SoftBank Group said it will invest up to €75 billion to build and operate as much as 5 gigawatts of artificial intelligence data‑centre capacity in France, announcing an initial €45 billion phase to deliver 3.1 GW in the Hauts‑de‑France region by 2031. The first three sites named include Dunkirk (Loon‑Plage), Le Bosquel and Bouchain; state energy firm EDF and Schneider Electric are reported partners, with EDF offering a former power plant site for conversion and Schneider supplying modular equipment and a manufacturing hub at the Port of Dunkirk. The commitment, unveiled ahead of France’s Choose France investment summit, is framed by SoftBank founder Masayoshi Son as the group’s largest AI infrastructure push in Europe and follows the firm’s other global AI projects, including large US data‑centre plans and heavy investment in OpenAI. Announcements highlight France’s nuclear‑heavy grid as an energy advantage for high‑density computing, while analysts note questions over financing and SoftBank’s broader capital commitments.

A wave of AI-driven demand is forcing major shifts across cloud and datacenter markets, exposing constraints in power, planning and capital. Analysts warn developers face grid and community pushback in the United States even as firms race to add capacity; UK planning reform proposals could insulate datacenter builds from judicial review. Large commercial bets are underway: Snowflake is reported to be committing about $6 billion to AWS Graviton CPUs and AI accelerators, while vendors and providers point to cheaper bare-metal cloud and evolving offerings as alternatives to on-prem hardware. National labs and research centres are also repurposing excess supercomputing capacity to offer private AI inference services, and engineers have open-sourced tools aimed at slashing AI operating costs. At the same time, cloud platforms continue to fold new models into their stacks — reports say AWS may integrate Elon Musk’s Grok into Bedrock — underscoring provider-led consolidation. The combined trends — soaring compute demand, energy and siting bottlenecks, heavy platform investments and cost-reduction efforts — are reshaping where and how AI workloads will be hosted.

DuckDuckGo reported a sharp uptick in downloads and traffic after Google unveiled major AI-led changes to Search at I/O 2026. In the United States the app’s installs rose by an average of 18.1% week-on-week between May 20 and May 25, peaking around May 25 at roughly 30% above normal levels; iPhone downloads averaged a 33% weekly increase and saw a single-day spike of almost 70% on Apple devices, according to company data and independent analytics cited by media. Traffic to DuckDuckGo’s opt-in “No AI” search page also climbed markedly — the company reported roughly 23% weekly growth while other outlets said visits have tripled since Google’s announcement. The movement follows user unease over Google’s shift from traditional blue links to AI-generated answers and reports that Chrome began installing a 4 GB Gemini Nano model on devices. DuckDuckGo positions itself as offering choice: an AI-free default search experience plus optional AI via duck.ai, while continuing to rely in part on Microsoft’s Bing index. Though DuckDuckGo still holds a small share of the US search market (~3%), the shift signals user resistance to mandatory AI integration.

Daily spoilers and walkthroughs for The New York Times word games on May 31, 2026: Strands (game #819) and Connections (game #1085). Strands’ theme was “Places to go,” with solver clues provided and the spangram revealed as TOURISM; non-spangram answers included MUSEUM, BEACH, CASTLE, MARKET, MONUMENT and RESTAURANT. Connections’ four groups were published with hints and full answers: Yellow (things that are yellow) — BUTTER, PIKACHU, RUBBER DUCK, SCHOOL BUS; Green (billiards terms) — BREAK, CUE, POCKET, RACK; Blue (slang for a sailor) — JACK, SALT, SEA DOG, TAR; Purple (kinds of wood plus “s”) — SASH, SOAK, SPINE, STEAK. Coverage appeared across outlets including CNET and TechRadar on May 30–31, 2026, offering step-by-step hints, recommended clue words to unlock in-game assistance, and brief difficulty ratings to help players maintain streaks and improve scores.

U.S. prosecutors have charged Michele Spagnuolo, a Google software/security engineer, with insider trading after he allegedly used confidential Google data to win roughly $1.2 million trading on prediction market Polymarket. According to a complaint unsealed in late May 2026, Spagnuolo — who operated as “AlphaRaccoon” on the platform — placed bets tied to Google’s 2025 Year in Search and other internal search-data insights between October and December 2025. He was arrested in New York and faces criminal counts including commodities fraud, wire fraud and money laundering; the Commodity Futures Trading Commission filed a parallel civil complaint. Google said the employee accessed marketing tools available internally, has been placed on leave, and is cooperating with authorities. Polymarket said it worked with the U.S. Attorney’s Office and the CFTC and that blockchain transaction records helped trace the activity. The case follows a recent arrest of a U.S. servicemember accused of similar wagering and has drawn fresh scrutiny from lawmakers and regulators over how prediction markets are monitored and policed.