Marvell's 51.2T Switching Revenue to Double in FY27

Marvell Technology is seeing strong demand for its 800G optical DSPs and rapid growth in its 1.6T optical DSP business during the second quarter of fiscal 2027. The company expects to more than double its scale-out switching revenues in fiscal 2027 through its 51.2T switching products. Management also expects 1.6T solutions adoption to accelerate in fiscal 2028 as data centers upgrade for growing AI workloads.

Marvell is pursuing scale-across and scale-up networking markets as hyperscalers build larger AI clusters across multiple data centers. The company's optical interconnect business continues to grow because of AI-driven demand, with optical DSPs playing a key role in high-speed optical communication systems.

Credo Technology Group and Oracle are collaborating on a project started in January 2026 to create a common software interface for optical transceivers and host systems. The project aims to standardize optical link health monitoring, telemetry sharing and automated fault responses. Credo contributed its ZeroFlap host specification as the technical foundation and is helping develop micro-emitter optics specifications expected to be approved in 2027.

Arista CEO Jayshree Ullal says intelligent networks can pay for themselves by finding and using idle AI accelerators. She explained that AI accelerators are often not used but networks can spot this waste and help recover that computing capacity, leading to major cost savings. Ullal also discussed how AI can help with anomaly detection, predictive analytics, network security and compliance.

OpenAI reported over 900 million weekly ChatGPT users and more than 50 million subscribers. President Greg Brockman said enterprise sales now make most of the company's revenue. OpenAI closed a funding round with $122 billion in committed capital at an $852 billion valuation, with Amazon, NVIDIA and SoftBank leading the round. The company also expanded its credit facility to about $4.7 billion.

OpenAI described a single AI workflow run that scanned the Linux kernel and found 2,308 potential vulnerabilities, proving 948 as truly dangerous and writing patches for each one. The company says the API cost for this one run would be $1,213,430 in tokens, arguing this is cheaper than paying bug bounties ranging from $20,000 to $70,000 per exploit.

DeepSeek raised CNY 100 million in funding on November 5, 2026, from Tencent Holdings and Contemporary Amperex Technology. The company issued convertible preferred stock in the transaction. Meanwhile, Leah launched Harness Builder, a self-service feature that lets customers build their own AI-powered tools from plain language descriptions.

Key Takeaways

  • <li>Marvell expects to more than double scale-out switching revenues in fiscal 2027 through 51.2T products</li>
  • <li>Credo and Oracle are collaborating on optical connectivity standards starting January 2026</li>
  • <li>Arista CEO says intelligent networks can recover wasted AI computing capacity and save costs</li>
  • <li>OpenAI has over 900 million weekly ChatGPT users and 50 million subscribers</li>
  • <li>OpenAI closed a funding round at $852 billion valuation with Amazon, NVIDIA and SoftBank leading</li>
  • <li>OpenAI says one AI cyber run cost $1,213,430 in API tokens but found 948 critical vulnerabilities</li>
  • <li>DeepSeek raised CNY 100 million from Tencent and CATL on November 5, 2026</li>
  • <li>Leah launched Harness Builder for customers to build AI-powered tools via plain language</li>
  • <li>Marvell expects 1.6T solutions adoption to accelerate in fiscal 2028</li>
  • <li>Arista CEO highlights idle AI accelerators as a major waste problem networks can address

Marvell sees strong growth in AI networking with new DSP and switch products

Marvell Technology reported strong demand for its 800G optical DSPs and rapid growth in its 1.6T optical DSP business during the second quarter of fiscal 2027. The company expects to more than double its scale-out switching revenues in fiscal 2027 through its 51.2T switching products. Marvell is also pursuing scale-across and scale-up networking markets as hyperscalers build larger AI clusters across multiple data centers. Management expects 1.6T solutions adoption to accelerate in fiscal 2028 as data centers upgrade for growing AI workloads.

Marvell's optical networking business expands on AI demand

Marvell's optical interconnect business continues to grow because of AI-driven demand. Optical DSPs are key parts in high-speed optical communication systems that send large amounts of data while keeping signals clear. The company is also growing its scale-out networking through high-capacity switches that connect servers and racks inside data centers. Marvell is targeting new scale-across and scale-up networking opportunities as hyperscalers build bigger AI clusters spanning multiple locations.

Credo and Oracle team up on AI optical connectivity standards

Credo Technology Group and Oracle are working together on a project started in January 2026 to create a common software interface for optical transceivers and host systems. The project aims to standardize optical link health monitoring, telemetry sharing and automated fault responses. Credo contributed its ZeroFlap host specification as the technical foundation. The company is also helping develop micro-emitter optics specifications expected to be approved in 2027 for next-generation AI data center connectivity.

Arista CEO says smart AI networks can save money by recovering wasted computing

Arista CEO Jayshree Ullal says intelligent networks can pay for themselves by finding and using idle AI accelerators. In an interview, she explained that AI accelerators are often not used but networks can spot this waste and help recover that computing capacity. This approach could lead to major cost savings for organizations. Ullal also stressed the need for network operators and AI developers to work together for better energy efficiency and performance.

Arista CEO Jayshree Ullal explains how AI can make networks more efficient

In an exclusive interview, Arista CEO Jayshree Ullal says intelligent networks can pay for themselves by recovering wasted AI computing capacity. She notes that idle AI accelerators are a big problem and that networks can detect and use them. Ullal also discusses how AI can help with anomaly detection, predictive analytics, network security and compliance. She believes AI will play a key role in making networks more efficient and self-sustaining in the future.

PersonaCodeBench tests coding agents on personal user needs

Researchers created PersonaCodeBench to test whether coding agents can meet personalized user requirements during multi-turn coding sessions. The benchmark turns user traits like skills and preferences into verifiable tasks. It evaluates agents on both correct code output and how well they follow persona-specific rules. Major frontier agents were tested and ranked on a comprehensive leaderboard.

OpenAI says one AI cyber run cost $1.2 million in API tokens

OpenAI described a single AI workflow run that scanned the Linux kernel and found 2,308 potential vulnerabilities. The system proved 948 of these as truly dangerous and had patches written for each one. OpenAI says the API cost for this one run would be $1,213,430 in tokens. The company argues this is cheaper than paying bug bounties ranging from $20,000 to $70,000 per exploit.

AI stock pick Ardent Health surges 19% in one day

An AI-powered stock pick from InvestingPro flagged Ardent Health in December 2025 when shares traded at just $8.79. The stock surged 19% in one morning and is up 79% since it was first selected. The model identified strong revenue growth and improving profitability as key signals. ProPicks AI uses machine learning to analyze over 60,000 stocks worldwide.

SEC focuses on AI cyber defenses and private credit valuations

Legal experts say firms should review their cybersecurity and private credit valuation policies due to increased SEC scrutiny. The SEC is examining how companies value private credit investments and protect against cyber threats. Firms that fail to comply risk facing fines and penalties. Experts emphasize the need for robust programs in both areas as threats grow more frequent and sophisticated.

AI political spending had mixed results in midterm elections

AI-funded super PACs had mixed effectiveness in the midterm elections. Guardrails Alliance PAC, which promotes AI rules, won all five races it spent on. Leading the Future helped elect one Democrat but lost support for Republican candidates. About 40 local ballot measures curbing data center development also passed across the country, surprising AI industry advocates.

OpenAI Enterprise Sales Grow as Brockman Addresses IPO Plans

OpenAI President Greg Brockman said enterprise sales now make most of the company's revenue. He spoke in Napa, California on November 4, 2026. OpenAI reported over 900 million weekly ChatGPT users and more than 50 million subscribers. The company closed a funding round with $122 billion in committed capital at an $852 billion valuation. Amazon, NVIDIA, and SoftBank led the round, and OpenAI expanded its credit facility to about $4.7 billion.

Leah Launches AI Tool-Building Feature for Customers

Contract lifecycle management provider Leah announced Harness Builder on Thursday. The self-service feature lets customers build their own AI-powered tools. It can generate a prototype from a plain language description. Human engineers can then refine the result. The tool is designed to help customers create custom applications with enterprise-grade controls.

DeepSeek Receives CNY 100 Million Funding from Tencent and CATL

Hangzhou DeepSeek Artificial Intelligence Co., Ltd. raised CNY 100 million in funding on November 5, 2026. Tencent Holdings Limited and Contemporary Amperex Technology Co., Limited provided the funding. The company issued convertible preferred stock in the transaction.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

Marvell Optical DSPs AI Workloads Data Centers Hyperscalers AI Accelerators Networking OpenAI ChatGPT Funding Vulnerability Detection AI Tools Cost Savings Collaboration Standards AI-driven Demand AI Cyber Run AI Networks AI Computing Capacity

Comments

Loading...