TSMC Q3 Revenue Surges 50% to $46.7B on AI Chip Demand

Key Takeaways

• <ul><li>TSMC Q3 revenue hit NT$1.49 trillion ($46.7B), beating estimates and growing ~50% year over year</li><li>TSMC serves major customers including Nvidia, AMD, Apple, and benefits from AI chip demand</li><li>North American customers accounted for over 75% of TSMC sales in the first half of 2026</li><li>TSMC is expanding production in the US, including a $265B Arizona campus commitment</li><li>NVIDIA was named a key holding by Andrew Hill Investment Fund, with revenue expected to grow over 70%</li><li>NVIDIA launched the Open Agent Safety Platform to test and deploy AI agents safely</li><li>Anthropic released Claude Haiku 5.5, a small model with 1M token context and 75% lower pricing</li><li>Meta debuted AI tools for advertisers, including audience discovery and generative AI video features</li><li>Healthleap raised $38M to expand its AI screening platform for undiagnosed health conditions</li><li>AegisAI launched Proteus, a model that identified over 99% of 3,020 phishing attacks in benchmarks</li></ul>

TSMC Q3 revenue hits NT$1.49 trillion on AI chip demand

TSMC reported third-quarter revenue of about NT$1.49 trillion ($46.7 billion), beating analyst estimates. Revenue grew 50% from a year earlier due to strong demand for AI chips. The company benefits major customers like Nvidia, AMD, and Apple. TSMC is expanding production capacity and investing overseas, including in the United States.

TSMC posts record Q3 sales as AI demand stays strong

TSMC, the world's largest contract chipmaker, reported record third-quarter revenue of T$1.49 trillion. The results beat forecasts as customers like Nvidia and Apple continued placing orders. The strong quarter highlights how artificial intelligence demand continues to drive growth in the semiconductor industry.

TSMC Q3 revenue surpasses expectations on AI chip sales

TSMC's third-quarter revenue reached about NT$1.49 trillion ($46.7 billion), up 50% year over year and above the NT$1.46 trillion analyst estimate. September revenue rose 55% year over year to NT$511.86 billion. The results reflect strong demand for advanced chips used in AI accelerators and data centers. TSMC revenue for the first nine months of 2026 reached NT$3.90 trillion, up 41.1% from a year earlier.

TSMC Q3 sales exceed estimates amid AI chip boom

TSMC reported third-quarter revenue of about NT$1.49 trillion ($46.7 billion), surpassing market expectations. Revenue grew 50% from the prior year as demand for AI chips remained strong. Customers benefiting from TSMC's advanced chips include Nvidia, Advanced Micro Devices, and Apple. The company is also expanding overseas, including in the United States, to strengthen semiconductor supply chains.

TSMC third-quarter revenue rises 39% to record high

TSMC's third-quarter revenue reached approximately NT$1.49 trillion, a roughly 39% increase from a year earlier. September monthly revenue climbed 54.6% year over year to NT$511.86 billion but dipped 0.6% from August. Revenue for the first nine months of 2026 totaled NT$3.9 trillion, up 41.1% from the same period in 2025. TSMC stock fell 1.35% before the full results were made public.

TSMC Q3 revenue beats estimates on strong AI chip demand

TSMC reported third-quarter revenue of about NT$1.49 trillion, or $46.7 billion, beating analyst estimates of NT$1.46 trillion. Revenue grew 50% compared to the same period last year. The strong results were driven by continued demand for advanced AI chips. TSMC serves major customers including Nvidia, AMD and Apple. The company is also expanding production capacity in the United States.

TSMC revenue rises sharply due to growing AI hardware demand

TSMC saw September revenue rise 54.6% year on year to 511.86 billion New Taiwan dollars, roughly 14.3 billion euros. Despite this strong annual growth, revenue fell slightly by 0.6% compared to August. Total third-quarter revenue is estimated at 41.7 billion euros. TSMC is a key supplier to tech giants like Apple and Nvidia. Investors are waiting for the full quarterly report on October 15 for more details on profit margins.

TSMC quarterly revenue jumps 51% as AI demand stays strong

TSMC reported a 51% rise in quarterly revenue totaling NT$1.49 trillion, about $46.7 billion, for the three months ending in September. This beat the analyst expectation of NT$1.46 trillion. The strong growth supports the outlook for AI demand worldwide. TSMC is considering further expansion in Texas after pledging $265 billion for an Arizona campus. North American customers accounted for more than 75% of the company's sales in the first half of the year.

Utah startup uses AI to prescribe acne medication

A Utah healthcare startup launched a pilot program where AI decides if patients need an acne medication prescription. The program will last one year, with the first 100 prescriptions overseen by a board-certified physician. If successful, the AI will eventually treat up to 400 patients with a physician reviewing decisions weekly. The CEO says human physicians should always be involved in patient care. The program represents a limited and specific use of AI in medicine.

Healthleap raises $38M for AI screening platform

Healthleap raised $38 million in seed and series A funding to expand its AI screening platform. The platform uses AI to identify patients at risk of undiagnosed conditions like malnutrition. It works with electronic health records and can be integrated into existing healthcare workflows. The funding will help expand the platform to cover cardiovascular disease, diabetes and cancer. Sequoia Capital led the funding round with participation from Hummingbird Ventures.

NVIDIA remains a top AI investment pick says fund

Andrew Hill Investment Fund named NVIDIA (NVDA) as a key holding in its Q3 2026 investor letter. The fund praised NVIDIA as a leader in AI infrastructure and computing hardware. Revenue is expected to grow over 70% in the next year. The stock is seen as attractively valued compared to chipmaking rivals.

NVIDIA launches AI agent safety platform

NVIDIA (NVDA) introduced the Open Agent Safety Platform in late September 2026. The tool helps developers test and deploy AI agents safely. It can quarantine rogue agents in milliseconds if they try to leave secure environments. This follows incidents where autonomous AI agents went rogue at Hugging Face. The platform strengthens confidence in NVIDIA's AI business.

Study tests stacked AI agent safety gates

Researchers tested stacked runtime gates on 1,119 labeled AI agent actions from three datasets. The setup included one deterministic rule layer and four LLM judges. They found that combining layers does not always multiply safety effectiveness. Solo accuracy does not predict what each added layer contributes. The study highlights challenges in measuring how well these stacked gates catch agent errors.

Anthropic releases Claude Haiku 5.5 small AI model

Anthropic released Claude Haiku 5.5, a small model with a 1 million token context window. It is priced at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100K tokens. The model targets high-volume tasks like summaries, classification, and subagent work. It scored 72.4% on OSWorld 2.1, up from 15.7% for Haiku 4.5. Pricing is about 75% cheaper on average than the previous version.

Meta launches AI tools for advertisers

Meta debuted new AI tools to help advertisers find and reach customers. The tools include AI-enabled audience discovery features like Customer Lifecycle Strategy. Marketers can set campaigns to target new customers and exclude existing ones. Meta also expanded generative AI video tools for business use. The company is partnering with retailers to enable purchases directly through AI conversations.

AegisAI launches Proteus model for email security

AegisAI released Proteus, a foundational model that investigates every email like a security analyst before it reaches the inbox. In benchmarks, Proteus identified more than 99% of 3,020 human-reviewed phishing attacks, outperforming leading general-purpose and open-weight models. The launch addresses the rise of AI-generated spear phishing, which AegisAI says could make most attacks AI-written by 2028.

Lawmakers warn Google could expose Spirit Airlines data

More than 100 members of Congress warned Google about a proposed $10 million deal to buy Spirit Airlines data for AI training. The data could include about 100 million emails, 500 million Teams messages, and sensitive employee records. Lawmakers said deidentification may not fully protect privacy and urged limits on data use and independent confidentiality review.

Study finds personal memory errors hurt AI language agents

A research paper found that AI language agents often fail before generating answers because their stored memory is stale, misattributed, or late. On a controlled benchmark, 70.3% of prompts exposed outdated values when updates were not resolved. The study argues that agent memory should be evaluated for accuracy and timing before generating responses.

Dell AI accelerators aim to cut high token costs

Dell introduced Pro Precision with GB10 and GB300 AI accelerators designed for local agentic AI use. The devices let organizations run open-weight models on deskside hardware instead of paying variable cloud token fees. Dell says this approach can reduce costs and keep sensitive data within local infrastructure.

Sources

• TSMC Q3 sales beat estimates on surging AI chip demand
• TSMC’s Record Q3 Sales Show AI Demand Is Still Hot
• TSMC Q3 sales beat estimates on surging AI chip demand
• TSMC Q3 sales beat estimates on surging AI chip demand
• TSMC's third-quarter revenue rose 39% to a record on AI chip demand
• TSMC Q3 sales beat estimates on surging AI chip demand
• Chip manufacturer TSMC benefits from growing demand for AI hardware and sees revenue rise sharply
• TSMC’s Quarterly Revenue Jumps 51% After AI Demand Holds Up
• Utah Now Allows AI to Scan Your Face and Prescribe Acne Medication
• Healthleap banks $38M for AI screening platform that catches undiagnosed conditions
• NVIDIA (NVDA) Is Still One of the Most Compelling AI Plays: Here’s Why
• From Agentic AI Concerns to Solutions: How NVIDIA Corporation’s (NVDA) Safety Platform Strengthens Bull Case
• Evaluate the Stack, Not the Layer: Do Deterministic and LLM Gates for Agent Actions Fail Independently?
• Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million Input Tokens
• Meta Debuts AI Tools to Help Advertisers Find Customers
• AegisAI launches Proteus, the first foundational model for email security that investigates every email like a security analyst, before it reaches the inbox
• Lawmakers warn Google could expose Spirit Airlines data in $10 million AI training deal
• Stale, Misattributed, or Late: Where Personal Memory Fails Before Generation
• $3k a day on tokens? Dell’s deskside agentic AI Accelerators will save you money on AI development