Anthropic Is Lying About Its First Profitable Quarter
Two days ago, the Wall Street Journal dropped what looked like the biggest story in AI since ChatGPT launched: Anthropic is on track for its first-ever operating profit in Q2 2026.
But peel back the curtain, and the picture is far less flattering. This “historic milestone” rests on a two-month accounting gimmick, a glaring contradiction with sworn testimony, and suspiciously perfect timing for a massive funding round.
Anthropic isn’t lying in the legal sense — it’s a private company that doesn’t have to follow public-company GAAP rules — but it is aggressively shaping a story that collapses the moment real costs hit.
The SpaceX Deal That Changes Everything
In May 2026, Anthropic signed what may be one of the largest compute contracts in history: $1.25 billion per month to SpaceX (via xAI’s Colossus clusters) through May 2029. That’s $15 billion a year, confirmed in SpaceX’s own S-1 IPO filing. The deal includes a deliberate “ramp-up” discount for the first two months — May and June — before the full freight kicks in July.
The only quarter in which Anthropic’s single largest expense is artificially suppressed just happens to be the quarter it chose to leak as “profitable.” Coincidence? Or convenient optics?
Even the Wall Street Journal, which broke the story, quietly noted that Anthropic “may not remain profitable for the full year” because of planned expense growth. Translation: the profit evaporates the moment the discount ends.
The Revenue Math That Doesn’t Add Up
Here’s where it gets worse.
On March 9, 2026, Anthropic CFO Krishna Rao testified under oath in federal court that the company had generated “over $5 billion” in revenue in its entire history since 2021. That figure was GAAP revenue—real, recognized, auditable money.
Either the CFO understated revenue to a federal judge, or the numbers being shopped to investors right now are inflated. Both cannot be true.
Anthropic’s defenders will wave this away as “run-rate vs. recognized revenue.” But Rao’s court filing explicitly referred to actual revenue generated “to date.” The discrepancy is not semantic — it’s enormous.
The Funding Round Calendar Doesn’t Lie
Anthropic “found” its first profit in the exact window when its biggest bill was on sale, then rushed the numbers to the Journal during peak fundraising season.
The company itself doesn’t dispute that these figures come from internal materials shared with investors. It’s not fraud. It’s just extremely convenient marketing.
Private Company, Private Math
Unlike public companies, Anthropic gets to define its own “operating profit.” The leaked metric conveniently includes model training costs but excludes stock-based compensation — one of the biggest real expenses at any high-growth tech firm. Compute costs alone were eating 71 cents of every revenue dollar in Q1; they dropped to 56 cents in the projection. Impressive efficiency gains, sure. But when your rent jumps by $1.25 billion a month in July—and that’s before adding Amazon, Google, and Microsoft cloud bills — the margin math turns ugly fast.
Conservative estimates put Anthropic’s total monthly compute burn north of $3.5 billion once the SpaceX deal is at full price. A $559 million quarterly profit disappears in weeks.
- JPMorgan Eyes Prediction Markets: Jamie Dimon Says the World’s Largest Bank Is Studying a Launch
- Anthropic's Project Glasswing Update: Mythos AI Uncovers Over 10,000 Critical Vulnerabilities in One Month, Highlighting Urgent Cybersecurity Risks
- SEO Is Dying – But Google Is the Murderer
- Jamie Dimon Warns: The Perfect Storm for a Financial Crisis Is Here
Q3 Is When the Real Story Begins
No more ramp-up pricing. Full Colossus capacity, full $1.25 billion monthly invoice, plus everything else.
That’s when we’ll see whether Anthropic’s enterprise adoption of Claude is genuinely explosive enough to cover the burn—or whether the “first profitable quarter” was a one-time accounting mirage designed to juice the next valuation round.
The AI industry desperately wants to believe the unit economics are working. Investors, reporters, and executives all have skin in the game. But facts are stubborn. A two-month discounted window, sworn testimony that no longer reconciles, and a perfectly timed leak do not equal sustainable profitability.
Anthropic has built extraordinary models. Claude is excellent. The engineering is world-class. But turning that into a business that actually prints money — without relying on temporary discounts and creative definitions — remains unproven.
Q3 won’t have the luxury of a ramp-up grace period. Neither should the narrative.