onsemi Q3 Revenue $1.72B, AI Sales Surge 175%

Key Takeaways

• onsemi Q3 revenue hit $1.72 billion with AI data center sales reaching $180 million
• AI data center sales surged 175% year over year and 39% sequentially
• onsemi non-GAAP gross margin expanded to 41.2% for the fifth straight quarter
• Salesforce agreed to buy AI startup Cresta for about $2.7 billion
• CrowdStrike acquired Fiddler AI and Dreadnode for approximately $500 million
• China plans broad AI safety guardrails covering data, computing power and algorithms
• EU court adviser says Google Gemini may breach copyright using press articles
• Garland County Library adopts AI policy for employees
• Samsung expands AI training program to reach 100,000 teachers over five years
• New study finds terminal task benchmarks lack real depth

onsemi Earnings Surge on AI Data Center Demand and Margin Growth

onsemi reported strong earnings driven by AI data center growth and wider margins. Gross margin expanded for a fifth straight quarter. Earnings grew much faster than sales for the period. The company designs power and sensing semiconductors for automotive, industrial and AI data center uses.

onsemi Q3 Revenue Climbs 7% as AI Sales Jump 39%

onsemi reported third-quarter 2026 revenue of $1.72 billion and non-GAAP diluted EPS of $0.90. AI data center sales rose 39% sequentially and 175% year over year to $180 million. Non-GAAP gross margin reached 41.2% and the company guided fourth-quarter revenue to $1.715 billion to $1.815 billion. onsemi also introduced its Embedded Power Platform and is sampling 700-volt and 1,200-volt vertical gallium nitride devices.

Onsemi Q3 Profit Rises on AI Sales and M&A Expansion

Onsemi lifted third-quarter operating profit 43% year over year to $378.3 million on sales of $1.7223 billion. AI data center sales surged 175% year over year and 39% from the prior quarter. The company is acquiring a firm that owns power supply technology for AI Semiconductor and raised its 2030 revenue target for the data center business to over $3 billion. Free cash flow reached $489 million, up 31% from a year earlier.

China Plans Broad AI Safety Guardrails After Global Lapses

China said it will speed up legislation to govern AI development and use across the full life cycle of models. The rules will cover data, computing power, algorithms and key application scenarios. The government will review how the rules work and update outdated provisions. China also plans to watch AI's impact on jobs and provide protections for affected workers while encouraging open AI development through 2030.

China Prepares New Safety Standards for AI Firms

China is set to establish wide-ranging AI safety standards according to comments from its economic planner. The government will accelerate legislation covering data, computing power and algorithms. It will also regulate key application scenarios and set ethical guidelines for the full AI life cycle. China will monitor AI's effect on employment and provide worker protections. The official called for international cooperation on AI safety governance.

New study finds terminal task benchmarks lack real depth

Researchers studied 734 terminal tasks from six datasets and found that most synthetic tasks added work but not real depth. They defined depth as the number of sequential turns required to complete a task. Only 94 percent of solved synthetic tasks passed within three turns. A new method called DEPTH was proposed, creating 105 tasks that exceed all audited datasets. None of six current models passed half of these tasks, and fine-tuning Qwen3-8B improved performance on two benchmarks.

Garland County Library adopts AI policy for employees

The Garland County Library Board of Trustees voted to establish an AI policy for library employees. The policy does not require employees to use AI. It sets guidelines for workers who may choose to use AI as part of their workflow. The decision comes as artificial intelligence use grows across industries. The board met on Monday afternoon to approve the measure.

EU court adviser says Google Gemini may breach copyright

An adviser to the European Union's highest court said Google's Gemini chatbot may infringe copyright when using press articles. Advocate General Maciej Szpunar advised that reusing publications to generate answers can violate publishers' rights. This applies when chatbots bypass paywalls or copy the expression of a work. The case was brought by Hungarian website operator Like Company against Google. However, using copyrighted material to train large language models is permitted under text and data mining exceptions.

ADIPEC 2026 opens in Abu Dhabi with record attendance

ADIPEC 2026 opened in Abu Dhabi with roughly 239,000 participants from over 170 nations. The four-day event is hosted by ADNOC under the patronage of UAE President Sheikh Mohamed bin Zayed Al Nahyan. Dr Sultan Al Jaber outlined five priorities for energy security including storage, market access, and AI-enabled systems. OpenAI CEO Sam Altman spoke about AI's growing link to energy demand. The gathering addresses supply chains, investment, and infrastructure needs.

CrowdStrike acquires two AI security firms for $500 million

CrowdStrike acquired Fiddler AI and Dreadnode for approximately $500 million to strengthen its AI agent security offerings. Fiddler AI provides observability for monitoring AI agent behavior in production. Dreadnode focuses on automated security testing for AI systems to find vulnerabilities. The acquisitions come after CrowdStrike's July global outage that crashed millions of Windows systems. The move positions CrowdStrike in the emerging AI agent security market that has grown rapidly over the past two years.

Salesforce spends $2.7 billion on AI startup Cresta

Salesforce agreed to buy AI startup Cresta for about $2.7 billion. The deal is expected to close in early 2027. Cresta will work with Salesforce's Service Cloud to improve phone support. Salesforce has made five major acquisitions this year. The company's stock fell 1.32% and closed at $226.18.

Samsung expands AI training program for students

Samsung is expanding its Solve for Tomorrow program to teach students AI skills. Research shows 95% of U.S. high school students have used AI, but only 16% received formal training. The program will reach 100,000 teachers over five years. The national winner could receive up to $50,000 in technology. The competition deadline is November 23, 2026.

Pulse method improves AI memory for long tasks

Researchers created Pulse, a streaming personalization method for AI agents. It uses a lightweight memory agent to track user preferences during long tasks. Pulse scored 71.9% on pi-Bench and 46.4% on MCP-Persona. These were the highest scores among compared methods. The approach uses reinforcement learning to improve when the system intervenes.

Sources

• onsemi (NASDAQ:ON) AI Data Center Surge and Wider Margins Lift Earnings Well Above Sales
• onsemi Q3 Revenue Rises 7% as AI Sales Jump
• Onsemi boosts Q3 profit as AI data center sales surge 175% and M&A expands
• China plans broad AI guardrails after global lapses
• China Prepares to Impose New Safety Standards for AI Firms
• Is Your Long-Horizon Benchmark Really Deep? Measuring and Synthesizing Depth in Terminal Tasks
• Garland County Library board approves AI policy, change order for new branch
• Google AI chatbots use of press items may infringe copyright rules, EU court adviser says
• ADIPEC 2026 Opens in Abu Dhabi with Record 239,000 Participants to Address Energy Security, AI and Investment
• CrowdStrike Drops $500M on AI Security Duo as Agent Market Explodes
• The Selloff Didn’t Stop Salesforce From Spending $2.7 Billion on an AI Startup
• Samsung Finds 95% of Students Use AI, but Few Get Formal Training
• Pulse: Long-Horizon Tasks Need Streaming Personalization