AI hardware and compute news
Hardware & Compute follows the physical side of AI: GPUs and custom chips from Nvidia, AMD, Intel, Google, Amazon and the start-ups, the data centres being built to run them, and the power, water and supply chains they depend on. It is for engineers, infrastructure buyers, investors and energy watchers who need to understand the compute race behind every model launch. Stories cover chip announcements, export rules, capacity deals, data-centre sites and prices, with the numbers as reported and links to the companies' own figures where they publish them.
This week in Hardware & Compute
seven days to 31 October89 Hardware & Compute stories this week, from 70 publishers. The biggest: SpaceX says it closed $40B debt for GPUs arriving in 2027. Musk confirmed SpaceX closed $40 billion in debt financing for GPU purchases. This massive debt financing signals a significant expansion of AI compute capacity, potentially reshaping the market for GPU access and intensifying competition among AI infrastructure providers. Also this week: Nvidia secures 2.5x substrate capacity and 15x optical engine expansion by 2028; xAI leases 1.2M GPUs, earns $41.1B yearly from Microsoft, Anthropic and others and Google's Project Suncatcher satellite delivers first AI-generated haiku from orbit using Gemma 4 model on TPU.
Key topics
most mentioned this weekMost important
this week, by importance-
SpaceX says it closed $40B debt for GPUs arriving in 2027
Elon Musk confirmed on X that SpaceX has closed $40 billion in debt financing to purchase graphics processing units, with the hardware expected to arrive in the first half of 2027. The disclosure came in response to…
3 sourcesfool.comTom's Hardwarewccftech.com -
Nvidia secures 2.5x substrate capacity and 15x optical engine expansion by 2028
Nvidia announced multi-billion-dollar deals on October 21 to expand its AI accelerator supply chain, addressing critical bottlenecks. The company signed agreements with Ibiden, Unimicron, and Samsung Electro-Mechanics…
4 sourcesTom's Hardwarefinance.yahoo.comcryptobriefing.com -
xAI leases 1.2M GPUs, earns $41.1B yearly from Microsoft, Anthropic and others
xAI, SpaceX’s AI division, is positioning itself as an AI infrastructure provider by leasing out its GPU capacity. On October 1, 2026, reports emerged that Microsoft was in early discussions with xAI about leasing…
1 sourcenote.com -
Google's Project Suncatcher satellite delivers first AI-generated haiku from orbit using Gemma 4 model on TPU
Google announced on October 30, 2026, that its Project Suncatcher prototype satellite successfully generated and downlinked its first AI response from orbit: a haiku created by the onboard Gemma 4 model running on a…
5 sourcesThe Verge AIUnite.AITechCrunch AIinterestingengineering.com +1 more -
Amazon plans to sell Trainium AI chips to third‑party data centers starting late 2027
Amazon’s cloud unit reported its strongest growth in nearly five years, with AWS revenue jumping 39.3% to $46.0 billion and operating margin reaching a record 38.7%. The quarter’s operating income rose 61% to $28.1…
3 sourcesfinance.yahoo.comtechtimes.com247wallst.com
Latest Hardware & Compute stories
newest first-
Powerlattice raises $86M to cut AI chip power use by over 50%
Powerlattice, a Vancouver-based startup, secured $86 million in funding, according to an SEC filing. The round nearly triples the $31 million raised in November 2025, including a $25 million Series A led by Playground…
1 sourcegeekwire.com -
OpenAI CEO Sam Altman says home PC RAM is quite terrible for global allocation
OpenAI CEO Sam Altman argued in a Cheeky Pint interview that having DRAM in personal computers is inefficient from a global allocation perspective, stating it's 'quite terrible' for everyone to have limited memory…
6 sources HN 41The Verge AITom's Hardwarefrvr.comandroidheadlines.com +2 more -
Google’s cash burn drags down AI optics suppliers by 10-15%
Google’s $9.1 billion quarterly cash burn—double analyst expectations—sent shockwaves through AI optics suppliers this week. Applied Optoelectronics, Coherent, Lumentum, and Marvell all saw their stocks tumble 10% to…
1 source247wallst.com -
Microsoft’s capex shortfall drags down CoreWeave, Nebius shares
CoreWeave and other neocloud providers saw their stocks fall in extended trading after Microsoft reported quarterly capital expenditures of $45.2 billion, below the $50.7 billion average analyst estimate. The shortfall…
3 sources247wallst.comcnbc.comCNBC Technology -
Fujifilm plans fivefold increase in magnetic tape production for AI
According to sources cited by Nikkei Asia, Fujifilm is planning to expand its magnetic tape production capacity by five times. This move is intended to meet the growing storage demand driven by artificial intelligence…
3 sourcesArs Technica AITom's Hardwareasia.nikkei.com -
Foxconn plans new OpenAI collaboration and forecasts double-digit growth for 2027
Foxconn chairman Young Liu announced a new partnership with OpenAI on October 30, 2026, at the company's Hon Hai Tech Day in Taipei. Liu met OpenAI CEO Sam Altman earlier that day but did not disclose the scope,…
3 sourcesinvezz.comcryptobriefing.comwsau.com -
Taiwan Semiconductor Manufacturing could provide AI exposure despite bubble fears
Investors worried about an AI bubble are being urged to look at Taiwan Semiconductor Manufacturing (TSM). The company runs the world’s largest third‑party foundry, producing chips designed by Nvidia, AMD and other…
1 sourcefool.com -
Nvidia CEO sees Kazakhstan as potential AI hub for Europe
Nvidia CEO Jensen Huang visited Kazakhstan for the first time, calling it a potential AI computing hub for Europe during a fireside chat with Deputy Prime Minister Zhaslan Madiyev. Kazakhstan plans to nearly double its…
1 sourceyahoo.com -
PsiQuantum unveils 8x8 optical switch cutting AI data center losses to 1.2 dB
PsiQuantum, a quantum computing firm, introduced an optical switch for AI data centers that loses only 1.2 dB of light in lab tests—about a quarter less than typical industry switches. The 8x8 prototype, which uses…
1 sourceTom's Hardware -
Eclypsium partners with Nvidia to secure AI infrastructure against hardware-level attacks
Eclypsium announced a partnership with Nvidia on October 29, 2026, to integrate its AI Infrastructure Assurance platform with Nvidia's BlueField DPUs and Open Agent Safety Platform, aiming to build a chain of trust…
4 sourcestech.yahoo.comUnite.AIcryptobriefing.comfinance.yahoo.com -
Tom’s Hardware explains key AI concepts for local model users
Tom’s Hardware breaks down technical terms for users running AI models locally, focusing on text-based LLMs. The article defines core concepts like weights—the numerical values determining token relationships—and…
1 sourceTom's Hardware -
Runware plans factory to mass‑produce AI data center pods
Runware, an AI infrastructure firm, announced plans to build a factory that will assemble its Sonic Inference Pods on a production line, aiming for 30 megawatts of data‑center capacity per day by 2028. The company…
1 sourceTechCrunch AI -
CXMT plans to start HBM3E production with 20%‑30% yields
China’s memory maker CXMT announced it will begin manufacturing HBM3E, an advanced high‑bandwidth memory used in AI accelerators. The company expects early‑stage production yields of only 20% to 30%, a figure cited by…
2 sourcesreuters.com -
Tom's Hardware benchmarks Nvidia Vera CPU against AMD Turin and Intel Granite Rapids-AP
Tom's Hardware tested Nvidia's new 88-core Vera CPU, built on the custom Olympus core and Armv9.2-A architecture, against AMD EPYC 9755 and 9965 Turin chips and Intel Xeon 6980P Granite Rapids-AP in a dual-socket…
1 sourceTom's Hardware -
Finland probes Google over 816-acre forest clearance for AI data centers
A criminal investigation has begun into Google’s Finnish subsidiary, Tuike Finland Oy, over unauthorized deforestation at its Muhos data center project. Police confirmed five complaints and said the probe is backed by…
1 sourceTom's Hardware -
Bank of America says open-weight AI models boost Nvidia’s chip demand
Bank of America analyst Vivek Arya argues open-weight AI models—where weights are freely shared—do not cut costs for businesses but still require heavy computing power. The analyst notes these models now account for…
2 sourcesfool.comfinance.yahoo.com -
Foxconn aims for 50% of global AI server market in a couple years
Foxconn Chairman Young Liu told CNBC that the company expects its share of the global AI server market to reach 50% within a couple of years. He noted that the current market share is approximately 40%, with a…
1 sourceCNBC Technology -
Complete Dell workstation with 96GB NVIDIA GPU may cost less than graphics card alone
NVIDIA RTX PRO 6000 Blackwell graphics cards have become so expensive at US retailers that buyers are finding cheaper options by purchasing complete workstations. One example comes from Dell, where a Pro Max Tower T2…
1 sourcevideocardz.com -
NVIDIA invests $3 billion in Meta's equity offering to expand AI collaboration
NVIDIA will invest $3 billion in Meta's equity offering as part of an expanded collaboration on AI models and infrastructure. The partnership includes deploying NVIDIA's Vera Rubin systems, Vera CPUs, and Spectrum-X…
12 sourcesbloomberg.comcio.economictimes.indiatimes.comeconomictimes.indiatimes.comfinance.yahoo.com +4 more -
Analysis maps six AI chip camps: NVIDIA GPUs vs in-house ASICs
A new analysis compares the semiconductor strategies of six major tech players, highlighting the trade-offs between NVIDIA's high-power GPUs and custom in-house ASICs. The report notes that while NVIDIA's Blackwell…
1 sourcenote.com -
Iluvatar claims Tiangai 300 GPU matches Nvidia’s H200 in AI workloads
Iluvatar CoreX, a Chinese AI chipmaker based in Shanghai, says its new Tiangai 300 GPU matches or outperforms Nvidia’s H200 in real-world AI tasks. In a live demo for Chinese media on Wednesday, the company showed the…
1 sourcescmp.com -
OpenAI may purchase Google AI chips, report says
OpenAI is reportedly exploring a direct purchase of artificial‑intelligence chips designed by Google, according to a Tuesday report from The Information that cites a person familiar with the matter. The potential deal…
4 sourcesfinance.yahoo.comnote.comfool.com -
Alphabet introduces TPU 8t and 8i with lower AI chip costs
Alphabet launched its TPU 8t and 8i chips in October, claiming better performance-per-dollar than the previous TPU v7 (Ironwood). The company highlights lower token costs, positioning its offering as among the most…
1 sourceseekingalpha.com -
$10,000 in AMD or Nvidia by 2030 yields very different returns, model predicts
$10,000 invested in Advanced Micro Devices (AMD) or Nvidia (NVDA) could grow to vastly different amounts by 2030, according to a price-prediction model from Search: NVIDIA. The model, based on reference prices of…
3 sources247wallst.comfinance.yahoo.comfool.com -
AMD releases ROCm HRX Loom SDK for AMD CPUs, GPUs and NPUs
AMD has released the ROCm HRX Loom SDK, a specialized compiler and intermediate representation (IR) stack as part of its ROCm advancements. Loom compiles kernels and asynchronous device programs, links model programs,…
1 sourcephoronix.com -
Piq Energy report shows home batteries may ease data center grid overloads
Piq Energy’s Unlocking Flexibility Without Curtailment report, published October 29, explores how coordinating residential batteries could relieve transmission bottlenecks for large electricity users like AI data…
1 sourceUnite.AI -
Wall Street analysts predict Broadcom may outperform Nvidia in AI chip growth
Analysts at 24/7 Wall St. rate Broadcom stock a buy with a $416 price target, citing a 22% potential upside despite its 0.77% year-to-date gain. The company’s AI semiconductor revenue jumped 221% to $17 billion in the…
1 sourcefinance.yahoo.com -
Grace Blackwell’s SkyPilot boosts NVL72 GPU usage by 27%
NVIDIA’s NVL72 systems connect 72 GPUs per rack via NVLink, cutting communication bottlenecks for distributed AI workloads. However, maximizing their efficiency requires careful job placement, failure recovery, and…
1 source HN 46skypilot.ai -
Nvidia’s RTX Spark N1X laptops ship with up to 128GB memory for local AI
Nvidia’s RTX Spark N1X is the first Windows 11 laptop chip combining an Arm CPU and Blackwell GPU, designed for local AI workloads. Released on 16 October 2026, it powers laptops like the Microsoft Surface Laptop Ultra…
1 sourcei-scoop.eu -
PCBs and chip substrates may constrain consumer electronics supply, insiders say
Industry insiders report that printed circuit boards (PCBs) and chip substrates have become critical supply constraints for consumer electronics manufacturers. As AI demand surges, it is increasingly consuming…
2 sourcesTom's Hardwareasia.nikkei.com -
Cisco and NVIDIA introduce 400G optics for AI networking with fourfold bandwidth density
Cisco and NVIDIA announced a partnership at the Open Compute Project Global Summit 2026 to develop an open industry specification for 400G-per-lane pluggable optics. The new design aims to deliver four times the…
1 sourceSmall Business Trends -
KDnuggets lists five Docker containers for a home AI lab
KDnuggets outlines five Docker containers to build a reliable home AI lab, focusing on modularity and avoiding common pitfalls like lost models or forgotten ports. The containers cover model serving (Ollama), user…
1 sourceKDnuggets -
NVIDIA and Oracle expand OCI Superclusters with Vera Rubin NVL72 systems to multi-gigawatt scale
NVIDIA and Oracle announced plans to expand Oracle Cloud Infrastructure (OCI) Superclusters with NVIDIA Vera Rubin NVL72 systems to multi-gigawatt scale, supporting growing customer demand for AI training, inference…
9 sourcesfinance.yahoo.comthefastmode.comTom's Hardwarewccftech.com +4 more -
SWI Compute Debuts With Plans for NVIDIA AI Factories Across Europe
SWI Compute, a new AI infrastructure company within Euronext Amsterdam-listed SWI Group, emerged from stealth on October 26, 2026, announcing plans to build and operate NVIDIA AI factories starting in Europe, backed by…
3 sourcescryptobriefing.comUnite.AIaol.com -
Amazon adds Alexa AI to photo frame for Prime members
The Alexa Photo Frame is Amazon’s new digital picture frame with built-in Alexa AI, designed to automatically curate and display photos from Amazon Photos. The frame uses Alexa+ to filter images based on factors like…
1 sourcepetapixel.com -
FuriosaAI raises $620M Series D at a $2.9B valuation
Seoul-based AI chip startup FuriosaAI raised $620 million in a Series D funding round, Bloomberg reported. The transaction values the company at $2.9 billion, nearly quadrupling its previous $735 million valuation…
1 sourcebloomberg.com -
Oracle adds Google and Cerebras chips to Nvidia AI cloud deal
Oracle has expanded its partnership with Nvidia to deploy Vera Rubin AI servers at a multi-gigawatt scale, while simultaneously diversifying its hardware suppliers. The cloud provider announced it will begin offering…
25 sourcesfinance.yahoo.combetanews.comCNBC Technologyandroidheadlines.com +16 more -
Microsoft ends Copilot support on LG and Samsung smart TVs in 2026
Microsoft will stop supporting its Copilot AI assistant on LG and Samsung smart TVs during 2026, according to a notification seen on a webOS-powered LG TV. The feature, introduced in 2025, allowed users to chat with…
1 sourcetheverge.com -
Nvidia reportedly explores buying edge-AI chipmaker Ambarella for physical AI expansion
Nvidia may be considering acquiring Ambarella, a fabless semiconductor company specializing in low-power edge-AI chips for cars, cameras, and robotics, according to a post by Betaville and Goldman Sachs. The move would…
2 sourcesfinance.yahoo.com -
AWS releases Ray Train DLC for faster PyTorch training on EKS with EFA
AWS has published the Ray Train Deep Learning Container (DLC), a pre-built PyTorch training image that bundles CUDA, NVIDIA Collective Communications Library (NCCL), Elastic Fabric Adapter (EFA), and Ray libraries. The…
1 source primary sourceAWS Machine Learning Blog -
Google and Amazon launch PyTorch tools for their AI chips to compete with Nvidia
Google and Amazon are challenging Nvidia’s dominance in AI hardware by releasing open-source tools that enable developers to run PyTorch workloads natively on their custom chips. Google’s TorchTPU integrates with…
3 sourceslivemint.comyourstory.comcryptodaily.co.uk -
AI-driven chip shortages push HP, Asus, Acer to shift PC production back to China
HP, Asus, and Acer are reportedly relocating some PC manufacturing back to China due to soaring component costs, primarily driven by AI demand. Memory and storage chip prices surged 500% in 12 months, while PCBs and…
1 sourceTom's Hardware -
Ggml-org merges CUDA optimizations for MoE gate/up matmuls and GLU in MMQ
The open-source repository ggml-org/llama.cpp merged a pull request (31188) on October 26, 2026, adding CUDA optimizations. The changes fuse MoE gate/up matrix multiplications and GLU (Gated Linear Unit) operations…
1 source primary sourcegithub.com -
Nvidia older chips prices rise, signals AI demand
Brian Sozzi at Yahoo Finance noted that prices for older Nvidia chips are rising. The observation suggests that buyers are willing to pay more for existing silicon rather than wait for newer releases. This behavior…
1 source247wallst.com -
Nvidia signs supplier deals to triple AI chip capacity by 2028
Nvidia has announced agreements with six suppliers to expand manufacturing capacity for its AI systems, citing extraordinary demand and capacity constraints. The chipmaker stated that these deals are part of the…
1 sourcefinance.yahoo.com -
Core 4 Solutions lists 300-rack AI supercomputer with 9,600 Nvidia A100 GPUs on eBay for $59.7M
Core 4 Solutions is selling a 300-rack AI supercomputer on eBay for $59.7 million, including $50 for overnight shipping. The cluster contains 9,600 Nvidia A100 GPUs and 4,200 AMD EPYC Rome 64-core processors, totaling…
2 sources HN 5Tom's Hardwaretech.yahoo.com -
Nvidia and TSMC both hit records but stocks fall amid supply risks
Nvidia and TSMC posted record quarters in October, yet both stocks dropped: Nvidia fell 4.03% and TSMC lost 5.82%. Nvidia’s revenue hit $96.22 billion, up 105.8% year-over-year, with networking revenue at $40.31…
7 sources247wallst.comfinance.yahoo.comfool.comseekingalpha.com +1 more -
Nvidia’s China AI chip shipments drop 44% as export limits and production constraints tighten
Nvidia’s shipments of AI chips to China fell 44% year-over-year in the first half of 2026, per IDC, despite reported orders totaling over $35 billion for RTX PRO 5500 graphics cards. The decline reflects U.S. export…
4 sourcescoinpedia.orgfool.comfinance.yahoo.comtweaktown.com -
Nvidia proposes $2,000 annual fee per GPU for AI software
Nvidia CEO Jensen Huang suggested an annual software fee of $2,000 per GPU at GTC Berlin on October 22, but the company has not confirmed terms, scope, or timing. The fee would apply to software managing AI agents in…
8 sources HN 38geo.tvtechspot.comfinance.yahoo.comtechpowerup.com +3 more -
Moore Threads reports Q3 revenue up 1,774% to ~$240M
Chinese AI chipmaker Moore Threads Technology Co. reported that its third-quarter revenue surged 1,774% year-over-year to approximately $240 million. The company also posted a net profit of roughly $7.4 million for the…
1 sourcebloomberg.com -
Cowboy Space unveils Big Iron rocket to launch 20,000 AI data center satellites
Cowboy Space, a California-based company founded in 2024 by Robinhood co-founder Baiju Bhatt, announced its plan to deploy 20,000 AI data center satellites in low Earth orbit (LEO). The project, called Stampede, relies…
1 sourcespace.com -
NVIDIA reportedly halts RTX 5090 orders for board partners
NVIDIA has denied reports that the GeForce RTX 5090 has been discontinued, stating the card will continue shipping despite production constraints. However, new information from an unnamed board partner suggests the…
2 sourcesnotebookcheck.netvideocardz.com -
Mazda CX-5 adds Gemini AI assistant via free Google Play update
Mazda announced that its third-generation CX-5 now supports Google’s Gemini AI assistant. The update, available for free through Google Play, extends to cars in the US, Canada, Europe, and Japan. Owners can control…
1 sourcepaultan.org -
AMD’s MI455X offers 432 GB HBM4 vs Nvidia’s 288 GB in AI chips
AMD and Nvidia are competing in AI data centers with their latest GPUs: the Instinct MI455X (432 GB HBM4) and Vera Rubin NVL72 (288 GB HBM4). The memory gap—50% more in AMD’s chip—is the only concrete figure both…
1 sourceshattered.io -
Foxconn launches AI factory powered by Nvidia supercomputing cluster
Foxconn has deployed an Nvidia DGX SuperPOD based on the Blackwell architecture, using DGX GB200 systems, to power its new AI factory. The compute feeds FoxBrain, Foxconn's in-house large language model, which the…
3 sourcesintelligentliving.cocryptobriefing.comdatacenterfrontier.com
421 Hardware & Compute stories in all; this page shows the most important and the newest 55.