SED News: Apple Bets on Gemini, Google’s AI Advantage, and the Talent Arms Race

February 3, 2026

51 min episode · 2 min read

Episode

51 min

Read time

2 min

Topics

Artificial Intelligence

AI-Generated Summary

Published Feb 3, 2026

Key Takeaways

✓Google's TPU Advantage: Google's tensor processing units strip away GPU architectural baggage designed for gaming graphics, focusing solely on massive matrix multiplication for deep learning. This allows Google to avoid NVIDIA's 75% gross margin markup that competitors pay, creating significant cost advantages in model training and inference at scale.
✓OpenAI Vesting Strategy: OpenAI removes traditional one-year vesting cliffs, allowing monthly or quarterly equity vesting from day one. This unprecedented approach aims to attract top AI talent without fear of losing unvested equity, though it risks attracting mercenaries over mission-driven employees and costs $6 billion annually in stock compensation.
✓Starlink Aviation Impact: Starlink provides gate-to-gate satellite internet on flights with no login friction or charges to consumers per contract terms. The service delivers sufficient bandwidth for productive work on eight-hour flights, representing a fundamental shift in aviation connectivity compared to traditional paid, limited WiFi systems.
✓AI Coding Tool Evolution: Senior engineers now derive more value from AI coding agents than junior developers, reversing earlier patterns. Effective use requires architectural knowledge and understanding of system integration. This creates concerns about junior developers never gaining deep expertise if insulated from implementation details by AI assistance.
✓Manus Acquisition Strategy: Meta acquires Chinese AI startup Manus for $2 billion after the company relocated to Singapore in 2024 to derisk Chinese ties. Despite generating $100 million revenue, Manus struggled with expensive token costs and unclear billing models that frustrated users requiring frequent credit top-ups for AI-generated presentations and websites.

What It Covers

OpenAI declares code red as Google's TPU advantage and Gemini 3 model demonstrate superior AI capabilities. Apple partners with Gemini for Siri improvements. OpenAI eliminates one-year vesting cliffs to compete in talent wars, spending $6 billion annually on stock compensation—nearly 50% of projected revenue.

Key Questions Answered

•Google's TPU Advantage: Google's tensor processing units strip away GPU architectural baggage designed for gaming graphics, focusing solely on massive matrix multiplication for deep learning. This allows Google to avoid NVIDIA's 75% gross margin markup that competitors pay, creating significant cost advantages in model training and inference at scale.
•OpenAI Vesting Strategy: OpenAI removes traditional one-year vesting cliffs, allowing monthly or quarterly equity vesting from day one. This unprecedented approach aims to attract top AI talent without fear of losing unvested equity, though it risks attracting mercenaries over mission-driven employees and costs $6 billion annually in stock compensation.
•Starlink Aviation Impact: Starlink provides gate-to-gate satellite internet on flights with no login friction or charges to consumers per contract terms. The service delivers sufficient bandwidth for productive work on eight-hour flights, representing a fundamental shift in aviation connectivity compared to traditional paid, limited WiFi systems.
•AI Coding Tool Evolution: Senior engineers now derive more value from AI coding agents than junior developers, reversing earlier patterns. Effective use requires architectural knowledge and understanding of system integration. This creates concerns about junior developers never gaining deep expertise if insulated from implementation details by AI assistance.
•Manus Acquisition Strategy: Meta acquires Chinese AI startup Manus for $2 billion after the company relocated to Singapore in 2024 to derisk Chinese ties. Despite generating $100 million revenue, Manus struggled with expensive token costs and unclear billing models that frustrated users requiring frequent credit top-ups for AI-generated presentations and websites.

Notable Moment

A conference attendee publicly challenged the Manus CTO about unsustainable costs, stating they spent excessive amounts on credits that depleted weekly. This highlighted the fundamental tension in consumer-facing AI products between delivering value and managing expensive token costs, ultimately contributing to the company seeking acquisition rather than continued independence.

Know someone who'd find this useful?