Google DeepMind's Major Restructure: Hassabis Goes Full AGI
Sundar Pichai announced sweeping changes at Google DeepMind. Demis Hassabis becomes Chair of GDM and Chief Scientist of Alphabet while continuing to lead Isomorphic Labs — a role designed to let him focus entirely on "shaping the future of AGI." Koray Kavukcuoglu steps up as SVP of Google DeepMind to run day-to-day operations across Gemini development, frontier research, and the Gemini app. Most surprising: Jeff Dean, after 27 years at Google, is leaving to launch an independent public benefit corporation with Sanjay Ghemawat focused on ML and science. Google will be a founding investor. This is Google explicitly prioritizing the AGI race at the leadership level.
Read on Google Blog →IDC: AI Compute Spending to Hit $702B by 2029
New IDC analysis reveals a dramatic split in AI economics: inference costs have fallen 300x since GPT-3's launch, yet AI compute hardware spending will surge from $56B in 2023 to $702B in 2029 — a 12.5x increase in six years. The report also projects that by 2035, 48% of enterprise applications will effectively become "agents as apps," with another 24% being "agent-led." Software pricing is shifting from 60% subscription today toward consumption and outcome-based models. The economics are inverted: AI is cheap to consume but requires massive infrastructure to support.
Read on InfotechLead →MIT/Gartner: 95% of GenAI Pilots Fail to Deliver ROI
A sobering reality check from BigDataWire: MIT research found that 95% of generative AI pilots are failing to deliver measurable financial returns. Gartner projects that over 40% of agentic AI projects will be cancelled by end of 2027. The culprit is what they call the "first mile gap" — enterprise data isn't ready for AI. Context is killing enterprise AI, and it's creating M&A opportunities for companies that can solve the data preparation problem. The hype cycle is hitting the wall.
Read on Woodside Capital →6sense Pipes Buying Intelligence Directly Into AI Agents via MCP
6sense announced a new MCP server that pushes real account and intent data directly into MCP-compatible AI agents — Claude, ChatGPT, Writer, Agentforce — without custom integration. Sales, marketing, and revenue ops agents can now act on live buying signals like predicted buying stages and qualified account status inside their existing workflows. This is the infrastructure layer that makes agentic GTM practical: agents grounded in the same data the analytics team relies on, updated in real time.
Read on AI Agent Store →Ads Designed for AI Agents Are Emerging
Business Insider reports on a new marketing channel: ads designed to influence AI agents rather than human consumers. As AI agents increasingly handle shopping, research, and purchasing decisions, advertisers are exploring how to get their products surfaced when an agent — not a person — is making the choice. It's early, but it's a signal: the customer journey is bifurcating into human and machine decision paths. PMMs will need playbooks for both.
Read on Business Insider →💡 My Take
The IDC numbers tell a structural story: AI token costs have collapsed 300x, but infrastructure spending will grow 12x by 2029. Cheap consumption + expensive infrastructure = massive concentration of power in companies that can build the compute layer. Meanwhile, 95% of pilots failing and 40%+ of agent projects expected to die by 2027 isn't doom — it's a clearing event. The winners will be those who solve the data readiness problem, not those who deploy the shiniest model. And Google putting Hassabis in a "Chief Scientist of Alphabet" role to focus on AGI? That's the clearest signal yet that the big labs see AGI as a near-term race, not a research horizon.