← BACK TO LATEST ISSUE
★ POW! ★ ZAP! ★ WOW! ★ KAPOW! ★ BZZZT! ★ POW! ★ ZAP! ★ WOW! ★ KAPOW! ★ BZZZT! ★ AUTO-PUBLISHED VIA REVDOKU

AI DAILY COMICS

TODAY’S TOP AI NEWS — ILLUSTRATED
THURSDAY · JUNE 25, 2026 · ISSUE #12
#1
ANTHROPIC PAYS $1.25B/MONTH FOR SPACEX COLOSSUS — THE $45B COMPUTE MEGA-DEALCOMPUTE
THE MEGA-DEALPAY TO:SPACEX / xAI$1,250,000,000…per month. Every month.SIGNED BY: ANTHROPIC220,000+ NVIDIA GPUsH100 / H200 / GB200 mixTHROUGH MAY 2029$45B+ total commitmentLargest compute deal ever
COLOSSUS 1MEMPHIS, TENNESSEE300+ megawatts capacityxAI MOVED TO COLOSSUS 2Was running at only 11% use!ANTHROPIC TAKES 100%Claude needs ALL the GPUs
COMPUTE WARWHO PAYS SPACEX MOST?ANTHROPIC $1.25B/moGOOGLE $920M/moBoth locked in for 3 years$2.17B / MONTHSpaceX collects from AI labsELON WINS EITHER WAYAI labs pay rivals for computeNo choice: no one else has GPUs
THE STORYAnthropic has contracted to pay SpaceX $1.25 billion per month — over $45 billion total through May 2029 — to lease all of the Colossus 1 data center in Memphis, Tennessee, with over 220,000 NVIDIA GPUs. xAI had migrated its own training to the newer Colossus 2 facility, leaving Colossus 1 running at just 11% utilization. Google separately signed a $920M/month deal for Colossus compute to power Gemini Enterprise, bringing SpaceX’s combined monthly AI revenue to $2.17 billion.
Sources: TechCrunch · Anthropic
· · · NEXT STORY · · ·
#2
OPENAI LAUNCHES GPT-ROSALIND — FIRST AI AGENT TO CO-AUTHOR PUBLISHED DRUG RESEARCHSCIENCE
NEW MODELOPENAIROSALINDPURPOSE-BUILT FOR:DRUG DISCOVERYCHEMmoleculesPROTEINengineeringGENOMEanalysisAMGEN · MODERNA · ALLEN INST.Thermo Fisher · more partners
HISTORY MADEWORLD FIRST:AI AGENT IN PEER-REVIEWEDmedicinal chemistry journalJ. MEDICINAL CHEMISTRYAutonomous AI co-author, 2026WHY IT MATTERS:AI helps design drugs, not justread papers about them
THE SPEEDUPBEFORE AI10–15 YEARSDiscovery to approval: very slowWITH GPT-ROSALINDGOAL: MONTHSAI handles target to moleculeALSO: GPT-5.5-CYBERHighest CyberGym scoreever recorded by any model
THE STORYOpenAI unveiled GPT-Rosalind, a purpose-built AI model for drug discovery and pharmaceutical research, partnering with Amgen, Moderna, and the Allen Institute. An autonomous OpenAI AI agent became the first ever to contribute to a published medicinal chemistry study in the Journal of Medicinal Chemistry. Separately, OpenAI also launched GPT-5.5-Cyber, claiming the highest CyberGym security benchmark score ever recorded.
Sources: PYMNTS · Drug Target Review
· · · NEXT STORY · · ·
#3
FABLE 5 BAN ENTERS WEEK 2 — DARIO REFUSES TO BUDGE, WHITE HOUSE HOLDS FIRMPOLICY
HOW WE GOT HEREJUN 9: FABLE 5 LAUNCHESWorld celebrates new AIJUN 10: AMAZON ALERTSJassy flags jailbreak to govJUN 12: FULL BANExport control - offline globallyJUN 13-24: STANDOFFDario refuses to fix jailbreakJUN 25: WEEK 2Still no deal in sight
THE SHOWDOWN🧔DARIO“Not our fault.”VS🏠WHITE HOUSE“Fix it or stay off.”CORE ISSUE:Fable 5 can describe cyberattacksNEW: SHARED FRAMEWORKGov + Anthropic co-writingrules for AI intervention
WHAT’S NEXT?SCENARIO A ✅Anthropic patches guardrails,Fable 5 comes back onlineSCENARIO B ❌Standoff drags on for months,Fable 5 stays dark globallyTHE BIG PICTURE:First time a government hasswitched off a major AI modelNEW GLOBAL PRECEDENT
THE STORYThe standoff between Anthropic and the White House over Fable 5 entered its second week with no resolution. The US government’s export control directive took Fable 5 and Mythos 5 offline globally on June 12 after Amazon CEO Andy Jassy flagged a security jailbreak. CEO Dario Amodei refused to patch it. The two sides are reportedly co-writing a framework governing when the US government may intervene to disable AI models — a historic first.
Sources: Anthropic · Nat’l Law Review