Anthropic signs $517bn compute deals, Qualcomm joins Amazon on AI chips, and more - Silicon Brief #87

September 14, 2026 - Silicon Brief
Hi there,

This edition covers the expanding infrastructure behind AI, from Anthropic's reported $517 billion in compute agreements to Oracle's 121% growth in cloud infrastructure revenue and new AI cloud contracts. Chip developments include Fujitsu's Monaka CPU, mass production of HyperAccel's Bertha inference accelerator, and a custom AI inference silicon partnership between Qualcomm Technologies and Amazon.

Power, cooling and networking remain central constraints. Massachusetts now requires large data centers to procure enough clean power to cover full electricity use, while Munters expands cooling production in Virginia, iPronics raises $125 million for optical networking, and ABB and Boston Consulting Group examine 800 VDC power distribution. The edition also tracks new AI data center capacity plans in Texas, Malaysia and Missouri, alongside edge, defense and photonic hardware developments.
DIMAAG-AI and Toshiba collaborate on SCiB storage for AI data centers
DIMAAG-AI and Toshiba International will develop energy storage systems for AI data centers using Toshiba's SCiB lithium titanium oxide battery cells. The systems will be integrated into DIMAAG-AI's ZettaWatt Power Platform, with capabilities subject to design, testing and validation. Read more
Fujitsu launches Monaka CPU for sovereign AI infrastructure
Fujitsu will begin global sales of its Japan designed Monaka CPU in November 2026, alongside a Monaka server for data centers, enterprises, academic institutions and high performance computing users in Japan and Europe. The company says the CPU provides higher AI inference throughput and power efficiency, while the server supports resource pooling and air cooling. Read more
Fermi Founder Group Seeks Independent Strategic Review
Founder affiliated shareholders holding about 22% of Fermi seek an independent investment bank review of options including a sale or majority recapitalization. The proposal also calls for governance and ownership limit changes before the October 30 annual meeting. Read more
DeepCtrls raises Series B+ led by CATL
DeepCtrls completed a Series B+ funding round led by CATL, with participation from Aramco Ventures, Taiping Innovation Investment, GF Xinde Investment Management and Fosun Capital. The company develops Physical AI for energy systems and is expanding into intelligent liquid cooling and computing energy coordination. Read more
Nscale appoints former OpenAI executive Fidji Simo to board
Nscale appoints former OpenAI executive Fidji Simo as an independent director while preparing for a potential IPO. Simo remains an adviser to OpenAI and serves on Shopify's board. Read more
Q.ANT appoints IBM veteran Joerg Behrend to lead photonic hardware
Q.ANT appoints former IBM senior manager Joerg Behrend as Vice President Hardware. He will oversee development of the company’s photonic processors, from chip architecture through production readiness. Read more
Aitech introduces rugged AI boards for defense and aerospace
Aitech introduces the 3U VPX C530 and SOSA aligned 3U VPX U C5300 rugged AI GPGPU boards for defense and aerospace systems. The boards support AI inference, sensor fusion, radar processing, electronic warfare and unmanned aircraft applications. Read more
Sophia Space and SLI set $300 million framework for 10 satellites
Sophia Space and aerospace leasing company SLI establish a nonbinding $300 million financing framework for a 10 satellite constellation. SLI would purchase and lease the satellites, which could launch from 2028 and provide orbital computing capacity for Earth observation, weather and security applications. Read more
Massachusetts requires large data centers to use 100% clean power
Massachusetts requires data centers with peak electricity demand above 25 MW to procure enough clean energy to cover their use. The framework also includes requirements on local power generation support, water use, pollution, community engagement and disclosure. Read more
Alif Semiconductor launches $49 edge AI StartKit boards
Alif Semiconductor launches the $49 SK-B1 and SK-E1C StartKit evaluation boards for its Balletto and Ensemble microcontrollers. The boards include an Arm Cortex-M55 CPU and Ethos-U55 neural processing unit for on device AI inference. Read more
iPronics raises $125 million for AI data center optical networking
iPronics raises $125 million in a Series B round co-led by Maverick Silicon and Light Street Capital, with NVIDIA among participating investors. The company will use the funding to expand production and deployments of its ONE optical networking platform, and open an office in Santa Clara. Read more
Spirox launches SP7000G TGV crack inspection system
Spirox introduced its SP7000G inspection system at the 2026 Advanced Packaging Technology Forum in Hsinchu, Taiwan. The system scans glass substrates for internal microcracks after metallization and ABF build up processes in Glass Core and TGV manufacturing. Read more
CargoSense acquires Adapt-IP for wireless 3D tracking
CargoSense acquires Adapt-IP to add wireless connectivity, real time asset tracking and 3D location data to its AI supply chain platform for data center construction and operations. Adapt-IP CEO Michael McNamara will lead CargoSense's hardware strategy. Read more
Veridue raises $4 million for renewable energy due diligence AI
Veridue, a London based AI company, raised $4 million in pre seed funding led by Episode 1 Ventures. Its platform supports due diligence for renewable energy and data centre transactions, and the company plans to expand the product with the funding. Read more
IREX unveils StreamVLM for public safety video analytics
IREX introduces StreamVLM, a cloud based vision language model for creating video detectors through plain language prompts. The platform supports real time alerts, investigations, collaboration and audit controls across existing camera networks. Read more
Dell'Oro reports 19% rise in campus switch prices
Dell'Oro Group reports that AI related component shortages reduced campus switch availability and increased average selling prices by 19% in the second quarter of 2026. The firm expects memory shortages to continue into 2027, while Cisco, Arista, Ubiquiti and Extreme gained market share. Read more
ADP expands AWS partnership for AI powered HCM
ADP expands its partnership with AWS, which will become its strategic cloud provider. ADP will use AWS cloud and AI services across its HR, payroll, compliance and workforce management platforms serving more than 1.1 million clients. Read more
Samsung Electronics partners with Mistral AI on semiconductor AI
Samsung Electronics and Mistral AI will develop AI systems for chip design, defect detection, equipment optimisation and manufacturing. Samsung Electronics will also lead Mistral AI's Series D funding round and take an equity stake. Read more
Telit Cinterion introduces on module AI inference SDK
Telit Cinterion introduces an edge AI SDK for running machine learning models directly on selected 4G, 5G RedCap and high performance 5G cellular modules. The Linux based SDK uses LiteRT, supports standard .tflite models and is planned for release in Q4 2026. Read more
Munters expands Virginia data center cooling campus
Munters has opened an expanded 565,000 square foot cooling campus in Daleville, Virginia. The site adds manufacturing, testing, training and chiller production capacity for data center, AI and high performance computing customers, and employs nearly 600 people. Read more
Besxar completes first semiconductor test flight with SpaceX
Besxar flew semiconductor wafer samples in two canisters aboard a SpaceX Falcon 9 mission on July 5, 2026. Both canisters were recovered intact, though one recorded a flight data anomaly, marking the first of 12 planned flights. Read more
Innventure removes DarkNX booking and forfeits earnout shares
Innventure removed a purchase order booking from DarkNX related to an Accelsius 300 MW data center project. The company says senior management and directors will forfeit associated earnout shares, while Accelsius 2026 revenue and cash flow targets are suspended alongside cost and funding changes. Read more
Bain report projects AI will use 0.7% of global energy
Bain & Company projects that AI will account for 0.7% of global energy use, despite substantially higher estimates from executives and consumers. The report also finds that many sustainable technologies have fallen short of expected deployment levels despite $17 trillion in investment over the past decade. Read more
OpenAI GPT-6 Astra becomes available on Amazon Bedrock
OpenAI makes GPT-6 Astra generally available through Amazon Bedrock. Organizations can access the model through Bedrock APIs for reasoning, software development, document analysis and AI agents, with AWS security and audit controls. Read more
Owl Autonomous Imaging selected for Pentagon led DUTCH thermal imaging award
Owl Autonomous Imaging will develop and demonstrate thermal imaging capabilities under the Pentagon led DUTCH initiative. The work uses its uncooled long wave infrared technology and KnightOwl HD platform for autonomous, defense, automotive and industrial applications. Read more
CNCF and SlashData report 400,000 cloud native AI developers in China
CNCF and SlashData estimate that China has 1.75 million cloud native developers, including about 400,000 working on cloud native AI. The report also links cloud native infrastructure to production AI workloads and distributed inference. Read more
Kawasaki Heavy Industries and EdgeCortix partner on AI aerial defense
Kawasaki Heavy Industries and EdgeCortix agree to develop AI enabled aerial defense systems from 2026 to 2028. The program will use EdgeCortix's chiplet hardware and MERA software for onboard sensor fusion, object recognition and threat assessment. Read more
Hyosung develops 22.9kV solid state transformer for U.S. AI power market
Hyosung Heavy Industries developed a 22.9kV, 1.05MVA solid state transformer designed to connect directly to urban distribution networks. The company plans to target power demand from U.S. AI and data centers, supported by production in Tennessee and a Pennsylvania joint venture with Quanta Services. Read more
AWS requires AI cost observability from MSP partners in 2027
DoiT says its Channel Management tool supports the AWS MSP Program Validation Checklist 8.0, which introduces requirements for generative AI and agent monitoring, AI cost optimization, and customer evidence of AI business results. The checklist takes effect on January 1, 2027, replacing the previous version. Read more
SEMIFIVE begins mass production of HyperAccel's Bertha AI accelerator
SEMIFIVE begins mass production of HyperAccel's Bertha accelerator for large language model inference, manufactured on Samsung Foundry's 4nm process. The chip exceeds 500 mm² and is SEMIFIVE's first large scale mass production program on the process. Read more
Google Cloud and Accenture form Gemini Enterprise unit
Google Cloud and Accenture form the Accenture Gemini Enterprise Business Group to help enterprises deploy Gemini Enterprise. Google Cloud will train up to 1,000 Accenture engineers to develop custom AI applications with customers. Read more
Galaxy receives conditional ERCOT classifications for 4.2 GW of Texas data centers
Galaxy Digital says five Texas data center projects received conditional classifications through ERCOT's Batch Zero large load process. The 4.2 GW capacity remains subject to an audit and a community impacts review. Read more
ABB and BCG report highlights 800 VDC for AI data centers
ABB and Boston Consulting Group identify 800 VDC distribution as an emerging power architecture for AI data centers. Their report cites growing AI demand, electrification and renewable energy, while noting standards, safety and workforce gaps that could slow wider DC adoption. Read more
NVIDIA CEO calls OpenAI Astra AGI, researcher disputes claim
NVIDIA CEO Jensen Huang describes OpenAI's Astra as artificial general intelligence, citing its training on more than 100,000 NVIDIA Grace Blackwell systems. Researcher Gary Marcus disputes the characterization, pointing to limited evidence and no agreed definition of AGI. Read more
NourNet launches Saudi hosted AI agent platform
NourNet introduced a beta AI Agent Platform hosted in Saudi data centers at LEAP 2026. The platform supports Arabic tuned and multilingual models, while the company also outlined data center expansion plans and new partner agreements. Read more
Firmus signs OpenAI deal for Malaysian data centre capacity
Firmus signs an agreement to provide OpenAI with computing capacity from two planned data centres in Malaysia. OpenAI will serve as an anchor customer, while the contract value and operating dates remain undisclosed. Read more
AidenAI launches AiDAP for Azure modernization
AidenAI launches AiDAP, an Azure native platform that uses specialized AI agents to support application migration and modernization. The company also introduces Aerie for cloud cost optimization and makes AiDAP available through the Microsoft commercial marketplace. Read more
DFI introduces Mini-ITX robotics platform with Intel Core Ultra options
DFI introduces the ARH171 and ARH173 Mini-ITX platforms for space constrained robotics and industrial automation. The boards support 16 Intel Core Ultra processors across the Meteor Lake and Arrow Lake generations, with Arrow Lake H configurations providing up to 99 platform TOPS. Read more
Island Roadhouse Plans 154 MW Nuclear Powered AI Campus in Missouri
Island Roadhouse Data Centers plans a Missouri AI data center campus supplied by behind the meter advanced nuclear reactors. Its first phase would provide up to 154 MW of IT capacity and is scheduled to begin commercial operations in 2034. Read more
Nvidia expects 70% revenue growth next year
Nvidia CEO Jensen Huang says the company expects revenue to rise about 70% year over year next year, from an estimated $400 billion to roughly $680 billion. He attributed the forecast to demand across AI models, data centers, suppliers and AI companies. Read more
Qualcomm and Amazon partner on custom AI inference chips
Qualcomm Technologies and Amazon will develop custom silicon for large scale AI data centers, initially for inference workloads. Their collaboration also covers optical connectivity supporting up to 1.6T bandwidth, while Qualcomm plans to use AWS AI infrastructure and Amazon Bedrock for chip design. Read more
Quantinuum finalizes over $100 million in US CHIPS funding
Quantinuum signs an agreement for more than $100 million in US CHIPS funding to expand domestic manufacturing capacity for trapped ion quantum computers. GlobalFoundries and Monarch Quantum will support development of ion traps, electronics, lasers and optical components. Read more
Cloud Native Computing Foundation graduates Karmada
The Cloud Native Computing Foundation graduates Karmada, an open source platform for managing applications across multiple Kubernetes clusters, clouds and regions. Its v1.19 release adds distributed AI training job scheduling, with planned support for multi cluster queuing, GPUs and other accelerators. Read more
Anthropic signs $517bn in compute agreements in 11 months
Anthropic signed compute agreements worth $517 billion over the past 11 months, according to DCD. The reported deals highlight the scale of infrastructure commitments supporting large AI model development. Read more
Lambda signs White House ratepayer protection pledge
Lambda signed the White House Ratepayer Protection Pledge on September 14, 2026. The AI cloud infrastructure company supports measures requiring data center developers to cover their electricity use and contribute to grid and power generation expansion. Read more
Knox Systems launches Space Force sponsored IL5 environment for defense AI
Knox Systems launches an Impact Level 5 cloud environment sponsored by the U.S. Space Force for commercial software and AI applications handling sensitive Department of War workloads. The service supports Controlled Unclassified Information and mission critical National Security Systems. Read more
Oracle Q1 Revenue Rises 30% as Cloud Infrastructure Grows 121%
Oracle reports first quarter fiscal 2027 revenue of $19.345 billion, up 30% from a year earlier, as cloud infrastructure revenue reached $7.388 billion. The company also signed more than $30 billion in AI cloud contracts and expanded data center capacity. Read more
Zettabyte and Raku Advanced Tech sign MOU for green AI data centers
Zettabyte and Raku Advanced Tech sign an agreement to develop renewable energy powered AI data centers in Taiwan and Thailand. Zettabyte will operate AI infrastructure through zSUITE, while Raku Advanced Tech will provide facilities, solar power, energy storage and energy management systems. Read more
Unisys Named Leader in NelsonHall AI Cloud Services Evaluation
NelsonHall named Unisys a Leader in its 2026 evaluation of AI enabled cloud infrastructure management services. The assessment cited Unisys capabilities in hybrid cloud automation, predictive analytics, self healing operations and the Unisys Intelligence Accelerator. Read more

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.

Also, consider following us on social media:

AI Chips & Datacenters AI Brief AI Brief (X)
Free newsletter

Silicon Brief

Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.

AI Token Webinar