Four days after closing the largest US IPO in history ($2T cap, $135 → $199 in 3 days), SpaceX exercised the April option on Anysphere: acquire Cursor at $60B or pay $10B for partnership. Cursor hit $2B ARR in March 2026, doubling in 90 days — the fastest SaaS ramp in history (Slack took 7 years to $1B). The deal plugs Cursor's coding agent into Grok, explicitly to close the gap with Claude Code and GPT-5.4 on the frontier-model coding market.
Cursor's Series D in Nov 2025 closed at $29.3B post-money on $1B ARR. By March 2026, ARR had doubled to $2B. At SpaceX's $60B purchase price, the implied revenue multiple moved from ~29x to ~30x — a *flat* multiple on a doubled revenue base. The market is paying for growth velocity, not profit margins.
*Mar 2026 implied value uses 30x trailing revenue from June's deal multiple, applied retroactively to $2B ARR. No new round was priced.
SpaceX didn't surprise Cursor with an offer — it had a binding April option: acquire at $60B or pay $10B for partnership. SpaceX chose to own. The Memphis data center complex (xAI) was already powering Cursor's training. Acquiring collapses the partnership premium into equity and gives Grok direct access to Cursor's coding-agent surface.
Jeff Bezos is back in an operating role for the first time since 2021, co-CEOing a physical-AI startup with Stanford physicist Vik Bajaj. The $12B round closed Thu Jun 12 from JPMorgan, BlackRock, Goldman Sachs, DST Global, Arch Venture — the traditional capital allocators placing their largest AI bets not on chatbots but on design-to-manufacturing AI for jet engines, medical devices, and rockets. Talent is pulled from OpenAI, Google DeepMind, and NVIDIA.
Not robotics. Bezos explicitly denied building physical robots. This is software-AI for engineering — compressing the design-to-manufacturing cycle. Combined with NEURA Robotics' $1.4B Series C the same week, "physical AI" is now a distinct mega-round category with its own LP base.
At $60B / 30x trailing revenue, Cursor sets the new floor for AI coding tool valuations. The next-round comp moves for every competitor — public and private.
*Implied next-round values at 0.5–0.55x Cursor's multiple. No public pricing for these names.
The most radical Search redesign in Google's history. The query box is gone. Users now interact with "information agents" that monitor the web 24/7 and proactively push updates. Gemini 3.5 Flash powers the new AI Mode, which has passed 1 billion monthly users. Queries in AI Mode already average 3x longer than traditional searches — the surface is shifting from retrieval to agent-style reasoning.
The April meeting had 3 dissents — not on the rate, but on the easing-bias language. Warsh (confirmed 54-45, sworn in May 22) is expected to (1) formally drop easing-bias wording, (2) skip his own dot plot on principle, and (3) deliver his first press conference with a more restrictive tone. April CPI at 4.2% YoY (3-year high, gas +23.5%) gives him cover.
Polymarket shows only 99% hold on the *June* decision, but the curve has repriced: 41% hold-through-year, 42% one hike, 14% two hikes (Jun 16). The H2 2026 macro trade is hawkish re-anchor, not recession.
Why this matters for TTL: The capital allocators have sorted AI into distinct sub-markets with their own multiples, LP bases, and exit paths. Frontier models trade on cap-table (SPCX, Anthropic, OpenAI). Coding agents trade on growth velocity (Cursor at 30x). Physical AI trades on JPM/BlackRock/GS balance-sheet capacity. Agent infra trades on bundle economics (Cloudflare's Workers + R2 play). Regulated-AI trades on pain-point specificity. The companies that pick one lane and dominate it will be the next $10B+ winners.
With SPCX live, OpenAI having confidentially filed (Jun 8), and Anthropic S-1 expected H2, the public market is being asked to underwrite the entire frontier-model stack within 6 months.