2026-05-18
By Vadym · Generated with AI, curated by me
Google is unwrapping its AI hand on the eve of I/O while the infrastructure question looming beneath the whole industry gets harder to ignore. Energy costs are reshaping where AI gets built — and money is flowing toward molecules and factory floors, not just chat interfaces.
Google held “The Android Show: I/O Edition” on May 18, one day before the full I/O keynote, rolling out Gemini Intelligence — a proactive AI layer that translates user intentions into actions across Android, Wear OS, Android Auto, and Android XR. Two new hardware categories were revealed: Googlebooks, AI-first laptops running a unified Aluminum OS that merges Android and ChromeOS, and Android XR smart glasses in partnership with Samsung, Warby Parker, Gentle Monster, and XREAL. Chrome on Android is getting an agentic “auto browse” mode. Google is clearly choosing the I/O eve moment to set terms before tomorrow’s keynote can be overshadowed by competitor news.
OpenAI rolled out a personal finance preview for $200/month Pro users in the US, letting them connect bank accounts via Plaid — a service plugged into more than 12,000 institutions including Chase, Fidelity, Schwab, and American Express. Once linked, ChatGPT surfaces a spending dashboard, subscription tracker, and portfolio overview, and answers questions grounded in real transaction data. The model can read balances, transactions, and liabilities but cannot move money. An Intuit integration is planned. This is the clearest signal yet that OpenAI sees ChatGPT as a financial OS, not just a chatbot — and that it intends to own the personal finance layer that banks and fintech apps have been fighting over for a decade.
Data center costs across Europe’s five largest markets — Frankfurt, London, Amsterdam, Paris, and Dublin — are set to rise 12% in 2026, CNBC reported today. Industrial energy prices in Europe average roughly double US rates and 50% above China and India. OpenAI paused its Stargate expansion in the UK partly over energy costs. Franklin Templeton’s global investment strategist put it plainly: “If I were making the next $7 billion data center, it would be in the US or China.” Analysts are floating nuclear as Europe’s long-term fix, but that is a decade-scale bet. In the near term, AI geography is being decided by the electricity bill.
Isomorphic Labs closed a $2.1 billion Series B led by Thrive Capital on May 12, bringing total outside capital raised to roughly $2.7 billion. New investors include Temasek, MGX (Abu Dhabi), CapitalG, and the UK Sovereign AI Fund; Alphabet and GV participated again. The company’s IsoDDE platform extends AlphaFold’s protein structure work into a unified drug design engine that integrates binding affinity prediction and pocket identification. Active drug-discovery deals are live with Novartis, Eli Lilly, and Johnson & Johnson. Target: clinical trials before end of 2026 — a deadline Hassabis had already pushed once. Life sciences is now the vertical where AI money is writing the largest checks, and Isomorphic is the clearest proof that DeepMind’s scientific bets have commercial legs.
The US Commerce Department’s CAISI (Center for AI Standards and Innovation) formalized pre-deployment evaluation agreements with Google DeepMind, Microsoft, and xAI, joining OpenAI and Anthropic in a program that has now completed more than 40 model assessments. The agreements are voluntary, not regulatory mandates — built around information sharing and collaborative research rather than enforcement. Labs frequently share models with reduced safeguards for national security evaluation. The significance is structural: pre-deployment government review is quietly becoming the baseline expectation for any lab operating at the US frontier, even without a formal legal framework requiring it.
Mind Robotics, founded by Rivian CEO RJ Scaringe, closed a $400 million round led by Kleiner Perkins on May 13, pushing total funding past $1 billion and valuation to $3.4 billion. The company builds AI-powered robots for industrial manufacturing environments, using Rivian’s own factories as a live training and deployment site. The round is notable for its pace: $115 million seed in late 2025, $500 million Series A in March 2026, now another $400 million nine weeks later. Physical AI venture funding has already hit $37 billion in 2026, shattering the previous full-year record of $21 billion. The bet behind all of it: inference-grade hardware on the factory floor is the next platform, and whoever gets there first writes the playbook.
Six stories, one meta-theme: capability is no longer the binding constraint. Google is weaving Gemini into every Android surface before a competitor can claim the space. OpenAI is moving into your bank account. DeepMind is helping design molecules. A Rivian founder is betting that the real AI prize is the factory floor. Meanwhile, the US government quietly formalized review of every major frontier model before it ships — and Europe is watching all of this from a cost structure that makes large-scale AI infrastructure economically irrational to build there. The next phase of the AI race is not a benchmark competition. It’s a fight over energy, geography, distribution, and trust — and those are much harder problems than training a better model.
— Boba
Curated by Vadym