Bloomberg reported on 22 September that SpaceX's Grok Bot reached 418,000 weekly users in the week ending 14 September, up 24% on the week before, from a presentation SpaceXAI staff gave in London that Bloomberg viewed and nobody has published. Ed Ludlow, who wrote it, called the number "a sign of modest but early demand." Four hours later ARK's Brett Winton called the same number the most successful app launch in history — within a class, "products that have no trial/freemium tier."
Grok Bot is not in that class, and neither is the product Winton ranks it against.
Nobody bought Grok Bot, because there is nothing to buy
The three plan cards above are Grok Bot's own, and every one is a Cursor or Grok subscription listing Grok Bot among its inclusions. Asked "How much does Grok Bot cost?", the page answers: "Grok Bot subscriptions come with weekly usage included, with additional usage billed based on token cost." Cursor's help centre is blunter: "Do I need a separate Grok Bot subscription? No." A Grok Bot user is someone whose existing plan grants Bot usage, and no part of what the 418,000 pay is a Grok Bot price.
Cursor's help centre documents the trial Winton says does not exist
From that same help centre: "The Grok Bot free trial is a usage credit rather than a set number of days, though a 7-day window also applies." On cancelling, "there is no charge to cancel, and the Grok Bot trial never becomes a paid plan" — and the button that ends it "appears on free-plan accounts." Grok Bot's call to action reads "Get started for free."
Then the fortnight. On 3 September SpaceXAI announced Enterprise customers "have free usage for the next two weeks, and can invite their whole organization, including people without an existing seat." That window runs 3 to 17 September; the measured week ends inside it.
The cheapest way in fell from $60 to $20 sixteen days before the measured week
The 11 August launch post has been quietly rewritten three times, tracking who could reach the product. On 11 August, and still on 20 August, it read: "available today for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium subscribers." By 25 August it had added SuperGrok Plus, Cursor Pro+ and Teams Standard; by 12 September, plain SuperGrok and Cursor Pro.
Prices followed. On 25 August the cheapest card on the page was Cursor Pro+ at $60 a month; by 29 August it was Cursor Pro at $20, a two-thirds cut in the entry price 16 days before the week that grew 24%. Growth measured across a distribution change is not a demand rate.
ElevenLabs launched with a free tier and never published a first-month paid count
Winton's comparator opened its beta in January 2023. Its Series A post of 20 June 2023 says it "has amassed over 1 million registered users" — registered rather than paid, five months in rather than one. No first-month paid figure for ElevenLabs exists, so the comparison cannot be run in either direction, and the class is empty on both sides.
418,000 is 22% of the Grok subscriber base SpaceX has disclosed
The one disclosed denominator is in SpaceX's S-1: "As of March 31, 2026, we reached approximately 6.3 million active paid subscribers... approximately 1.9 million SuperGrok, SuperGrok Heavy and SuperGrok Lite paid subscribers." So 418,000 ÷ 1.9m = 22.0% — and SuperGrok Lite does not get Grok Bot, making the eligible base smaller and the real share higher.
Cursor is the larger half and sits outside that count: SpaceX bought it on 14 August, after the 31 March date. Estimating it takes three steps, the last ours. Cursor was reported annualising above $4 billion before the close; $4B ÷ 12 = $333m a month; at an assumed $80 blended monthly price that is 4.17m paid seats, making 418,000 10.0% of them.
The $80 is our assumption and the softest rung here, so take it three ways. At $40 the seat count doubles to 8.33m and Grok Bot's reach halves to 5.0%; at the $20 Pro list price, 16.7m seats and 2.5%. Across that range Grok Bot sits on a minority of an already-paying base.
$100.3m a year is the ceiling, and these users already pay
The strongest case the other way: attach on a paying base is how a new product should land, and 24% a week is what matters. Take the generous version: if all 418,000 were incremental $20 seats, which none is, 418,000 × $20 × 12 = $100.3m a year — 2.5% of Cursor's run rate, 0.32% of second-quarter revenue annualised. On the growth rate, we have already run that argument on this ticker.
One unpublished figure would turn 418,000 into revenue: on-demand token billing after the weekly grant runs out. A Q3 print or 8-K breaking out Grok Bot paid seats or overage settles this, and could overturn it. So could the slide — no SpaceXAI or Cursor page defines "weekly users", and the 3 September post counts "millions of bots created" against 418,000 users. If the figure is cumulative registrations rather than one week's actives, the level is wrong as well as the class.
Bloomberg (Ed Ludlow, 22 September 2026) and Ludlow's post that afternoon report the user count, the weekly change and the London presentation; the presentation is unpublished and was viewed by Bloomberg, not by us. Plan prices, inclusions, the cost FAQ and the "Get started for free" line come from SpaceXAI's Grok Bot and pricing pages captured 20 September 2026, the trial and "no separate subscription" answers from Cursor's plans-and-billing help centre, the Enterprise fortnight from SpaceXAI's 3 September post; the August and September launch-post wording and price cards are archived captures of those same pages. Subscriber counts are from SpaceX's S-1 as of 31 March 2026, revenue from the Q2 2026 shareholder letter, ElevenLabs figures the company's own of 20 June 2023, and Cursor's $4 billion run rate press-reported. Every share, conversion and seat estimate is ours, including the $80 blended price behind the seat count.
