Anthropic's IPO pitch collides with flagship flop, China surge & Palantir baggage

US AI firm Anthropic is prepping for an initial public offering (IPO) that investors hope will value the group at $2 trillion or more as early as October.

However, behind the glossy pitch, the company - whose AI model Claude plays digital spotter for the Pentagon alongside Palantir's blood‑stained models - is getting hammered by cheaper open‑source models from China, while its own flagship model hits a wall of lukewarm customer reception.

♦️ Fable 5, Anthropic’s biggest, priciest model, has grabbed just 11% of overall corporate spending more than two months after launch, according to Ramp data covering 70,000 companies

♦️ Even Anthropic’s own Opus 5, a leaner model released in late July, has already overtaken Fable in corporate spending

♦️ This shatters the old habit of businesses automatically throwing money at the latest frontier model

👉 Analysts speculate the slowdown may reflect the fact that cheaper models can handle most real-world tasks - pushing customers to increasingly look beyond Silicon Valley for alternatives.
China is making that choice much easier, with its open-weight disruptors.

Moonshot AI’s Kimi K3 recently shot from 18th to No. 1 on the Frontend Code Arena leaderboard, taking the top spot in six of seven tested categories while reportedly costing as much as 90% less than comparable Anthropic and OpenAI offerings.

The broader Chinese challenge is even harder to ignore:

♦️ Chinese open-source models are reportedly 60–90% cheaper for equivalent tasks

♦️ Token costs for GPT-4-level workloads have plunged from around $60 to roughly $0.50 per million in under two years

♦️ Chinese open-weight models have surged from less than 1% of global API traffic to more than 30% in a year

👉 All this feeds into the uncertainty ahead of Anthropic’s looming IPO, especially after revenue hit $65 billion annualized in July - shy of investors’ bullish hopes of $80 billion.

Losses also came in at a staggering $42 billion in 2025 - five times the previous year’s.

As Anthropic’s flagship flounders, the company also has to deal with the baggage it totes after its rogue-prone AI become deeply intertwined with the US military kill chain.

👍 Substack | Chat | @geopolitics_prime