Deutsche Bank: AI Premium Pricing Faces Disruption from Open-Source Models
Miles Bennett
Deutsche Bank warns the performance gap between open-source and frontier proprietary AI models is closing fast, posing a systemic challenge to premium pricing — and direct pressure on AI companies whose valuations still depend on it.
Same task, 65× the price — what's the difference?
Anthropic's flagship Claude Fable 5 scores 60 on the Artificial Analysis intelligence index at roughly $3.25 per task; DeepSeek V4 Pro scores 44 at about $0.05.
Frontier models still lead on complex reasoning and agentic applications — software that lets AI carry out multi-step tasks autonomously. But Deutsche Bank argues low-cost models already handle the vast majority of everyday work.
In plain terms = it's a supercar versus a reliable sedan. Most commuters never need supercar performance — and paying supercar prices for a sedan's job wears thin.
Where is the real dividing line?
Deutsche Bank notes the low-cost AI race has moved beyond the "U.S. vs. China" narrative. Meta's Muse Spark, Nvidia's Nemotron 3 Ultra, and OpenAI's open-weight model gpt-oss all sit in the low-cost camp.
The report states directly: "The real dividing line is not the U.S. versus China — it is frontier proprietary models versus open-weight models."
This means → investors still framing valuations around a U.S.–China AI race may be looking at the wrong fault line. The pressure comes from the open-source camp, regardless of nationality.
How is the pricing model shifting?
Major AI developers are moving from subsidized flat-rate plans to usage-based billing, driven by financial scrutiny ahead of IPOs — the loss-making subsidy phase is nearing its end.
Uber reportedly burned through its AI-token budget within months, then began capping employee AI spending. Deutsche Bank treats this as an early signal of enterprise cost discipline.
This means → once pay-per-use rolls out broadly, enterprise customers will be forced to do the math: is it worth paying 65× more for a frontier model on routine tasks?
What does this mean for AI valuations?
Deutsche Bank does not predict frontier-model premiums will vanish — it expects the premium to migrate upward with each new generation, always present at the cutting edge.
But the report's core judgment: as operating costs draw more scrutiny, the market's reassessment of proprietary-AI valuations may prove more lasting than the brief correction after the "DeepSeek moment."
This reflects a deeper signal: the early-2025 DeepSeek shock was digested quickly, but this time the cost pressure is structural — not a single event, but a steadily narrowing gap.
Content is for reference only, not financial advice.