AI News Roundup — June 11, 2026
SpaceX priced the largest IPO in history at $135/share Wednesday evening — SPCX begins trading on Nasdaq Friday at a $1.77 trillion valuation — while Anthropic and Microsoft confirmed early talks to run Claude on Microsoft's custom Maia 200 chips, and OpenAI deepened its Oracle partnership to let enterprise customers access frontier AI models directly through existing Oracle Cloud credits.
SpaceX Prices the Largest IPO in History at $135/Share — SPCX Trades Friday
SpaceX officially priced its IPO Wednesday evening at $135/share — a fixed price, no book-building range — raising $75 billion at a $1.77 trillion valuation, the largest initial public offering in recorded history by deal size, nearly 2.5x Saudi Aramco's 2019 record. Demand reached $250 billion at 3.5–4x oversubscription, with over $10 billion in institutional orders. MSCI confirmed it will apply early inclusion treatment, meaning every passive fund tracking MSCI indexes will be a forced buyer of SPCX after listing — structural demand independent of fundamental valuation. The 366-day insider lockup is longer than standard, limiting initial supply. Notably, Elon Musk will not sell a single share, and a dual-class structure preserves his full voting control. Morningstar's independent valuation: $780 billion — 44% of the IPO price. Trading opens on Nasdaq under ticker SPCX on Friday, June 12.
AI Tools Recap · Yahoo Finance · Reuters
Anthropic and Microsoft in Early Talks to Run Claude on Maia 200 Custom Chips
CNBC and Reuters confirmed Anthropic and Microsoft are in early-stage talks — no deal closed — for Claude to run inference workloads on Microsoft's Maia 200 custom AI accelerator chips via Azure. Maia 200 launched in January 2026 on TSMC's 3nm process with 216GB HBM3e memory and 10+ petaflops FP4 performance, claiming 30%+ better inference cost per dollar than competing silicon. It currently runs OpenAI's GPT-5.2 internally. For Anthropic, closing the deal would add a fourth custom silicon option alongside Nvidia GPUs, AWS Trainium, and Google TPUs — reducing per-token costs at exactly the moment its unit economics will face scrutiny from public market investors ahead of an expected October 2026 IPO. For Microsoft, it would make Maia 200 the first chip externally validated by a frontier AI lab that isn't its own partner OpenAI. Risks: FP8 quantization may reduce output quality by 0.5–1.5% on some tasks; talks have been ongoing for three weeks without closing.
OpenAI Models Now Accessible via Oracle Cloud Credits — Stargate Distribution Play
OpenAI officially announced that enterprise customers can access its frontier AI models and Codex through existing Oracle Universal Credits, removing the need for a separate procurement relationship. Companies with multi-year Oracle Cloud Infrastructure commitments — often running into the tens or hundreds of millions — can now route AI workloads to OpenAI models without a new vendor contract. The move transforms the Stargate project's $500 billion infrastructure commitment (Oracle building data centers across the US) into an enterprise distribution channel. Procurement friction — not model quality — is the real barrier to enterprise AI adoption, and Oracle's deep penetration in Fortune 500 contracts gives OpenAI access to accounts that might otherwise build on Azure or AWS.
Claude Fable 5 Holds Up in Real-World Developer Testing — Free Until June 22
Two days into developer availability, Claude Fable 5's 80.3% SWE-Bench Pro benchmark is holding up in real-world coding tasks. Developers are reporting significant improvements on hard multi-file refactors, autonomous debugging sessions, and complex codebase analysis compared to Opus 4.8. Extended autonomous work sessions — Fable 5's stated strength — are showing in longer uninterrupted coding runs before needing human input. The safety classify routing (cybersecurity and biology queries falling back to Opus 4.8) is triggering in under 5% of sessions. API developers note the refusal returns HTTP 200 with stop_reason "refusal" and is not billed if no output was generated. Critical deadline: Fable 5 is free on Pro, Max, Team, and Enterprise plans until June 22 only — usage credits required after June 23.
ChatGPT Crosses 1 Billion Monthly Active Users — Claude Growing 640% Year-Over-Year
ChatGPT became the fastest application in history to reach 1 billion global monthly active users, crossing the milestone in May 2026 — roughly three years after its November 2022 launch, outpacing Google Maps, TikTok, Instagram, and YouTube. As of Q2 2026, ChatGPT has 900 million weekly active users while Claude has 56 million global monthly active app users. But the growth trajectories tell a different story: ChatGPT's MAU growth is 62% year-over-year, while Claude's is approximately 640%. Sensor Tower also found that US ChatGPT users who installed the Claude app in Q1 2026 spent 5% less time on ChatGPT one month later — a small but measurable attention shift. One billion users is not a ceiling; it is a floor. The more consequential metric is Claude's trajectory.
DeepSeek V4 Pro Makes 75% Price Cut Permanent — Frontier AI Pricing War Intensifies
DeepSeek made its 75% price discount on V4 Pro permanent this week, after the promotional period was set to expire. Permanent pricing: $0.003625–$0.87 per million tokens, down from $0.0145–$3.48. For context: OpenAI's GPT-5 charges $2.50/$10 per million tokens, Anthropic's Opus 4.7 charges $5/$25, Google's Gemini 3.5 Flash charges $0.15/$0.60. DeepSeek V4 Pro now sits below every major competitor on a per-token basis, runs on a 1M-token context window, and is built on Huawei's Ascend 950 chips — reducing its exposure to Nvidia supply constraints and US export controls. DeepSeek's explanation: "V4 Pro runs at roughly a quarter of the single-token compute and a tenth of the memory footprint of its predecessor at very long context. This is why the price cut is permanent rather than promotional."
Amazon Signs $17.5 Billion AI Infrastructure Financing Deal
Amazon has signed a deal to borrow $17.5 billion from banks, adding to the mounting evidence that the AI infrastructure race is being financed at a scale that has no modern precedent in enterprise technology. The financing is part of Amazon's broader AI buildout, which includes AWS data center expansions, investments in Anthropic (now heading toward IPO), and its own proprietary Trainium chip development. The sum is notable even against Amazon's scale — $17.5 billion in a single financing tranche suggests the infrastructure requirements for the next generation of AI training and inference are outpacing even the largest tech companies' operating cash flows.