromise could give Google much of the efficiency of model-specific silicon without making the hardware obsolete every time Gemini changes.
A spokesperson for Google tells The New Stack, “Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers.”
A spokesperson for Google tells The New Stack, “Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers. While not every project moves into production, this rigorous exploration is central to our full stack approach. By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimized for real-world workloads.”
According to the reporting, Google hopes the chip will help relieve the AI compute crunch that’s made it harder for cloud providers to keep up with demand, while also making Gemini much cheaper and more efficient to serve. Internal projections reportedly estimate the design could deliver six to ten times more tokens per watt than Google’s current generation of AI chips.
If the project moves forward, it would represent a different approach to AI infrastructure because Google would be building hardware specifically for Gemini. For developers, that’s an early indication that future AI systems may be designed with much tighter integration between the model and the hardware beneath it.
Currently, the vast majority of AI inferences run on Nvidia GPUs or Google’s own Tensor Processing Units (TPUs). Because these chips are intended to accommodate a wide variety of AI models, they carry a high degree of processing load.
There’s precedent for this kind of shift. Bitcoin mining followed a similar path, starting with CPUs, then GPUs, before ultimately settling on ASICs. AI inference may be headed in the same direction. Training still benefits from the flexibility of GPUs, but once a model reaches production, the priority shifts to serving as many requests as possible while using less power.
Training still benefits from the flexibility of GPUs, but once a model reaches production, the priority shifts to serving as many requests as possible while using less power.
Google isn’t the only company looking beyond general-purpose GPUs for AI inference. As inference becomes a larger share of AI workloads, more companies are experimenting with specialized hardware designed to improve performance while using less power. Even Nvidia, whose GPUs dominate the AI market, has invested heavily in inference, striking a reported $20 billion deal with Groq last year to license the startup’s technology.
Among the more ambitious efforts is Canadian startup Taalas. The company has demonstrated a chip with an entire 8-billion-parameter Llama model embedded directly into the silicon. By keeping the model on the chip instead of constantly moving data back and forth from external memory, Taalas says it can dramatically speed up inference, claiming throughput of roughly 17,000 tokens per second. Meanwhile, other companies are optimizing the spatial distance between memory and compute to avoid hardware lock-in.
d-Matrix’s new Corsair platform uses an SRAM-based in-memory compute architecture rather than relying on standard high-bandwidth memory (HBM) packaging. Similarly, SambaNova is deploying custom dataflow technology with a three-tier memory architecture to maximize tokens per watt for advanced workflows.
Google’s Frozen v2 sits uniquely in this landscape — borrowing the hyper-efficiency of hardwired architectures like Taalas, but retaining just enough flexibility to remain viable across multiple product cycles.
The “Frozen” name reportedly comes from the idea of permanently etching part of Gemini’s design into the chip itself. According to The Information, Google’s engineers have spent years trying to find the right balance between efficiency and flexibility.
An earlier concept, led by Google DeepMind Chief Scientist Jeff Dean, reportedly would have embedded Gemini’s model weights directly into the silicon. That idea was eventually abandoned because it would have tied the hardware to a single version of the model, dramatically limiting its useful life as Gemini continued to evolve.
Frozen v2 reportedly takes a different approach. Rather than locking in the weights, the chip would hardwire parts of Gemini’s underlying design while still allowing the weights to be updated over time. That would let Google continue improving the model without needing to replace the hardware every time Gemini receives a new version.
Rather than locking in the weights, the chip would hardwire parts of Gemini’s underlying design while still allowing the weights to be updated over time.
By moving some of Gemini’s execution into the chip itself, Google could reduce some of the overhead that comes with running models on more general-purpose hardware. That could translate into lower latency, especially for applications that depend on near real-time responses.
Google’s reported goal of delivering six to ten times more tokens per watt is ultimately about efficiency, and those savings could eventually make their way to enterprise teams building on Gemini, through lower API costs or more available capacity. While Google hasn’t said how it would pass those gains along, reducing the cost of inference has become a priority across the industry.
The post Google just bet its inference future on a chip built for one model appeared first on The New Stack.
Imagine you just finished a two-hour meeting. You were wearing a small AI work companion that promised to capture the conversation, structure it, and let you ask questions about it later. Two hours after the meeting, you open the app and ask for the transcript. You wait. The spinner turns. The whole reason you bought the device was so you’d never have to remember a meeting again, and now the one thing it promised to do well — recall what was said — is the thing that’s making you wait.
This is the failure mode that fascinates me about AI hardware products, because it’s not a model problem. The transcription was perfect. The summarization was good. The thing that frayed was the part nobody markets: getting the right bytes off disk and onto the screen at the moment the user asks. That’s a data problem. And for a product whose entire promise is memory, a data problem is a product problem.
“For a product whose entire promise is memory, a data problem is a product problem.”
I want to walk through a real version of this because the team that hit it, Plaud, makes the most popular AI notetaker on the market, and the architecture that broke is the architecture almost every team in this category starts with. It feels reasonable at launch. It becomes a liability at scale. And the reasons why are worth understanding before you ship, not after.
It would be easy to tell this story as “they made a mistake.” They didn’t. The original architecture was a sensible set of decisions, and I want to be precise about that before I dissect what went wrong, because the lesson is in the gap between “reasonable” and “right at scale,” not in anyone’s competence.
Plaud’s product generates two very different kinds of data per recording. There’s structured metadata: who recorded it, when, how long, what tags, what state the processing pipeline is in. And there’s the unstructured payload: the audio file and its transcript, which can run tens of megabytes per session. The team did the textbook thing. They put the structured metadata in MySQL, where relational queries and transactions are cheap, and they put the large objects in S3, where storage is cheap and effectively infinite.
If you’ve built systems, you’ve made this exact call. Object storage for blobs, a relational database for everything you need to query. It’s in every architecture diagram. It’s the default. And for a long stretch of the product’s life, it worked fine.
The problem is not that the decision was wrong. The problem is that it contained a hidden assumption: that the metadata and the content could live in separate systems because they would never need to be consistent with each other in real time. For most applications, that assumption holds. For a product whose core interaction is “give me back exactly what I recorded, right now,” it doesn’t. And the gap between those two worlds is where everything started to fail.
In my article on RAG retrieval, I identified an anti-pattern I called the vector sidecar: the habit of standing up a separate vector database alongside your primary store, only to discover that the two systems can’t answer a single query together. The Plaud architecture is the same shape, applied to an AI hardware product. Call it the transcript sidecar.
The transcript sidecar is what you get whenever you split structured metadata from unstructured content across two systems with no shared transaction guarantee. The metadata says a recording exists, is complete, and is ready. The content lives elsewhere, reached via a separate call, with its own latency and failure modes. Nothing ties the two together. There is no transaction boundary that spans “the row that says this transcript is ready” and “the object that contains the transcript.”
This produces three distinct compounding problems.
The unifying insight is the one I keep coming back to throughout this whole series: any time you split data across systems that can fail independently, you have handed the consistency problem to the application layer, where it is not solved so much as managed, and where it compounds over time.
The most interesting failure in the Plaud story isn’t the latency. It’s the schema freeze, because it’s the one that teams least expect and feel most acutely.
With around 300 million rows, the team’s MySQL setup reached a point where schema changes such as adding a column or changing an index — the routine evolution of any growing product — could no longer be performed online without risk. DDL operations on a table that large, on that architecture, meant locking, replication strain, and the real possibility of downtime. So changes had to be batched into maintenance windows.
Sit with what that means. A maintenance window for a schema change is the time the database allows the product team to ship. The roadmap now bends around the database’s limitations. A feature that needs a new column will wait until the next window. The product’s cadence is set by the operational fragility of the data layer. Most teams don’t notice this inversion because it happens gradually, and then one day a product manager asks why a small change is a two-week project, and the answer is “the database.”
“A maintenance window for a schema change is the time the database allows the product team to ship. The roadmap now bends around the database’s limitations.”
The 300-million-row ceiling is a forcing function that teams hit later than they expect and should hit later than they do. Expect, because single-instance MySQL feels limitless right up until it doesn’t. Should, because the architecture that gets you to 300 million rows is rarely the architecture that takes you past it, and the migration is far more disruptive at 300 million rows than it would have been at 30 million. The ceiling is real; it is predictable, and most teams plan for it only after they’ve hit it.
Plaud’s resolution was to consolidate the metadata layer into a distributed SQL database (they moved to TiDB), which removed the single-instance ceiling and restored online schema changes. I’m going to use their numbers, but not as a product pitch. I want them as evidence for a claim about where architectural debt goes in a product like this.
| Before (dual-store-brain at scale) | After (consolidated metadata layer) |
| Throughput strained under concurrent load | ~10x QPS headroom |
| Tail latency variable under load | P95 under 10 ms |
| Schema changes require maintenance windows | Online DDL, no downtime |
| Single-instance ceiling at hundreds of millions of rows | Horizontal scale across multiple clusters |
QPS headroom and tail latency matter, but the first result I’d point to is the online DDL. Restoring the ability to change the schema without a maintenance window is what handed the roadmap back to the product team. The database stopped governing the release cadence. That’s not a performance win you can put on a benchmark chart, but it’s the one the product organization feels every week.
Also notice what the fix did not do: it did not attempt to store 30-megabyte audio files in the database. Large objects can still be stored in object storage. The point was never “put everything in one system for its own sake.” The point was that the metadata layer, the part that has to be consistent, queryable, and evolvable in real time, needed to actually behave like a database at scale, instead of becoming the fragile half of a split-brain architecture.
Here’s the claim the Plaud numbers are evidence for. When the database is the product, architectural debt does not accumulate in some back-office system that only the platform team feels. It accumulates in the product experience, where every user feels it.
This is the structural difference between an AI note-taker and, say, an internal analytics tool. For a product like this, every meaningful user interaction is a database operation. Recording creates rows and objects. Transcription updates state. Retrieval is a read. Editing is a write. Search is a query. There is no part of the product that isn’t, underneath, the database doing something. Architectural problems surface as product problems, one-for-one. A consistency gap becomes a transcript that is briefly missing. Slow retrieval reads as a product that is slow to remember. A frozen schema means features arrive late.
No user will ever file a bug report that says “your metadata store and your object store lack a shared transaction boundary.” Left unaddressed, this class of debt manifests as a product that feels less dependable precisely when someone depends on it. For a product whose entire value proposition is “trust me to remember,” the debt lands on the promise itself. This is why Plaud’s decision to re-architect when it did matter. The team fixed the data layer before the failure became noticeable to users.
“You can have the best transcription model in the category and still ship a product that feels unreliable, because reliability for this kind of product is a data architecture property, not a model property.”
This is why I say the database is the product. Not as a slogan. As a literal description of where the product’s quality is determined. You can have the best transcription model in the category and still ship a product that feels unreliable, because reliability for this kind of product is a data architecture property, not a model property.
AI hardware is having a moment. Note-takers, wearables, ambient recorders, pendants, badges — each built on the premise that the device will remember so you don’t have to. The category is growing, and the products are getting better at the parts everyone talks about: the models, the form factor, the battery life.
But underneath, almost all of them start with the same architecture Plaud started with, because it’s the reasonable default. Structured metadata in a single-instance relational database. Large content in object storage. No shared transaction boundary. A schema that’s easy to change at ten million rows and frozen at three hundred million. The trap is identical, and it’s waiting at the same place on the growth curve for every team in the category.
The teams that will do well are the ones that recognize this early, understand that, for a memory product, the metadata layer is not back-office plumbing but the spine of the user experience, and plan their data architecture for the scale they’re trying to reach rather than the scale they’re at. Plaud’s migration is the pattern worth studying before you hit 300 million rows, not after. The lesson is cheaper to learn from someone else’s 300 million rows than from your own.
When your database and your product are the same thing, you don’t get to treat the database as someone else’s problem. It is the product. Build it like one.
The post “The database is the product”: What breaks when memory devices scale appeared first on The New Stack.
Meta is preparing to manufacture its own AI chip for the first time. According to an internal memo, the company expects production of its proprietary processor, Iris, to begin in September.
After clearing bug testing in about six weeks, the chip — reported on by Reuters — is expected to take on some of the inference work currently running on third-party GPUs, giving Meta more control over how it builds and scales its AI infrastructure.
It’s unmistakable that this could be the company’s most important move yet toward in-house silicon for AI workloads.
But anyone can see this isn’t really about the hardware. It’s unmistakable that this could be the company’s most important move yet toward in-house silicon for AI workloads. The timing, as Meta is locked in an aggressive multi-billion-dollar infrastructure race, is critical. It’s clear that the company’s CEO, Mark Zuckerberg, wants to grow into the AI titan he believes the company can be, but it’s nearly impossible when the competition controls the core infrastructure.
Iris is designed for a specific job inside Meta’s AI infrastructure as custom silicon optimized for Meta’s heavy workloads. Iris expands Meta’s Meta Training and Inference Accelerators (MTIA) program, which is intended to move targeted AI inference workloads onto custom silicon.
The processor would handle workloads that drive content ranking, recommendations, and generative AI services across Meta’s family of applications, including Facebook, Instagram, and WhatsApp.
By shifting these high-volume inference tasks to custom silicon, Meta can lower data center costs while bypassing the traditional hardware supply bottleneck for its day-to-day operations.
Meta’s modular, rapid-fire approach to custom silicon is aggressive versus traditional industry timelines. The company plans to drop a new iteration roughly every six months through 2027.
Meta is working with Broadcom to design Iris, while TSMC will manufacture the chip. But custom silicon is only one piece of the equation. Scaling AI infrastructure also requires a steady supply of memory, storage, and networking components at a time when demand for AI hardware continues to strain global supply chains.
To support that expansion, Meta has also been securing key components across its supply chain. The company has signed long-term agreements for high-bandwidth memory from Samsung Electronics, flash storage from SanDisk, and fiber-optic networking equipment from Sumitomo Electric.
The strategy mirrors similar investments by other hyperscalers. Google continues to expand its TPU program, while Amazon has developed its Trainium and Inferentia processors.
The Iris rollout is one component of Meta’s broader AI infrastructure expansion. The company plans to bring roughly 7 gigawatts of computing capacity online this year, then double that to 14 gigawatts in 2027. At that scale, Meta’s AI infrastructure would consume more electricity than many small countries.
At that scale, Meta’s AI infrastructure would consume more electricity than many small countries.
And, scaling AI infrastructure at this level comes with an enormous price tag. Meta has projected 2026 capital expenditures of between $125 billion and $145 billion, making it one of the largest single-year infrastructure investors in corporate history
Meta just pulled off something rare, which is essentially convincing Wall Street that spending more money is actually a good thing.
Yet Meta just pulled off something rare: essentially convincing Wall Street that spending more money is actually a good thing. Following a trillion-dollar wipeout in tech market cap amid investor nervousness about the sheer scale of AI spending, Meta’s shares climbed roughly 8%.
With new MTIA chips planned roughly every six months through 2027, Meta is betting that vertically integrated AI hardware can deliver lower inference costs and better performance than relying exclusively on merchant silicon. By bringing chip design in-house and securing critical components across its supply chain, the company is slated to scale AI infrastructure with greater control over cost, deployment, and optimization.
The post Meta’s Iris push signals the next phase of AI infrastructure appeared first on The New Stack.
This sponsored article is brought to you by Wetour Robotics.
A field technician on a wind turbine, harness clipped, both hands on a wrench, needs to send a command to the diagnostic device hanging at her belt. A logistics worker on a loading dock, gloves on, eyes on the pallet, needs to redirect a connected lift. A person using an assistive mobility device on a crowded street wants to nudge it forward without taking out a phone or speaking aloud. None of these moments call for a smarter robot. They call for a smarter way to be heard by the machines that already exist.
The past three years of Physical AI have been a story of remarkable progress on the robot side of the loop. Companies like Boston Dynamics, Figure, and Unitree have advanced actuators, locomotion, and dexterity to a level that would have seemed implausible a decade ago. Google DeepMind’s Gemini Robotics has redefined what vision-language-action models can do in unstructured settings. The trajectory of the hardware and the foundation models is real, and it is accelerating.
But there is another side to this loop, and it has been treated as a solved problem for too long. The interface between humans and machines has defaulted, for 40 years, to three input modalities: screens, buttons, and voice. Each of those assumes the user can stop, look down, and translate intent into structured commands. That assumption breaks the moment the work moves into a real environment. On a turbine. On a dock. On a sidewalk. In any setting where hands are occupied, eyes are committed, or speaking is impractical, the conventional interface stack quietly fails.
Spatial Intent Fusion is the simultaneous processing of three streams of human-centered information, namely spatial position, visual context, and gestural intent: Your body is the interface.
The bottleneck on the human side of the loop is becoming as important as the one on the machine side. And solving it requires a different question. Not how do we make the robot more capable, but how do we let the human participate in the computing system as naturally as the robot already does.
Wetour Robotics is betting that the next architectural leap in Physical AI is not about making the robot more capable. It is about making the human a first-class node in the computing network, with the same kind of low-latency, high-fidelity participation that connected devices already enjoy.
Wetour Robotics’ engineers frame the problem this way: a wristband that recognizes a gesture is not enough. A camera that recognizes a scene is not enough. The information a human carries about what they are about to do is distributed across multiple channels, including where their body is in space, what their eyes are attending to, and what their muscles are preparing to do, and any single channel observed in isolation is ambiguous. Reconstructing intent reliably means fusing those channels at the operating system level, with latency low enough that the loop feels closed rather than mediated.
This approach has a name. Wetour Robotics calls it Spatial Intent Fusion: the simultaneous processing of three streams of human-centered information, namely spatial position, visual context, and gestural intent, fused into a single real-time command for any connected physical device. It is the technical implementation behind a simpler positioning statement the company uses externally: your body is the interface.
Orchestra is a portable intelligent hub running the operating system that handles sensor fusion, intent inference, command translation, and safety arbitration. The reference compute platform is NVIDIA Jetson Orin Nano Super, which provides enough on-device inference capacity to keep the entire control loop at the edge, with no cloud dependency on the critical path. Wetour Robotics
Orchestra is not a single device but a layered platform, designed from the start to be sensor-flexible and actuator-agnostic. The architecture decomposes into three perception layers and four coordination engines.
Orchestra itself is the local compute and orchestration core: a portable intelligent hub running the operating system that handles sensor fusion, intent inference, command translation, and safety arbitration. The reference compute platform is NVIDIA Jetson Orin Nano Super, which provides enough on-device inference capacity to keep the entire control loop at the edge, with no cloud dependency on the critical path. Edge inference is non-negotiable for this application. Full-chain latency from biosignal acquisition to actuator command is held under 100 milliseconds, the envelope inside which closed-loop control feels natural rather than laggy.
VisionLink handles visual and spatial perception. Cameras feed into vision models that identify objects, estimate distances, and track environmental context. VisionLink is designed not as a passive recognition layer but as a real-time command generator: its outputs feed directly into Orchestra OS to be fused with biosignal data.
Conductor is the biosignal pipeline. It ingests raw surface electromyographic (sEMG) data from a wrist-worn device, classifies temporal patterns into discrete gestures or continuous control signals, and outputs actuator commands. The technically interesting property of sEMG for this use case is that the signal precedes visible motion. Motor unit action potentials appear at the skin surface roughly 50 to 80 milliseconds before a finger completes the corresponding gesture. Wetour Robotics calls this property pre-motion intent sensing, and it is what allows Orchestra to anticipate user intent rather than react to it.
On top of the three perception layers, Orchestra OS runs four coordination engines. The Perception Engine ingests and normalizes raw sensor streams. The Intent Engine performs Spatial Intent Fusion across modalities, resolving what the user is trying to do given where they are, what they are looking at, and what their hand is signaling. The Orchestration Engine translates intent into device-specific command sequences for any connected actuator. The Safety Engine arbitrates conflicting commands, enforces operational envelopes, and gates execution against runtime safety conditions.
No system that bridges the human body and the digital world is finished. Three engineering challenges remain open, and the company addresses each with a deliberate trade-off rather than a claim of having fully solved it.
Baseline stability of sEMG under motion. In a stationary user, continuous gesture recognition from sEMG is reliable. Once the user is walking, climbing, or otherwise moving, motion artifacts and electrode drift degrade the signal in ways that are difficult to fully compensate for. Rather than overpromise on continuous control in dynamic settings, Orchestra defaults to a smaller set of robust discrete gestures in complex operating environments, and reserves continuous control modes for contexts where the signal-to-noise ratio supports them.
Miniaturization of edge AI compute. Running the Orchestra control loop entirely at the edge requires real on-device inference, which has historically meant trading off between compute capacity, battery life, and form factor. Wetour Robotics’ approach has been a compact carrier board paired with a thermal design and a battery module sized for all-day wearability. The result is a hub that travels with the user rather than tethering them to a desk, and that performs the full perception-to-actuation loop without offloading to the cloud.
Heterogeneity of third-party device protocols. The actuator side of the loop is a fragmented landscape. Different manufacturers expose different command interfaces, different communication stacks, and different safety conventions, and a Physical AI operating system has to integrate with all of them. Wetour Robotics uses an AI-agent layer to negotiate connection and protocol translation adaptively, so that Orchestra OS can ingest data from a wide range of devices, run them through neural network models that infer human intent, and emit the right command on the right protocol for the device on the other end.
The history of computing is a history of interface revolutions. Command lines gave way to graphical user interfaces, which gave way to touch, which gave way to voice. Each transition expanded who could participate in the system and what they could do with it. The next transition is not about a new screen or a new microphone. It is about treating the human body itself as a participant in the computing network, capable of contributing intent at the same speed and fidelity that any other connected node can.
The history of computing is a history of interface revolutions. The next transition is not about a new screen or a new microphone — it is about treating the human body itself as a participant in the computing network.
This path is not a competitor to the work being done on humanoid robots, foundation models for embodied AI, and dexterous manipulation. It is the missing complement to that work. The hardest open problem for humanoid systems is the data: every natural interaction between a human and the physical world is a potential training signal, and most of those interactions are currently invisible to any computing system. As more humans become first-class nodes in the loop, those interactions become observable, structured, and ultimately useful for training the next generation of embodied AI, including the humanoid robots being developed today.
In other words: putting the human back into the computing loop is not just about better interfaces for individual users. It is about generating the kind of grounded, in-the-wild human-machine interaction data that the broader Physical AI ecosystem will need to keep advancing. The robot side and the human side of the loop are not two competing futures. They are two halves of the same one.
That is what Wetour Robotics means when it says: Your body is the interface.
Learn more at wetourrobotics.com.
Login