Google reshapes DeepMind leadership; Hassabis becomes Alphabet Chief Scientist
Koray Kavukcuoglu takes over as SVP of Google DeepMind, while Jeff Dean and Sanjay Ghemawat launch a public‑benefit AI lab with Google as a founding investor. Meanwhile, Reuters highlights DeepSeek’s 3¢ per‑test model cost, sharpening price pressure across AI workloads.
One-Line Summary
Google reorganizes DeepMind leadership and centralizes AI updates, while DeepSeek’s low-cost model intensifies pressure on model run costs.
Big Tech
Google reshapes DeepMind leadership for frontier AI focus
Google announced a leadership reshuffle at Google DeepMind: Demis Hassabis becomes Chair of Google DeepMind and Chief Scientist of Alphabet, while Koray Kavukcuoglu steps up as SVP of Google DeepMind reporting to Sundar Pichai. Kavukcuoglu will oversee Gemini model development, frontier AI research, and the Gemini app and developer teams, and Hassabis will continue to lead Isomorphic Labs while advising across models and research. 1
Pichai said Google must accelerate its AI work and cited demand for Gemini models, with the Gemini app reaching 950M+ monthly users alongside momentum in Search, YouTube, and Cloud. He added that Flash is in high demand, a Cyber model is live, and Gemma models have surpassed 900M+ downloads, underscoring Google’s full‑stack AI footprint. 1
Jeff Dean and Sanjay Ghemawat are launching an independent public benefit corporation to accelerate discoveries in machine learning, science, and engineering. Google will be a founding investor and Cloud partner, with plans to collaborate on a research framework for ML systems and related infrastructure advances. 1
Hassabis framed his new role around shaping the future of AGI and science, noting close work with Pichai and progress on new models including Gemini 4. The shift signals Alphabet concentrating frontier research leadership while pushing product teams to move faster on Gemini experiences like AI Mode and AI Overviews. 1
Industry & Biz
DeepSeek’s V4-Flash is the cheapest to run, study says
A research firm, Artificial Analysis, finds DeepSeek’s V4-Flash is by far the least expensive to run among well‑known models, with pricing at $0.14 per million input tokens and $0.28 per million output tokens. The firm estimates an average cost of 3 cents per test, versus 86 cents for Moonshot AI’s Kimi K3, $1.86 for OpenAI’s GPT-5.6 Sol, and $3.15 for Anthropic’s Claude Fable 5, roughly 105 times higher than V4‑Flash. 2
On performance, Reuters reports V4‑Flash ranks on par with Google’s Gemini 3.6 Flash but trails offerings from OpenAI and Anthropic. The company is seeking to regain momentum and, per prior Reuters reporting referenced in the article, is preparing for a potential IPO, underscoring intensifying price competition in model selection as of Aug 3, 2026. 2
New Tools
Google posts July 2026 AI updates hub
Google published a roundup page for July 2026 that aggregates links to AI updates across models and research, products, infrastructure and cloud, and technology topics like safety and health. It serves as a single starting point to find the latest posts. 3
The hub links out to areas such as the Gemini app and Gemini Notebook, developer tools, Workspace, Android and Google Play platforms, and devices like Pixel, Nest, Fitbit, and Chromebooks, as well as DeepMind, Research, Developers, Cloud, and Security blogs. For teams, this makes it easier to track which updates could affect their tools and workflows. 3
What This Means for You
Leadership clarity at Google DeepMind points to continued investment in frontier models and faster product shipping across Gemini surfaces. With Pichai citing 950M+ monthly users for the Gemini app, features tested at DeepMind can reach consumer and enterprise touchpoints quickly. 1
The cost data around DeepSeek V4‑Flash shows why procurement and product teams should model AI usage by tokens and tasks, not sticker price alone. If a task doesn’t demand top‑tier reasoning, a low‑cost model can materially reduce per‑request spend without changing your workflow. 2
Google’s July roundup page is a practical discovery tool: it curates updates across Workspace, Android, and Pixel so non‑developers can skim what’s new in one place and alert the right teammates. 3
Jeff Dean and Sanjay Ghemawat’s new public benefit corporation, with Google as founding investor and Cloud partner, suggests fresh ML systems research and infrastructure ideas could flow into Google’s ecosystem. Teams relying on Google Cloud or ML tooling may see new frameworks or guidance emerge from that collaboration. 1
Action Items
- Try the Gemini app: Use it to draft a one‑pager or summarize a meeting this week and evaluate whether the output fits your team’s tone.
- Build a quick cost comparison: Take 5 everyday prompts, note your current provider’s token usage, then project monthly spend using Reuters’ per‑million token figures for V4‑Flash versus your current model.
- Read the DeepMind leadership memo: Skim Pichai and Hassabis’s notes and write a 5‑bullet internal takeaway on implications for your product, marketing, or data roadmap.
- Bookmark Google’s July AI hub: Set a 15‑minute slot to scan it and flag any Workspace, Android, or device updates relevant to your team.
Comments (0)