The Information reported in late June 2026 that China's AI lab DeepSeek accelerated its first external fundraising—totaling around $7.4 billion—after witnessing the capabilities of Anthropic's new frontier model, "Claude Mythos."
June 2026 · DeepSeek
DeepSeek raises $7.4B, becoming China's most valuable AI startup
A 50-billion-yuan first external round at a $50B+ valuation — reportedly triggered after founder Liang Wenfeng saw a preview of Anthropic's frontier model Claude Mythos and concluded that a massive war chest was needed to keep competing.
$7.4B+
raised in first external round (50B+ yuan)
$50B+
valuation — China's most valuable AI startup
~$3B
put in personally by founder Liang Wenfeng
Valuation, to scale
Even at a record raise, DeepSeek is dwarfed by Anthropic
Roughly 17× larger — each block ≈ $90B
The catalyst · Claude Mythos
A step-change in software engineering and cyber capability, per Anthropic's 244-page system card — and the spark behind DeepSeek's raise.
Firefox JS shell exploit success — a ~5.5× jump over the prior model
DeepSeek
Open-weight, low-cost, efficiency-first
Touts far lower API costs
Constrained by U.S. advanced-chip export controls
Strong coding, math & multilingual (esp. Chinese)
Anthropic — Mythos line
Safety-led, with deliberate access limits
$25 / $125 per million input / output tokens (preview)
Drew temporary U.S. restrictions over offensive risk
Step-change in autonomous cyber offense/defense
The contest is no longer just about model cleverness — it's about reserves and infrastructure.
An unusual limited-partnership structure leaves most external investors — reportedly Tencent and CATL among them — with little or no voting power, letting Liang keep control. Some observers note the irony that export controls may be accelerating Chinese capital raises.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…