2026-06-20
By Vadym · Generated with AI, curated by me
Washington ordered the national power grid redesigned around AI data centers, half of Americans are now using chatbots despite distrusting the companies that build them, and the EU’s chatbot transparency clock hits zero in six weeks. The infrastructure layer of the AI transition is moving faster than the discourse about it.
On June 18, the Federal Energy Regulatory Commission issued six orders to the major regional grid operators covering most of the US — PJM, MISO, SPP, CAISO, ISO New England, and NYISO — requiring them to overhaul how they process large-load interconnection requests. Data centers are now seeking “hundreds of megawatts and in some cases gigawatt-scale” power commitments, but existing interconnection queues take years to navigate. Under the orders, operators have 60 days to justify their current tariffs or file revisions covering five areas: faster application processing, cost-shift safeguards, co-location rules, new transmission services, and dedicated study processes for large-load generation. Texas’s ERCOT is excluded due to limited federal jurisdiction. The national power grid is now, by formal regulatory order, a critical dependency of AI infrastructure — not just a background utility. The question of whether the grid can be reshaped fast enough to match the rate of AI compute demand is now a federal policy question, not an operational one.
Pew’s “Americans and AI 2026” report, released June 17 based on a February survey, maps a public that has adopted AI chatbots far faster than its comfort level has grown. 49% of US adults now use chatbots — up from 33% in 2024 — with ChatGPT leading at 44%, Gemini at 24%, Copilot at 17%, and Meta AI at 14%. Daily chatbot use sits at 24%. The sentiment data tells a different story: 63% believe AI is advancing too quickly, 71% expect it to make their personal information less secure, 59% distrust US companies to develop AI responsibly, and only 40% expect a positive personal impact. Confidence in government regulation dropped further, from 62% in 2024 to 33% today. Adoption and distrust are rising in parallel, not trading off. The industry instinct is to read high usage numbers as proof of acceptance. Pew’s data suggests something else: people are using the tools because they’re useful, not because they believe the companies behind them are trustworthy. That gap is going to matter when the next incident happens.
German company NEURA Robotics announced a $1.4 billion Series C on June 10, led by Tether and joined by NVIDIA, Amazon, Qualcomm, Bosch, Schaeffler, the European Investment Bank, and several other institutional investors. The round values NEURA at $7 billion — the largest funding round ever for a full-stack robotics company, and a milestone that makes it Europe’s most-funded humanoid robot maker. The full $1.4B is milestone-contingent on undisclosed targets. NEURA has a €1 billion order backlog and is targeting production of 5 million robots by 2030 across manufacturing, logistics, healthcare, and consumer segments. The investor lineup is not accidental: NVIDIA needs humanoid robots to absorb AI compute at scale; Amazon needs them for warehouse automation; Bosch and Schaeffler are industrial incumbents hedging their physical-world futures. The global robotics market has raised $55.8 billion so far in 2026, nearly double last year’s record. NEURA’s round suggests the ceiling has not been reached.
On August 2, 2026, Article 50 of the EU AI Act takes effect for all AI chatbot systems interacting with EU users. From that date, any AI system designed to converse with people must clearly disclose that the user is talking to an AI — before or during the interaction — unless it is obvious to a reasonable person. Non-compliance fines run to €35 million or 7% of global annual revenue, whichever is higher. A separate Annex III deadline for high-risk systems was pushed to December 2027 under the Digital Omnibus provisional agreement of May 7, giving enterprises more time on safety-critical applications. The chatbot disclosure rule was not extended. Six weeks out, the practical challenge is not the rule itself — labeling a chatbot as AI is not technically complex. The challenge is that audit trails, user consent flows, and documentation requirements under Article 50 are more involved than the headline disclosure suggests. Any company that assumed EU AI Act compliance was a 2027 problem now has a current-quarter deadline for at least this layer of it.
Today’s stories share a common thread: the infrastructure of the AI era is being built out in parallel with the models themselves, and it’s moving on its own timeline. FERC ordering six regional grids to redesign their interconnection rules is not a policy footnote — it’s a structural acknowledgment that AI compute demand has outgrown the power grid as it exists today. The Pew data is the human layer: half of Americans are using chatbots, but the trust gap between usage and belief in the companies behind the tools is widening, not closing. Neura Robotics and the $55.8 billion raised by robotics companies this year shows physical AI is moving past the demo phase. GitHub’s move to usage-based billing and the Windsurf-to-Devin-Desktop transition both reflect the same maturation in developer tooling — flat-rate models are giving way to consumption-indexed ones, and editor-centric IDEs are giving way to agent-centric ones. And the EU AI Act clock is running whether companies are ready or not. The AI transition is no longer waiting for the discourse to catch up with it.
— Boba
Curated by Vadym