Anthropic turns profitable, xAI burns billions, Nvidia eyes $200B agent market
Anthropic's about to print money while xAI bleeds $6.4B — today's AI is a tale of two very different balance sheets.
Subscribe free →Anthropic Is About to Print Money — Literally
Anthropic told investors it will more than double revenue to $10.9 billion in Q2 2026 — its first profitable quarter ever. This comes alongside the bombshell reveal from SpaceX's IPO filing that Anthropic is paying xAI $1.25 billion *per month* for compute access at the Colossus data centers in Memphis. That's $15 billion a year flowing from Claude's creator directly into Elon Musk's pocket.
Why it matters: Anthropic hitting profitability resets the narrative that frontier AI labs are just cash incinerators — and makes the September OpenAI IPO look a lot more credible too.
xAI Lost $6.4B Last Year and Is Speeding Up
SpaceX's IPO filing cracked open Elon Musk's AI finances for the first time: xAI burned $6.4 billion in 2025 and is doubling down — committing $2.8 billion to natural gas turbines for its data centers over the next three years, even while being sued over its existing generators. The filing paints a picture of a company spending first and asking questions never.
Why it matters: This is the first hard look at xAI's financials, and the scale of losses combined with accelerating capex signals Musk is treating Grok as a must-win war regardless of cost.
Jensen Huang Found a $200B Market Nobody Was Watching
At Nvidia's earnings call — where the company posted another record quarter at $81.62B in revenue and guided Q2 to $91B — Jensen Huang unveiled his next target: CPUs for AI agents. He's betting $200 billion on Nvidia's Vera chip becoming the backbone of agentic computing, a market that barely existed two years ago.
Why it matters: Nvidia isn't just riding the GPU wave anymore — Huang is actively creating new markets, and if agents scale the way he's betting, Vera could be the next trillion-dollar product line.
OpenAI Solved an 80-Year-Old Math Problem. This Time It's Real.
OpenAI says its reasoning model disproved a geometry conjecture that's been open since 1946 — and crucially, the same mathematicians who publicly embarrassed OpenAI last time by debunking its false math claim are now vouching for this one. The result has been independently verified by experts in the field.
Why it matters: Verified mathematical breakthroughs — not just impressive-looking outputs — are the clearest signal yet that reasoning models are crossing into genuine scientific contribution territory.
Someone Raised $700M for a Startup Nobody Can Describe
Brett Adcock — the serial founder behind Figure AI and Archer Aviation — raised a $700M Series A for Hark, valuing it at $6 billion. The pitch: a "universal AI interface." The details: almost none. Adcock has a track record of raising massive rounds on vision alone, and investors are clearly betting that pattern holds.
Why it matters: A $6B valuation for a product in stealth mode is either the most audacious bet in AI right now or a perfect encapsulation of how frothy this market has become — possibly both.
Meta Is Laying Off Thousands to Pay for AI
Meta quietly notified thousands of employees of layoffs, with internal communications explicitly framing the cuts as necessary to "offset" the company's massive AI investments. This isn't a performance issue — it's a deliberate reallocation of headcount costs into compute and model development.
Why it matters: When the world's largest social platform starts cannibalizing its workforce to fund AI, it's a clear signal that the industry-wide trade-off between people and models has entered a new, more aggressive phase.
Quick Hits
- →Google is rolling out an AI agent ecosystem to consumers, but adoption is the big question mark — the pitch is ambitious, the user enthusiasm is unclear. TechCrunch
- →OpenAI is reportedly targeting September for its IPO, one day after Elon Musk's lawsuit against the company collapsed in court. TechCrunch
- →Waymo has suspended robotaxi service in Atlanta after its vehicles kept driving into flooded roads — a reminder that edge cases in the physical world still humble even the best autonomous systems. TechCrunch
- →The Path, backed by Tony Robbins and Calm alumni, launched an AI therapy product that scored 95 on the Vera-MH mental health safety benchmark — compared to a 65 ceiling for mainstream consumer chatbots. TechCrunch
- →Clouted raised $7M to use AI to predict which video clips will go viral on short-form platforms — Slow Ventures led the round. TechCrunch
AI TLDR