SK hynix presented several next-generation memory solutions for AI servers at OCP Global Summit 2026, held October 12–15 in San Jose.
Modal announced its annualized revenue has surpassed $1 billion and it raised a $750 million Series D led by Accel at a $15.75 billion valuation. Usage has tripled since its Series C, growing more than 15x year over year. Sandboxes, launched in 2023, alone has reached $300 million in annualized revenue, with nearly 3 billion sandboxes launched cumulatively and the ability to scale to millions running concurrently.
At the AI in Science Summit in Dublin, NVIDIA joined the European Commission’s AI in Science Enablers initiative (part of RAISE), providing European researchers with tools, compute, and training.
Ayar Labs and GUC jointly ran thermal co-simulations on a CoWoS-L package integrating 8 TeraPHY optical engines, 2 XPUs, and 8 HBM4 modules; the package draws over 3kW of power and provides a full-duplex optical interface of more than 100 Tbps.
In a SEMICON West interview, Synopsys CEO Sassine Ghazi said the company has deployed more than 40 task agents, and some customers can hand a spec to the Synopsys Autonomous Agent to automatically generate and verify RTL — a process that used to take multiple engineers several months — though fully autonomous chip design remains out of reach today.
Agentic AI requires extensive if-then-else decision orchestration, work handled by CPUs; industry estimates suggest the CPU:GPU ratio could rise from roughly 1:4 in typical LLM scenarios to nearly 1:1, pushing up both CPU demand and prices.
Semiconductor Engineering examines how open-weight AI models affect chip architecture, noting the focus is shifting from peak TOPS to memory bandwidth and capacity, sustained throughput, numerical format flexibility, and deployment security.
AI training and inference have different demands for memory, compute, and interconnect, pushing AI infrastructure toward heterogeneity. UCIe provides die-to-die standardized connectivity, and with chiplet solutions lets compute, I/O, and memory interfaces be flexibly integrated in one package. At the system level, PCIe, CXL, UALink, and Ethernet must be combined with technologies like HBM4E and LPDDR6, using open standards and interoperability validation to leave room for architectural evolution.
At OCP Global Summit 2026, AMD CEO Lisa Su called this the year of inference, estimating inference accounts for about 60% of the market, and raised CPU TAM forecasts three times due to agent workloads, expecting the market to exceed $200 billion by 2030.
The Nordics are building a leading European technology ecosystem with NVIDIA technology, expected to deploy 1 gigawatts of AI cloud compute by the end of 2027.
Nvidia CEO Jensen Huang announced in Paris $2 billion in new investments across Europe over the coming week, and disclosed stakes in French startups including Doctolib, Genesis AI, Pigment and Entalpic, advocating that Europe use open-weight models to close the AI gap.
Semiconductor Engineering spoke with experts from Synopsys, Silvaco, Vinci, Keysight EDA and Siemens EDA at the Design Automation Conference about the thermal management challenges posed by chip stacking.
Multiple cities across the EMEA region, along with software partners, are using NVIDIA accelerated computing to build physical AI agents and digital twins for urban planning, predictive maintenance, and public services. France's Côte-d'Or department deployed a sovereign digital twin based on Nemotron and Omniverse across more than 500 public buildings, cutting document retrieval time from weeks to minutes, lowering operating costs by 10%, with a target of 30% at scale.
NVIDIA announced $250 million in compute resources over five years to support UK agent safety research, $50 million per year in accelerated computing for the University of Oxford, UCL, and Imperial College London through the BOLD program, and joined the Earthshot Prize as a founding Global Alliance Partner.
An SK hynix research team published a paper at the 2025 IEEE International Electron Devices Meeting (IEDM) proposing co-optimization of cell, non-cell, and page schemes to improve the energy efficiency of analog compute-in-memory (A-CiM) based on 3D FeNAND.
At TechCrunch Disrupt 2026, Cerebras co-founder and CEO Andrew Feldman criticized the AI industry's doomsday narrative, saying some AI leaders "maybe shouldn't speak publicly," and argued more should be said about AI's potential benefits, such as ensuring grandchildren "never meet a person who died of cancer."
UC Berkeley professor emeritus and machine learning pioneer Michael I. Jordan, in an interview with Libération, dismissed AI doom rhetoric, saying "superintelligence" is a science fiction term and that most related debates belong in the realm of sci-fi.
Engineer Rama Afullo, whose proposals for orbital data centers were rejected by both Google and SpaceX, founded Satlyt, which develops computing software for satellites and has raised an $8 million seed round.
AMD released the replay of its OCP 2026 keynote themed "Building the Next Era of AI Infrastructure." Downloads on the page require first accepting AMD's trademark usage terms, and the AMD trademarks may only be used to promote products containing AMD processors.
NVIDIA announced new investments in French startups spanning aerospace, finance, public services and research. Mistral released the private-preview open model Mistral Large 4; Bull will produce NVIDIA Vera Rubin NVL72 servers in Europe, planned for 2028; and the Altair-1 satellite carrying a Jetson AGX Orin is now in orbit.
At OCP 2026, AMD CEO Lisa Su highlighted open infrastructure for the future of AI. At the event, AMD emphasized supporting next-generation AI compute buildout with open architectures.
At the OCP Global Summit, Intel detailed its GPU D-series (codenamed Crescent Island), built on the Xe3P architecture with up to 480 GB LPDDR5X memory, 1.5 TB/s memory bandwidth and 1.3 petaflops of FP8 compute at 350 watts; a single card supports 300-billion-parameter models and an 8-card system supports over 1 trillion parameters.
Ukrainian drones struck Yandex data centers in Kaluga and Vladimir regions on October 9 and October 11, with the latter hosting Galushkin, one of its three supercomputers (ranked 134 on the Top500 with 20.64 petaflops), causing widespread disruption to Russian digital services.
At the OCP Global Summit, NVIDIA announced it is contributing Vera Rubin rack and compute tray designs to the OCP through its MGX open architecture. The MGX 1.2 spec adds rack-level blind-mate liquid cooling, a 5000A liquid-cooled busbar, and support for rack deployments of up to 250kW.
Why it matters: The piece collects multiple open specifications such as MGX 1.2 and Open Data Center Specification v0.7, along with accompanying software, giving readers a view into the open foundations of AI factories.
AMD launched the Helion agentic AI rack-scale solution, an open blueprint for agentic AI infrastructure designed based on OCP specifications.
Marvell released the second-generation Structera X 3406 and 3506 CXL 3.2 memory expanders, the industry's first six-channel CXL memory expansion products, with 50% more channels than the previous generation, supporting up to 9TB of DDR4 or 6TB of DDR5 per controller and delivering up to 256GB/s of host bandwidth over PCIe 6.0 x16.
Microsoft unveiled its third-generation in-house AI chip Maia 300, positioned as the "primary AI inference system" for its products and services, with production planned from 2027 and the Azure Maia family scaling to multi-gigawatt deployments starting in 2027.
AMD has launched the AMD Helion agentic AI rack, an open blueprint for agentic AI infrastructure designed to Open Compute Project specifications and powered by AMD compute.
In Q3 2026, 80 startups raised more than $8.5 billion combined, with two inference companies each raising $1 billion and nearly $7.5 billion going to 22 companies with rounds of $100 million or more. Investor focus centered on interconnect technology, networking chips and photonics, including co-packaged optics, millimeter-wave plastic waveguides, ferroelectric glass modulators and coherent DSPs for AI optics, while some companies explored replacing constrained HBM with ordinary DRAM or on-chip SRAM.
Amazon, following Microsoft's earlier move, announced it will no longer use non-disclosure agreements (NDAs) in negotiations with local governments over data centers, in a statement made in a blog post arguing that data centers benefit communities.
a16z says data from Ornn, Silicon Data, and Bloomberg through August 2026 shows a textbook Jevons paradox in the AI market: token prices keep falling, yet H100 GPU rental prices hold steady or rise, as cheaper tokens fuel AI agents and automation with usage growing faster than prices drop.
A Ukrainian drone strike destroyed an AI data center belonging to Yandex, known as "Russia's Google"; per Reuters, other affected Russian services include a dating app, a real estate aggregator, a self-publishing platform, and a Russian Premier League football club's website. The report notes data centers are becoming war targets in 2026, after Iranian drones destroyed multiple Amazon data centers in Bahrain and the UAE following the February US-Israel strikes on Iran.
Arista Networks (NYSE: ANET) announced an open Ethernet rack-scale portfolio on October 7, 2026, in collaboration with AMD, Arm, Broadcom, d-Matrix, Meta, Microsoft, Qualcomm and others, extending Ethernet into the scale-up domain based on its ESUN initiative.
AWS has released a multi-tenant reference architecture based on Amazon SageMaker HyperPod and EKS, letting multiple teams share a single GPU cluster while maintaining isolation and fair resource allocation.
Ambrosia Energy CEO Ben Longmier and Bill Thayer, head of data center solutions at Bloom Energy, will sit down for a conversation titled “Where the AI infrastructure boom is creating winners” on the Smart Systems Stage at TechCrunch Disrupt 2026.
Semiconductor Engineering gathered experts from Teradyne, Advantest, ISE Labs and Keysight to discuss CPO/NPO test challenges.
Semiconductor fabs are using real-time equipment sensor data and AI/ML inference models to move yield anomaly detection inside process modules, accelerating yield learning and improving advanced process control (APC). Equipment sensor sampling has risen from about once per second to once every few tens of milliseconds, data types have grown at least 10x compared with a decade ago, and some fabs generate hundreds of petabytes per year. How to process data in sub-second timescales while preserving context and connectivity is the main data management challenge shift left brings.
Advantest's Dong Ong talks with ISIG co-founder and CEO Salah Nasri about competitive advantage in the semiconductor industry over the next decade. In the AI era, industry value is shifting from transistor process nodes and single chips to advanced packaging, system integration, software and platform ecosystems; performance depends on how efficiently an entire system moves data, manages energy and coordinates compute, and future winners will be vendors that combine technology, platforms and partnerships into a value system.
NVIDIA and Microsoft announced at the Windows AI and Surface event in San Francisco that they are jointly building agent software and hardware for Windows PCs.
Why it matters: The article details the specs and launch timeline for RTX Spark and DGX Station for Windows, helping readers assess how realistic running large models locally with resident agents will be.
At the Design Automation Conference, Semiconductor Engineering spoke with experts from Synopsys, Silvaco, Vinci, Keysight EDA, and Siemens EDA about heat's impact on photonics.