Fabrinet, a manufacturing partner for AI data centers, reported a 36% revenue jump to $4.6 billion in fiscal 2026, though Cisco accounts for over 36% of total revenue. The company maintains an 11% non-GAAP operating margin but carries a P/E of 30, reflecting high investor expectations tied to AI hardware demand.
In the AI chip and hardware space, Astera Labs posted revenue of $852.5 million with 115.1% growth and $219.1 million in net income, though its top three clients make up 86% of revenue. Arm generated nearly $4.9 billion in revenue with 22.8% growth and $904 million in net income. Cerebras Systems reached $510 million in revenue with 75.7% growth but posted a $75.7 million operating loss, while Innodata hit $251.7 million in revenue with positive free cash flow of $35.6 million and no debt.
Huawei Cloud announced the global rollout of enterprise AI products on September 18, 2026, with availability across Asia-Pacific, Europe, and the Americas. CEO Dr. Peter Zhou also unveiled Agentic Infra, designed to improve token efficiency and secure agent operations, along with AICS supporting over 40 days of stable training and Context Memory Storage offering petabyte-scale capacity. AgentArts launched with over 5,000 general MCP assets serving more than 100 customers.
AT&T, T-Mobile, and Verizon are each pursuing AI-RAN with different strategies. AT&T uses GPUs and CPUs and has a working system, T-Mobile relies on a custom-built GPU still in testing, and Verizon uses GPUs and FPGAs and is the least advanced. NVIDIA and Intel serve as vendors across these efforts, and AI-RAN is expected to improve network performance and reduce latency.
Anthropic, the maker of Claude, is preparing for a major IPO with annualized revenue expected to exceed $100 billion this year. CEO Dario Amodei has called for more guardrails on AI development, a stance that has sparked debate. Nvidia CEO Jensen Huang, in a CBS News interview, argued that AI should be developed as fast as possible, saying Nvidia will never ship unsafe products but should push forward. His comments came amid safety concerns raised by Anthropic researcher Jacob Coxon and CEO Dario Amodei.
At Salesforce's Dreamforce conference, business leaders said current AI models already deliver enough value, with discussions centering on deploying existing tools rather than waiting for newer ones. Separately, Mistral AI ran advertisements inside New York City's Grand Central subway station, captured on May 21, 2026, as the French AI company promotes its brand in a major public transit hub.
A new research framework called Evidence-Gated Matched-Pulse Transport helps agents detect and recover from hidden changes in system dynamics using a small number of interactions, offering recovered policies or safe fallback decisions. Meanwhile, five data science certificate programs for 2026 were compared, including MIT Professional Education's 15-week online program for $3,900 and UC Irvine's 9-to-15-month certificate, covering topics from machine learning to Agentic AI and RAG.
Key Takeaways
- Fabrinet's fiscal 2026 revenue rose 36% to $4.6 billion, driven by AI data center manufacturing, with Cisco over 36% of revenue and a P/E of 30.
- Astera Labs posted 115.1% revenue growth to $852.5 million and $219.1 million in net income, but its top three clients make up 86% of revenue.
- Arm generated nearly $4.9 billion in revenue with 22.8% growth and $904 million in net income, while Cerebras reached $510 million in revenue but posted a $75.7 million operating loss.
- Innodata achieved $251.7 million in revenue with 47.6% growth, positive free cash flow of $35.6 million, and no debt.
- Huawei Cloud launched enterprise AI products globally on September 18, 2026, with Agentic Infra, AICS supporting 40+ days of stable training, and AgentArts with over 5,000 MCP assets serving 100+ customers.
- AT&T, T-Mobile, and Verizon are pursuing AI-RAN with different hardware strategies, using vendors including NVIDIA and Intel, with AT&T the most advanced.
- Anthropic is preparing for an IPO with over $100 billion in expected annualized revenue, while CEO Dario Amodei calls for more AI safety guardrails.
- Nvidia CEO Jensen Huang says AI development should move as fast as possible, countering safety concerns raised by Anthropic's Amodei and researcher Jacob Coxon.
- Salesforce Dreamforce attendees said current AI models provide enough value, focusing on deploying existing tools rather than waiting for newer ones.
- Mistral AI advertised in New York's Grand Central subway station on May 21, 2026, as the French AI company expands brand visibility.
Fabrinet expands facilities to meet surging AI hardware demand
Fabrinet reported strong growth driven by its role as a manufacturing partner for AI data centers. Revenue rose 36% to $4.6 billion in fiscal 2026. Cisco accounted for over 36% of total revenue. The company maintains a non-GAAP operating margin near 11% but faces risks from customer concentration and a high P/E of 30.
Cerebras Systems vs Innodata: AI tech stock comparison
Cerebras Systems and Innodata both benefit from AI industry growth but follow different strategies. Cerebras reported FY 2025 revenue of $510 million with 75.7% growth but posted an operating loss of $75.7 million. Innodata reached $251.7 million in revenue with 47.6% growth and a net income of $32.2 million. Innodata had no debt and positive free cash flow of $35.6 million.
Astera Labs vs Arm: AI semiconductor stock comparison
Astera Labs and Arm both serve the AI chip market but operate very differently. Astera Labs reported revenue of $852.5 million with 115.1% growth and turned profitable with $219.1 million in net income. Arm generated nearly $4.9 billion in revenue with 22.8% growth and $904 million in net income. Astera Labs has heavy customer concentration with its top three clients making up 86% of revenue.
Huawei Cloud launches enterprise AI products globally
On September 18, 2026, Dr. Eric Yu, President of Huawei Cloud Business Unit, announced the global rollout of Huawei Cloud enterprise AI products in Shanghai. The products aim to provide comprehensive cloud solutions including AI, big data, and IoT services. They will be available in Asia-Pacific, Europe, and the Americas. The rollout supports Huawei Cloud's strategy to become a leading cloud service provider.
Huawei Cloud unveils Agentic Infra and enterprise agent platform
Dr. Peter Zhou, CEO of Huawei Cloud, announced Agentic Infra designed to make tokens more efficient and run agents securely. The latest AICS supports over 40 days of stable training and will be available in China on September 30 and outside China on November 30. The Context Memory Storage offers petabyte-scale memory space. Huawei Cloud also launched AgentArts with over 5,000 general MCP assets, serving more than 100 customers.
5 Data Science Certificate Programs to Watch in 2026
Five data science certificate programs are compared, ranging from foundational data science to advanced Agentic AI topics. MIT Professional Education offers a 15-week online program covering ML, deep learning, and multi-agent workflows for $3,900. UC Irvine's certificate spans 9 to 15 months and focuses on data engineering, statistics, and databases. MIT IDSS, the University of Arizona, and Old Dominion University also offer programs covering RAG, NLP, explainability, and machine learning. The programs vary in cost, duration, and emphasis on newer AI architectures.
New Framework Diagnoses Hidden Dynamics Changes in AI Systems
Researchers introduced Evidence-Gated Matched-Pulse Transport, a method for detecting and recovering from hidden changes in system dynamics. The framework helps an agent identify where local dynamics have changed using a small number of interactions. It then provides recovered policies or safe fallback decisions for candidate tasks. The approach was tested across diverse dormant-actuator benchmarks spanning multiple simulators. It aims to improve deployment decisions by balancing performance, confidence, and interaction cost.
AT&T, T-Mobile, and Verizon Take Different Paths on AI-RAN
The three major US carriers are all exploring AI-RAN but are using different hardware and approaches. AT&T uses GPUs and CPUs and is the most advanced with a working system. T-Mobile uses a custom-built GPU and is still in testing. Verizon uses GPUs and FPGAs and is the least advanced. The carriers also work with different vendors, including NVIDIA and Intel. AI-RAN is expected to improve network performance and reduce latency.
Anthropic Pursues IPO While CEO Warns on AI Safety
Anthropic, the maker of the Claude chatbot, is preparing for a major initial public offering. The company is expected to generate over $100 billion in annualized revenue this year. CEO Dario Amodei has called for more guardrails on AI development, sparking debate. Anthropic may release financial documents in the coming weeks, with shares potentially trading as early as November. The plans could still change due to market conditions.
AI Hardware Core Links Viewed as Opportunity Amid Bottlenecks
A market analysis frames bottlenecks in AI hardware as potential opportunities for investors. The article references stock ticker 3499.HK and a South East Asia ETF new product. It was published on Longbridge's platform but the specific page is no longer available. The original content discussed one-click layout strategies for AI hardware core links. Limited details remain about the specific investment recommendations.
Dreamforce attendees say current AI models are enough
At Salesforce's annual Dreamforce conference in San Francisco, business leaders said older AI models already provide enough value. The event featured debates about AI safety and whether model development should slow down. Attendees focused on how to use AI tools that are already available rather than waiting for newer ones.
Mistral AI advertisements appear in New York subway station
Advertisements for Mistral AI, a French artificial intelligence company, were displayed inside the Grand Central subway station in New York City. The photo shows the company promoting its brand in a major public transit hub. The image was captured on May 21, 2026.
Nvidia CEO says AI should be developed as fast as possible
Nvidia CEO Jensen Huang said tech companies should push forward with AI development as fast as they can. He spoke in an interview with CBS News about ignoring calls to slow down despite safety concerns. Huang said Nvidia will never ship unsafe products but expects the company to succeed and help America prosper. His comments came amid growing worry about AI speed, partly spurred by concerns raised by Anthropic researcher Jacob Coxon and Anthropic CEO Dario Amodei.
Sources
- Fabrinet's 2026 Outlook: Capital Expansion Powers Surging AI Hardware Demand
- Cerebras Systems vs. Innodata: Which AI-Driven Tech Stock Is a Better Buy in 2026?
- Astera Labs vs. Arm: Which AI-Driven Tech Stock Is a Better Buy in 2026?
- Huawei Cloud Rolls Out Enterprise AI Products Across the Board, Building an Open Agentic Cloud
- Huawei Cloud Rolls Out Enterprise AI Products Across the Board, Building an Open Agentic Cloud
- From Python and Statistics to Agentic AI: 5 Data Science Certificate Programs in 2026
- Diagnose, Recover, Certify: Task Readiness under Hidden Dynamics Changes
- AT&T, T-Mobile and Verizon plans diverge on AI-RAN
- Anthropic Pursues IPO Despite Its A.I. Safety Warnings
- "Bottlenecks" are opportunities: One-click layout of AI hardware core links with 3499.HK [South East Asia ETF New Produc…
- AI safety debate meets reality at Dreamforce as business leaders say last year's models are enough
- Advertisements for Mistral AI, the French artificial intelligence company, are displayed inside the Grand Central subway station in New York City
- Nvidia CEO Jensen Huang calls for AI to be developed "as fast as we can"
Comments
Please log in to post a comment.