Apple's 2029 AI Server: M8 Ultra Chips, NVLink Fusion

Key Takeaways

• Apple is developing an AI server with M8 Ultra chips, set for a 2029 launch and aimed at enterprise and government customers.
• Nvidia's Vera Rubin AI Accelerator promises 2X profit per watt over Blackwell, targeting hyperscaler energy efficiency.
• Microsoft's AI education framework prioritizes safety, privacy, and educator control, citing gains from Microsoft 365 Copilot.
• Goldman Sachs investment banking chief Kim Posnett advises junior analysts to use AI tools thoughtfully and in balance.
• Mitsubishi Heavy Industries is investing $64 million in AI startup Preferred Networks for defense and security technology.
• Siemens stock may be 34 percent undervalued at around $90 per share, despite a 106.8 percent rise over three years.
• A new TAISE framework cuts catastrophe risk modeling costs by roughly ten times using AI weather models.
• TEDCO invested $250,000 in enterprise AI startup Syntes to help businesses optimize workflows with machine learning.
• Apple's AI server will extend the company beyond consumer hardware into enterprise AI infrastructure.
• Researchers developed a coverage-aware virtual IMU augmentation framework that improves human activity recognition on public benchmarks.

Apple builds AI server using M-series Ultra chips

Apple is developing an AI server that uses future M8 Ultra chips. The server could launch in 2029 and would be Apple's first new server in nearly 20 years. It will come in two versions with two or four M8 Ultra chips. The project is aimed at AI developers and enterprise customers.

Apple develops secret enterprise AI server project

Apple is working on a confidential enterprise AI server for third-party developers and government clients. The server will use future M8 Ultra chips and Nvidia NVLink Fusion networking technology. This would extend Apple beyond consumer hardware like iPhone and Mac into enterprise AI infrastructure. The move reflects Apple's plan to help power the compute layer that AI developers rely on.

Nvidia Vera Rubin AI accelerator delivers twice the profit per watt

Nvidia's next-generation Vera Rubin AI Accelerator is reported to deliver 2X profit per GW compared to Blackwell. The platform targets hyperscaler energy efficiency challenges. It is designed to help developers build and deploy AI models more easily. Many companies have already shown strong interest in the new AI platform.

Goldman banking leader shares how she uses AI

Goldman Sachs investment banking chief Kim Posnett shared how she uses AI in her work. She advised junior analysts to balance their AI use carefully. The guidance came as the company explores how AI tools can support workers at different experience levels. Posnett emphasized that new employees should use AI thoughtfully.

Microsoft outlines AI principles for education

Microsoft released a plan for using AI in education built on five key principles. The company stressed safety, privacy, and transparency in AI design. It also said educators should stay in control and that AI should support learning rather than replace thinking. Microsoft cited examples showing improved student outcomes with tools like Microsoft 365 Copilot.

New framework improves virtual sensor data for activity recognition

Researchers created a coverage-aware virtual IMU augmentation framework to improve human activity recognition. The method selects diverse and scarce data points in a learned sensor space. It generates virtual IMU candidates and weights them by reliability during training. Tests on public benchmarks showed consistent performance improvements over existing methods.

Siemens stock may be undervalued after industrial AI news

Siemens stock has risen about 106.8 percent over three years but may now be undervalued. Some analysts estimate the stock could be 34 percent undervalued at around $120 per share. The current price is roughly $90. Siemens has gained attention for AI-powered manufacturing solutions and industrial partnerships.

Mitsubishi Heavy invests in Preferred Networks AI partnership

Mitsubishi Heavy Industries is investing $64 million in AI startup Preferred Networks. The partnership focuses on developing domestic AI technologies for defense equipment and security systems. The companies announced their alliance in June 2026. The investment aims to strengthen research and development cooperation.

AI framework cuts catastrophe risk modeling costs sharply

A new TAISE framework uses AI weather models to generate extreme weather scenarios at lower cost. Traditional catastrophe risk models rely on expensive manual methods from the 1990s. The framework produces continuous atmospheric fields where extreme events emerge naturally. Tests showed about ten times less computational cost than conventional approaches.

TEDCO invests $250K in enterprise AI startup Syntes

TEDCO made a $250,000 seed investment in enterprise AI startup Syntes. The funding will help Syntes develop its AI platform that optimizes business workflows. The company uses machine learning to help businesses make data-driven decisions and cut costs. Syntes has worked with TEDCO since 2020 and plans to expand its team.

Sources

• Apple reportedly building server packed with M-series Ultra chips for AI
• Apple (AAPL) Builds Secret Enterprise AI Server Push Beyond Consumer Hardware
• NVDA Stock Rises Premarket: Vera Rubin AI Accelerator Delivers 2X Profit Per GW Than Blackwell, Semianalysis Says
• Goldman banking chief shares how she uses AI
• Microsoft’s commitment for AI in education
• Coverage-Aware Virtual IMU Augmentation for Low-Resource Human Activity Recognition
• Siemens (XTRA:SIE) Stock Could Be 34% Undervalued Following Fresh Industrial AI News
• Mitsubishi Heavy to boost Preferred Networks AI tie-up with investment
• From Manual Construction to AI-Driven Scenario Emergence: Rethinking Catastrophe Risk Modeling
• Dealroom.co | TEDCO backs enterprise AI startup Syntes with $250K seed investment