diff --git a/data/ai-infrastructure-projects.json b/data/ai-infrastructure-projects.json
index 85a2127..cb6d35b 100644
--- a/data/ai-infrastructure-projects.json
+++ b/data/ai-infrastructure-projects.json
@@ -1,6 +1,6 @@
 {
   "version": 2,
-  "generated_at": "2026-06-05T00:00:00Z",
+  "generated_at": "2026-06-15T00:00:00Z",
   "license": "TF curation CC-BY-4.0. Underlying public-record sources retain their original license.",
   "notice": "Editorial registry of disclosed major AI buildout projects. Each entry preserves canonical source URLs. Capacity, capex, and timeline figures reflect publicly announced or reported numbers at time of curation and may evolve. Status enum: announced (publicly committed, no ground broken), permitted (regulatory greenlights obtained), construction (active build), operational (online and serving traffic). Power source enum is the announced primary source; many projects mix multiple feeds.",
   "schema": {
@@ -319,6 +319,24 @@
       "primary_sources": [
         "https://tensorfeed.ai/originals/anthropic-spacexai-colossus-orbital"
       ]
+    },
+    {
+      "id": "softbank-france-hauts-de-france",
+      "name": "SoftBank AI Data Centers France",
+      "operator": "SoftBank Group",
+      "partners": ["Schneider Electric", "EDF", "SB Energy"],
+      "location": { "city": "Loon-Plage, Bosquel, Bouchain", "state": "Hauts-de-France", "country": "France" },
+      "status": "announced",
+      "capacity_mw_announced": 5000,
+      "power_source": "EDF grid (nuclear-rich, low-carbon)",
+      "announced_date": "2026-05-30",
+      "target_operational": "2031",
+      "capex_usd_billion": 87,
+      "context": "SoftBank Group announced at the 2026 Choose France summit that it will invest up to 75 billion euros (approximately $87 billion) to develop and operate AI data center capacity in France. Phase 1 commits 45 billion euros to deliver 3.1 gigawatts in the Hauts-de-France region by 2031, with sites in Dunkirk (Loon-Plage), Bosquel, and Bouchain. Total target capacity is 5 gigawatts. Schneider Electric is a manufacturing partner for power modules at the Port of Dunkirk. EDF provides low-carbon power leveraging France's nuclear-heavy electricity grid. The commitment represents the largest single European AI infrastructure investment announced by the Japanese technology conglomerate, signaling SoftBank's move from investor to operator of AI compute capacity.",
+      "primary_sources": [
+        "https://group.softbank/en/news/press/20260531_0",
+        "https://techcrunch.com/2026/05/30/softbank-says-it-will-invest-up-to-e75-billion-to-build-french-data-centers/"
+      ]
     }
   ]
 }
diff --git a/data/page-freshness.json b/data/page-freshness.json
index 16bafe4..f04d0c0 100644
--- a/data/page-freshness.json
+++ b/data/page-freshness.json
@@ -8,27 +8,27 @@
   "policy": "Editor bumps last_reviewed when the page gets a real review or content update. Pages with cadence other than 'editorial' get a Next-Review-Due stamp computed from last_reviewed + cadence interval, and surface an OVERDUE flag when today is past that date.",
   "pages": {
     "/ai-infrastructure": {
-      "last_reviewed": "2026-06-05",
+      "last_reviewed": "2026-06-15",
       "cadence": "weekly",
       "notes": "New projects added when announced; updates when status (announced/permitted/construction/operational) changes."
     },
     "/ai-crawler-access": {
-      "last_reviewed": "2026-06-02",
+      "last_reviewed": "2026-06-15",
       "cadence": "weekly",
       "notes": "Aggregate numbers refresh from the daily crawl via the live summary endpoint. This stamp tracks editorial review of the methodology and seed universe (worker/src/ai-crawler-access-seeds.ts)."
     },
     "/timeline": {
-      "last_reviewed": "2026-05-13",
+      "last_reviewed": "2026-06-15",
       "cadence": "monthly",
       "notes": "Historical record of AI ecosystem milestones. Monthly review for new entries."
     },
     "/models": {
-      "last_reviewed": "2026-06-05",
+      "last_reviewed": "2026-06-15",
       "cadence": "weekly",
       "notes": "Model catalog and pricing. Update when a flagship drops, when a tracked model is deprecated, or when a provider changes pricing. Reminder issue opens Mondays via weekly-models-check workflow."
     },
     "/ai-hardware": {
-      "last_reviewed": "2026-06-05",
+      "last_reviewed": "2026-06-15",
       "cadence": "weekly",
       "notes": "Hand-curated AI accelerator chip catalog in worker/src/ai-hardware.ts. Update when a new datacenter chip ships (NVIDIA, AMD, Google, AWS, Apple, Cerebras, Groq) or when specs are corrected. Reminder issue opens Mondays via weekly-ai-hardware-check workflow."
     },
@@ -38,7 +38,7 @@
       "notes": "Agent-only work directory and the ARB-backed agent surface. Refresh data/agents-directory.json when notable new agent products ship or when categories shift. Reminder issue opens every other Monday via biweekly-agents-directory-check workflow."
     },
     "/benchmarks": {
-      "last_reviewed": "2026-06-05",
+      "last_reviewed": "2026-06-15",
       "cadence": "weekly",
       "notes": "data/benchmarks.json. Add a row when a flagship lands or when benchmark rankings materially shift. Reminder issue opens Mondays via existing weekly-benchmarks-check workflow."
     },
diff --git a/data/timeline.json b/data/timeline.json
index 614af4c..84a02b6 100644
--- a/data/timeline.json
+++ b/data/timeline.json
@@ -28,6 +28,13 @@
       "category": "open-source",
       "description": "MiniMax launched M3, an open-weight coding and agentic model built on MiniMax Sparse Attention, which replaces full attention with KV-block selection and cuts per-token compute at 1M context to roughly one twentieth of the previous generation. M3 accepts text, image, and video input across a 1,048,576 token context window with up to 512K output tokens, priced at $0.30 per million input and $1.20 per million output. MiniMax reports 59% on SWE-Bench Pro and 83.5 on BrowseComp, though headline runs used its own infrastructure with agent scaffolding, so independent verification is pending. Weights and a technical report are due on Hugging Face within about ten days."
     },
+    {
+      "date": "2026-06-01",
+      "title": "NVIDIA Vera Rubin Chips Enter Full Production",
+      "provider": "NVIDIA",
+      "category": "hardware",
+      "description": "NVIDIA CEO Jensen Huang announced at Computex 2026 in Taipei on June 1 that Vera Rubin, the successor to the Blackwell GPU architecture, has entered full production. Each Rubin GPU delivers 50 petaflops of FP4 compute and 3.6 TB/s memory bandwidth, interconnected via sixth-generation NVLink. The rack-scale Vera Rubin NVL72 system pairs 72 Rubin GPUs with 36 custom Vera CPUs and aggregates 260 TB/s of NVLink scale-up bandwidth. Customer system deliveries are planned through major cloud providers including AWS, Google Cloud, Microsoft Azure, and Oracle in the second half of 2026, with Rubin marking the first post-Blackwell NVIDIA data center architecture in production at commercial scale."
+    },
     {
       "date": "2026-05-28",
       "title": "Anthropic Raises $65B Series H at $965B Valuation",
@@ -42,6 +49,13 @@
       "category": "model-release",
       "description": "Anthropic shipped Claude Opus 4.8 just six weeks after Opus 4.7, keeping the 1 million token context window. Anthropic's own measures put agentic coding at 69.2% (up from 64.3%) and knowledge work at 1890 (up from 1753), with the model roughly 4x less likely than 4.7 to let flaws in its own code slip through while using about 35% fewer tokens per task. A new fast mode runs roughly 2.5x quicker and about three times cheaper than before, and Dynamic Workflows adds large-scale parallel subagent support."
     },
+    {
+      "date": "2026-05-30",
+      "title": "SoftBank Commits Up to 75 Billion Euros to Build AI Data Centers in France",
+      "provider": "SoftBank",
+      "category": "infrastructure",
+      "description": "SoftBank Group announced at the 2026 Choose France summit hosted by President Emmanuel Macron that it will invest up to 75 billion euros to develop and operate AI data center capacity in France, one of the largest single AI infrastructure commitments in European history. Phase 1 commits 45 billion euros to deliver 3.1 gigawatts of capacity in the Hauts-de-France region by 2031, with sites in Dunkirk (Loon-Plage), Bosquel, and Bouchain. Total target is 5 gigawatts across the full program. Schneider Electric is the power modules manufacturing partner at the Port of Dunkirk; EDF supplies low-carbon electricity leveraging France's nuclear-heavy grid. The commitment marks SoftBank's strategic move from Vision Fund-style investing into owning and operating frontier AI compute infrastructure."
+    },
     {
       "date": "2026-05-21",
       "title": "Trump Postpones AI Executive Order Hours Before Signing",