Google Cloud CEO: Anthropic, TPUs, Mythos, NVIDIA and More

Matthew Berman
Matthew BermanApr 24, 2026

Why It Matters

Google’s control over both hardware and AI models creates a sustainable competitive moat, driving revenue growth while demonstrating AI’s tangible benefits to businesses and the public.

Key Takeaways

  • Google controls its own silicon, boosting TPU margins and capacity.
  • TPU manufacturing shift accelerates data‑center deployment, outpacing construction.
  • Google rents TPUs to finance AI research and fund Gemini.
  • Strategic AI use cases showcase societal benefits, easing job‑displacement fears.
  • Community‑focused data‑center strategy addresses energy, employment, and public sentiment.

Summary

Google Cloud’s CEO Thomas highlighted the company’s unique position in the AI race, emphasizing its end‑to‑end stack—from custom‑designed TPUs to the Gemini model suite—and how this integration fuels both internal projects and external revenue streams.

Long‑term planning has allowed Google to sidestep the compute crunch that haunts rivals. By shifting data‑center construction to a manufacturing model, diversifying energy sources, and pre‑securing real‑estate, the firm can deploy racks at scale in weeks rather than months. Owning silicon, now entering its eighth‑generation TPU era, lets Google sell hardware, lease inference capacity to labs like Anthropic, and monetize token usage across a broad customer base, all while preserving strong margins.

Thomas cited concrete examples: Citadel’s high‑frequency trading desks leveraging TPUs for ultra‑low‑latency inference, Germany’s Signal insurer cutting claim‑resolution time from minutes to seconds, and the American Society for Clinical Oncology using Gemini to verify complex treatment guidelines. He also described new code‑security tools that detect and repair vulnerabilities, underscoring practical, revenue‑generating AI applications.

The broader implication is clear: owning the compute stack transforms Google from a cloud provider into a profit‑center that can fund ambitious AI research, shape public perception of AI’s societal value, and mitigate regulatory or community pushback through localized, energy‑efficient data‑center deployments.

Original Description

Join My Newsletter for Regular AI Updates 👇🏼
My Links 🔗
👉🏻 Forward Future X: https://x.com/forwardfuture
Media/Sponsorship Inquiries ✅
Chapters:
0:00 – Intro
0:42 – Google's Insane Compute Capacity
03:17 – TPU Monetization
05:24 – Why Google Doesn't Hoard Compute?
08:02 – Datacenter Buildout
15:01 - Does AGI Mean Job Displacement?
17:55 - NVIDIA vs TPU (Total Cost of Ownership)
23:25 - 8th Gen TPU
24:32 - Training vs. Inference
30:53 - Google's "Extreme Co-design"
35:01 - Working with Anthropic
37:46 - Serving Mythos-sized Models (10T)
41:42 - Google engineering culture (Steve Yagge tweet)
48:27 - Cybersecurity
51:50 - What keeps Thomas up at night?

Comments

Want to join the conversation?

Loading comments...