Even as enterprises adopt open-source AI more widely, Anthropic's Claude still commands a majority of spending in high-value use cases and has not seen its revenue significantly eroded—according to an analysis based on real-world usage data published on July 7, 2026.
July 7, 2026 · AI Economics
Open-Source AI Is Winning Volume — But Anthropic Still Owns the Money
Cheaper open-weights models like DeepSeek V4 now process a third of all tokens across AI gateways. Yet spending stays concentrated on frontier models — leaving Anthropic's revenue largely unscathed, for now .
50%+
of total gateway spend captured by Anthropic
⅓+
of all tokens now processed by DeepSeek
70–80%
Anthropic spend share in coding & agent workloads
Per-Token Price: Opus 4.8 vs DeepSeek V4Flash
Priced per 1M tokens — the frontier model costs roughly 23× more, so a smaller volume still commands the bulk of dollars.
≈ 23× the price per token → high volume ≠ high spend
Weekly Usage on OpenRouter (tokens)
Open-source leads on raw volume — while frontier leads on revenue.
The "AI Lifecycle" Thesis
Why frontier and open-source models are complementary, not competing — so premium revenue persists.
Discovery
New, experimental use cases lean on frontier models
→
Production
Mature workloads migrate to cheaper open-source
→
New Demand
Fresh use cases keep frontier demand alive
Decagon shifted ~90% of its workloads to open-source — yet the industry-wide pattern of concentrated frontier spend still holds.
The Caveat: "Yet"
As open-source models grow more capable and costs keep falling, the open question is how long this lifecycle equilibrium — and Anthropic's spend dominance — can hold.
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…