OpenAI Files for IPO, Lovable Hits $500M ARR, and the Tokenpocalypse Looms

June 2026 is shaping up like no one predicted. Three of the most important tech companies in the world — OpenAI, Anthropic, and SpaceX — are all barreling toward IPOs within months of each other. The vibe-coding startups are printing revenue. And the bill for all those subsidized AI tokens? It just came due.

Let’s spaghetti this.

OpenAI Goes Public (Confidentially)

OpenAI filed confidentially for an IPO yesterday, riding the wave Anthropic kicked off just over a week ago. The filing is a draft registration with the SEC — no specifics on timing, pricing, or share structure yet.

But here’s the thing: OpenAI is burning cash faster than it can make it. The company raised a record $122 billion in March (including $3 billion from retail investors via bank channels), yet projects it will spend $85 billion in 2028 even after doubling sales. By its own math, the business won’t generate more cash than it spends for at least four more years.

And the timing is complicated. OpenAI recently missed internal targets for new users and revenue, per the Wall Street Journal. CFO Sarah Friar has reportedly raised concerns about whether the firm can sustain its data center spending.

In a surprising move, OpenAI also published a sweeping philosophical statement about AGI and its mission to benefit all of humanity — the kind of public-market-pre-IPO communication that companies in a quiet period normally avoid. That tells you something about the current regulatory posture: the SEC under this administration is hands-off, and OpenAI is reading the room.

The numbers that matter:

  • Last valuation: $852 billion post-money
  • 2028 projected burn: $85 billion
  • 2028 compute spending alone: ~$122 billion
  • Closest peer (Anthropic): near first quarterly profit, valued at $65B+

The billionaire’s IPO trifecta — OpenAI, Anthropic, and SpaceX ($1.75T) — is a concentration of high-stakes offerings the market hasn’t seen since dot-com.

Futuristic AI coding terminal in a dark server rack room with holographic code

Lovable Crosses $500M ARR: The Vibe-Coding Empire Grows

While the giants jockey for public offering slots, Lovable just announced it’s blown past $500 million in annualized revenue run rate. Up from $400M in February. The company was founded in late 2023 — it hasn’t even hit its third birthday.

The numbers are genuinely absurd:

  • 1 million new projects per week
  • 50M+ projects built total
  • Primary users are non-technical (founders, designers, salespeople)
  • They’re building CRMs, inventory systems, HR platforms, e-commerce storefronts

The narrative Lovable is pushing is the so-called SaaSpocalypse — why buy expensive SaaS annual contracts when vibe-code it yourself? Of course, the hard question nobody can answer yet: who maintains this stuff six months from now? Software is a living thing. Dependencies shift, APIs deprecate, things break.

Lovable will need to start reporting abandoned project rates as the platform matures. If those stay low, that’s the real signal that the SaaS disruption thesis holds. If they spike, the “build it yourself” dream hits the maintenance wall.

The Tokenpocalypse Is Here

This is the one that’s making devs sweat right now. Microsoft jacked up GitHub Copilot pricing — shifting from flat-rate to per-token pricing — and enterprise accounts are getting walloped. One Reddit user’s company started calling it the “Tokenpocalypse”.

The dynamic is simple and brutal: the AI ecosystem has been massively subsidized by investor dollars. Companies charged $20/month for ChatGPT and called it a business model. They called it a price point when “Let’s spit out a number” was closer to the truth.

Now Anthropic and OpenAI are prepping IPOs where they need to show a path to profitability. Enterprise customers like Uber are already “tokenmaxxing” one month and capping usage the next. The cost curve is still going up while the tech improvement curve — though steep — hasn’t caught up to sustainable unit economics.

The big question every investor will ask at the IPO roadshow: Can AI labs collapse costs fast enough to meet customers where they are on willingness to spend? Because right now, those two lines are diverging.

Related reading on the Copilot changes: the full Equity podcast breakdown is worth the listen if you’re budgeting your team’s AI spend.

WWDC 2026: Apple Bets Everything on Privacy-First AI

Meanwhile, Apple held its WWDC keynote. Tim Cook’s last hurrah before handing the CEO role to hardware chief John Ternus on September 1. And the star of the show was Siri’s AI overhaul.

The big reveal: Google Gemini is now under the hood of Siri. Apple’s Craig Federighi hammered the privacy message: “data is only used to execute your request, and outside experts can verify this at any time.”

The strategy is clear — Apple isn’t chasing raw model performance. It’s positioning as the “most private” AI ecosystem, full stop. Siri upgrades include:

  • Standalone Siri app (not just voice assistant)
  • Visual intelligence capabilities
  • Liquid Glass opt-ins (dial back the design, not just turn it on)
  • AI-powered reply suggestions in Messages
  • Cross-app context awareness in Phone, Safari tab management

Whether “most private” wins against “most capable” remains the defining consumer AI battle of the year. After last year’s $250M false ad settlement, Apple seems determined to let the tech speak louder than the marketing.

Split scene of smartphone with privacy shield next to abstract neural network with padlocks

Also Worth Noting

  • Worldcoin lays off staff even as OpenAI files for IPO — Sam Altman’s other moonshot isn’t immune to cost-cutting
  • **SBF filed for a pardon](https://techcrunch.com/2026/06/08/sam-bankman-fried-applies-for-a-pardon-from-trump/) from Trump — the convicted FTX founder is serving 25 years (officially on file now)
  • **Mercor’s founder called out Sequoia](https://techcrunch.com/2026/06/08/mercors-brendan-foody-calls-out-sequoia-over-dual-pricing-valuation-tricks/) for “dual-pricing” VC valuation games — the gloves are coming off
  • GitHub trending is dominated by agent tools: last30days-skill (36K stars), Agent-Reach (25K), whichllm and turbovec all surging

The Takeaway

We’re entering the accounting phase of AI. The subsidized token spree, the $20/month dream pricing, the “we’ll worry about unit economics later” mentality — it’s all running into the IPO calendar. OpenAI and Anthropic have to show shareholders a path to profitability. Enterprise customers feel the squeeze. Devs are rejiggering budgets.

And right in the middle of all this, Lovable proves that non-technical people will happily pay $500M/yr to build software without writing a single line of code. The demand is real. The question is whether the economics of serving it can survive the end of the free-money era.

Disclosure: Links to Amazon or AI tools may include affiliate links (e.g., ?tag=spaghettistor-20). As an Amazon Associate I earn from qualifying purchases. This doesn’t affect my editorial independence.

If you want this in your inbox every morning, subscribe to the Buttondown newsletter.