Two of the biggest AI-adjacent finance stories of 2026 dropped in the same window, and most coverage is treating them as separate events. They’re not. SpaceX’s record-breaking IPO trading as SPCX on Nasdaq and Anthropic’s first profitable quarter since founding in 2021 are connected by one specific deal buried in SpaceX’s own prospectus, and once you see it, the headlines make a lot more sense.
So let’s get into what actually happened, what’s real versus spun, and what it means if you’re tracking either company.
SpaceX completed the largest IPO in US history on June 12, 2026, pricing at $135 a share and debuting at $160.95 on Nasdaq under the ticker SPCX, a 19% first-day jump that briefly valued the company near $2.1 trillion. Days earlier, reports surfaced that Anthropic expects its first profitable quarter since founding, with projected Q2 revenue of $10.9 billion and an operating profit of roughly $559 million. The link between them: SpaceX’s prospectus revealed that Anthropic agreed to pay $1.25 billion a month through May 2029 for compute capacity at SpaceX’s Colossus 1 data center in Memphis — a deal that’s currently running at a discounted ramp-up rate, which happens to land exactly in the quarter Anthropic is calling profitable.
SpaceX’s IPO, by the numbers
Here’s what nobody disputes: SpaceX filed its S-1 with the SEC on May 20, 2026, confirmed Nasdaq as its exchange and SPCX as its ticker, and skipped the usual price-range dance that most IPOs go through. It went straight to a fixed price of $135 a share. Pricing happened June 11, the stock started trading June 12, and it closed that first day at $161, up 19% from the offer price. Total raised: somewhere around $75 billion, which is what makes the “largest IPO in US history” claim hold up — it’s not marketing copy, multiple outlets independently confirmed it.
Elon Musk became the world’s first trillionaire based on his combined Tesla and SpaceX stakes, and the IPO put him in charge of two separate trillion-dollar public companies at once. That’s a genuinely unusual position for one person to be in, and it’s worth sitting with for a second before moving on to the numbers.
What surprised a lot of people watching this closely: SpaceX’s prospectus disclosed a total accumulated loss of $41.3 billion since the company was founded in 2002. Musk has said on record that SpaceX has been cash-flow positive since around 2015, and the only consistently profitable division today is Starlink, the satellite internet business. Rockets and Starship development are not where the money comes from — they’re the mission, but Starlink pays the bills.
Then there’s the xAI complication. SpaceX folded its xAI subsidiary — and with it, the Grok models, the X platform, and X’s advertising business — into the company structure in February 2026. That merger is part of why this filing reads less like a pure aerospace prospectus and more like three companies stapled together. X’s advertising revenue actually fell by $100 million in the first quarter as the platform overhauled its ad tech, a sharp contrast to Meta, Pinterest, and Reddit, which all posted ad growth in the same period. Subscription revenue across X and Grok did better — up $365 million in 2025 and another $177 million in Q1 2026 alone, partly from people paying for Grok directly. If you’re trying to track Grok’s actual cost and access limits as a product rather than as a line item in a 400-page filing, our breakdowns of Grok’s free limits and Grok’s voice mode setup are more useful than anything in the prospectus.
The stock itself has been a wild ride since debut, which matters if you’re watching it rather than just reading about it. It hit an all-time high of $225.64 on June 16, then pulled back. As of June 18, it’s trading in the $190s, down from that peak but still well above the $135 IPO price. Crypto-native traders were pricing this in advance, too — perpetual futures contracts tied to SPCX were trading on platforms like Hyperliquid and BitMEX days before the actual listing, which tells you how much speculative appetite was built up ahead of time.
Why Anthropic’s first profitable quarter matters right now
Here’s the part that actually connects to SpaceX. Anthropic — the company behind Claude — told investors as part of an ongoing funding round that it expects second-quarter revenue of $10.9 billion, more than double the $4.8 billion it posted in Q1. Out of that, it’s projecting an operating profit of around $559 million. If that holds, it would be Anthropic’s first profitable quarter since it was founded in 2021. The Wall Street Journal broke the numbers, and CNBC and the Financial Times both independently confirmed them through their own sources.
That’s a real milestone if it lands. Revenue more than doubling quarter over quarter is rare at this scale — faster than Zoom grew during the pandemic, faster than Google or Facebook grew heading into their own IPOs, according to the reporting. Anthropic is also reportedly closing a $30 billion funding round that would push its valuation past $900 billion, narrowly ahead of OpenAI’s roughly $852 billion private valuation. For context on how unusual that is: OpenAI posted around $5.7 billion in Q1 revenue, more than Anthropic’s $4.8 billion, but has told investors it doesn’t expect actual profitability until 2030, after spending more than $600 billion building out infrastructure. xAI, meanwhile, reportedly lost $6.4 billion on $3.2 billion in revenue last year. Anthropic getting to “profitable” first, even briefly, is genuinely notable in that company.
Now here’s the part that should make you slow down before treating this as proof Anthropic has cracked the AI profitability problem. The company itself told investors it does not expect to stay profitable in the quarters after Q2. That’s an unusual thing to volunteer — most companies announcing a first profitable quarter don’t immediately add the footnote that it’s a one-off. Anthropic did exactly that, up front, which is either refreshing transparency or a tell that the number needs an asterisk. Probably both.
The compute deal that ties both stories together
This is the detail most coverage of either story missed individually. Buried in SpaceX’s IPO prospectus is confirmation that Anthropic agreed to pay SpaceX $1.25 billion a month, through May 2029, for compute capacity at SpaceX’s Colossus 1 data center in Memphis, Tennessee. That’s roughly $15 billion a year at the steady-state rate. SpaceX is essentially acting as a neocloud here, renting out xAI’s data center infrastructure to a direct AI competitor — which is its own strange dynamic, but not the focus today.
The part that matters for Anthropic’s “profitable quarter” claim: that $1.25 billion-a-month rate doesn’t kick in immediately. Anthropic is paying a discounted ramp-up rate during the deal’s early months, and those early months happen to be Q2 2026 — the exact quarter Anthropic is telling investors was profitable. Tech journalist Ed Zitron published a detailed critique arguing the operating profit figure is largely a function of that temporary compute discount, not a structural improvement in Anthropic’s economics, and that the company’s costs have historically scaled almost linearly with revenue — meaning once the full rate applies, the profit picture likely reverses.
That’s a fair point worth taking seriously, and it’s not an isolated take — multiple analysts have raised the same flag. But it’s not the whole story either. Anthropic’s revenue growth — from roughly $87 million annualized in January 2024 to north of $30 billion annualized by April 2026, according to figures tracked across multiple independent reports — isn’t something a compute discount can manufacture out of thin air. The revenue is real. Whether the profit is a preview of where the business is heading, or a one-quarter accounting artifact made possible by a temporary discount, depends entirely on what happens once the full $1.25 billion-a-month rate applies starting sometime after the ramp period. Nobody outside Anthropic’s finance team knows that yet, and the company itself isn’t required to follow public-company accounting standards, since it hasn’t gone public.
I’ve watched enough “first profitable quarter” press cycles over the years to know the pattern: the number gets leaked right as a funding round or IPO conversation is heating up, and it always comes with favorable timing. That doesn’t make it false. It just means you read it as one data point, not as a verdict.
What this actually means if you’re not a hedge fund
If you’re trading SPCX, here’s the honest version: this is one of the most volatile new listings in recent Nasdaq history, and treating any single day’s price action as signal is a mistake. The stock swung from a $135 IPO price to $225.64 within four trading days, then dropped back into the $190s. Prediction markets like Polymarket were pricing bets on whether SpaceX would close above a $2 trillion market cap on debut day alone — that’s the level of speculative intensity around this name right now. The next scheduled earnings call is expected in September 2026, and the lockup period — when early investors and employees can start selling — expires in December 2026. Both of those dates are realistic checkpoints for when the speculative froth either gets validated by real numbers or doesn’t. None of this is investment advice; it’s just what’s publicly known and worth knowing before you act on a headline.
If you’re evaluating AI providers rather than stocks, the more useful takeaway is different: the economics behind frontier AI labs are getting genuinely strange, and it’s worth understanding why before you build a business that depends entirely on one provider’s pricing staying stable. When a company’s “profitability” hinges on a temporary compute discount from a deal that converts to a $15-billion-a-year obligation in 2029, that’s a signal the underlying cost structure is still unsettled. Prices and rate limits on frontier models can and do shift when the economics behind them shift.
That’s part of why a lot of builders hedge by keeping a local or open-weight option in their toolkit instead of routing everything through one frontier API. If you’re already working in an editor like Cursor, our guide to running local models with Cursor AI walks through setting that up as a fallback. The same logic applies on the image side — tools like Venice AI’s free tier offer a different pricing and privacy model than the big labs, and there’s a reason interest in self-hosted setups, including uncensored local image AI models and private image generation setups, keeps climbing every time a major lab’s pricing or access policy shifts. None of these are about avoiding frontier models entirely. They’re about not having a single point of failure when the company behind your stack is one compute renegotiation away from a different cost structure.
As for Grok specifically — now sitting inside a public company’s financial filings for the first time — that visibility cuts both ways. You get real numbers on its ad and subscription revenue, but you also see how much of its economics still run through SpaceX’s broader balance sheet rather than standing on its own. If you’re actually using Grok day to day, the free limits and alternatives breakdown we put together is more practical than anything in a 400-page S-1.
Watch the September earnings call for SpaceX and whatever Anthropic discloses about its actual Q2 results, not the projections. Projections shared with investors during a funding round are a sales document. Audited or even informally confirmed actuals are the thing that tells you whether either story holds up past the headline.
(SpaceX’s S-1 filing is publicly available through SEC EDGAR if you want to read the primary source instead of secondhand coverage.) For more coverage like this, the homepage at thebizaihub.com tracks how these AI industry shifts play out for people actually building with these tools.
