Summary of Demis Steps Down, Apple’s Memory Problem, Microsoft’s Clever Trick

Overview of Demis Steps Down, Apple’s Memory Problem, Microsoft’s Clever Trick

This episode of Big Technology Podcast covers three major AI-era business stories: Demis Hassabis stepping down as CEO of Google DeepMind, Apple’s exposure to rising memory-chip costs amid the AI build-out, and Microsoft’s seemingly restrained AI spending that may actually be driven by accounting changes. Alex Kantrowitz and M.G. Siegler also dig into a broader strategic question: whether Google’s AI strategy is drifting away from the LLM race toward “world models” and other breakthrough research paths.

Demis Hassabis Steps Down from Google DeepMind

What happened

• Demis Hassabis stepped down as CEO of Google DeepMind and will become chairman and chief scientist.
• Google DeepMind tech chief Koray Kavukcuoglu is taking over responsibility for AI model development.
• Google veteran Jeff Dean is also leaving to co-found a new AI startup.

Why it matters

• The move signals a possible split between:
  • Hassabis’s long-term research vision, which emphasizes breakthroughs beyond today’s LLMs
  • Google’s commercial AI priorities, which are increasingly centered on shipping competitive products
• Hassabis has consistently suggested that current LLMs may not be enough to reach AGI and that additional breakthroughs are needed.

Main takeaway

• Alex argues the departure reflects a difference in opinion more than a simple leadership shuffle.
• He sees it as good for AI overall, because Hassabis can focus on scientific breakthroughs while the LLM race continues elsewhere.
• For Google, though, it raises serious questions about talent retention, strategy, and whether the company can keep pace in the most important tech battle right now.

Apple’s Memory Problem and the Coming Price Hike

The core issue

• Apple is being hit by rising memory and component costs because AI data centers are absorbing huge amounts of chip supply.
• Even though Apple is not spending like its AI rivals on infrastructure, it still relies on many of the same scarce components for its devices.

Why this is a problem for Apple

• Apple’s supply chain advantage is being strained by the AI boom.
• Tim Cook described the shortages as “very significant.”
• Apple already gave weaker guidance, and the concern is that margin pressure may worsen if it has to raise device prices.

Expected pricing strategy

• M.G. Siegler expects Apple to raise prices on future devices, likely by:
  • $100–$200 as a base case
  • potentially $250–$300+ for higher-end models
• Apple has already raised prices on some Macs and may do the same for iPhones, especially premium models like the rumored foldable/“Ultra.”

Apple’s likely response

• Apple is trying to soften sticker shock by shifting more customers into financing/subscription-style ownership:
  • The new Apple Upgrade program
  • Monthly payments for iPhone, Mac, iPad, and Apple Watch
• This could evolve into a more Amazon Prime-like ecosystem where users pay Apple a recurring fee for hardware and services access.

Services growth concern

• Apple’s services business, especially the App Store, is also under pressure.
• Slower services growth was partly blamed on:
  • App Store regulation and antitrust pressure
  • gaming slowdown
  • China-related softness
• That matters because services are a major part of Apple’s valuation and growth story.

Microsoft’s “Clever Trick” on AI Spend

What Wall Street saw

• Microsoft appeared to be spending less than expected on AI infrastructure, which Wall Street liked.

What may actually be happening

• Siegler explains that Microsoft changed how it accounts for some data center leases:
  • Previously: 15-year lease assumptions
  • Now: 25-year useful-life assumptions for facilities
• That shifts some spending from capex to opex, making capex look lower without necessarily reducing total spend.

Why it matters

• The move is likely legal and defensible, but it also has a clear optical benefit.
• It helps Microsoft look more disciplined than rivals like Google, Amazon, and Meta, even if the real AI build-out continues at a similar pace.

Bigger question

• There’s a growing concern that a lot of cloud growth is being driven by a handful of heavily funded AI players, especially OpenAI and Anthropic.
• If that spending slows, cloud growth could look much less impressive across the industry.

Google’s AI Strategy: LLMs vs. World Models

The debate

• Google appears to be leaning harder into world models and broader scientific AI research, rather than purely pushing LLMs.
• That aligns with Hassabis’s long-held view that LLMs alone may not get to AGI.

Why Google is falling behind

• The company has been slow to ship its next flagship model.
• It has also been less visible in coding agents and other frontier product categories.
• Internal competition and “fiefdoms” may be slowing execution.

Possible upside and downside

• Upside: Google could be investing in the next major AI breakthrough rather than just competing in today’s race.
• Downside: It risks losing the current market, especially if rivals use LLM progress to build better products first.

Notable Themes and Takeaways

• AI is reshaping every part of big tech, not just the companies directly building models.
• Apple is getting squeezed by the AI boom despite not being a front-line AI infrastructure player.
• Microsoft may be more aggressive than it appears, just with better accounting optics.
• Google is facing a strategic identity crisis: ship competitive LLM products now, or keep betting on deeper research breakthroughs.
• The recurring theme is that commercial incentives and scientific ambition are increasingly pulling AI companies in different directions.