What happened
DeepSeek and other open-source models have surged in token volume across major AI platforms — DeepSeek now processes just over a third of tokens on Vercel's gateway, and V4 Flash handles 5.3 trillion tokens weekly on OpenRouter. However, Anthropic still accounts for more than half of overall AI spend on Vercel, and Anthropic's Opus 4.8 commands roughly 23× higher token costs than V4 Flash ($1.37 per million tokens compared to 6 cents).
Why it matters
The data suggests frontier AI models and open-source alternatives are complementary rather than directly competing — new use cases keep emerging that demand expensive state-of-the-art models, even as mature deployments shift to cheaper open-source versions. This means frontier labs may be able to maintain profitability through premium pricing on high-value tasks, rather than being undercut into commodity status. For businesses, this suggests a two-tiered market: frontier models for discovery and novel problems, open-source for proven, repetitive work.
What to watch
Nvidia's Nemotron model is noted as a newest arrival poised to leap to the front of the usage pack by virtue of Nvidia's strong connections and the model's extreme adaptability. The stability of this two-tiered economy — and whether frontier labs can sustain their pricing premium — remains to be seen.
Summaries like this, in your inbox every morning.
The article presents a counterintuitive picture of the AI market's evolution. Rather than open-source models cannibalizing frontier labs' revenue, the two appear to occupy distinct roles in a maturing market. Vercel and OpenRouter data show that while open-source alternatives like DeepSeek have exploded in token volume—DeepSeek now leads Vercel's platform—this has not significantly eroded the spending share of frontier providers like Anthropic. The mechanism is pricing: frontier models command per-token costs many times higher than open-source equivalents, insulating them from direct competition even when their volume share declines.
The article frames this as a natural market segmentation: frontier labs are winning the "discovery" phase, identifying novel use cases and proving their value, while open-source models capture the "production" phase, handling proven, repetitive tasks at scale. This two-tiered structure appears stable because the addressable market for AI tasks is expanding fast enough that new high-value problems continuously emerge, sustaining demand for premium models. Notably, the article does not claim this dynamic will last indefinitely—it observes that frontier labs "aren't suffering too much from the rise of open source … at least not yet"—but the market structure it describes may represent a durable equilibrium where both tiers coexist profitably.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Vinci4D provides chip-design simulation software that trains on physics equations instead of customer data, an…
SEMI President and CEO Ajit Manocha told DIGITIMES that the semiconductor industry reached the US$1 trillion r…
CoreWeave fell more than 25% this week, its debt yield briefly rose above 16% on Thursday, and SK Hynix droppe…
Nvidia's AI Enterprise lists at $4,500 per GPU annually, and paid subscriptions add supported software, securi…
China's commerce and science ministries released a draft update adding 23 key AI technologies to its export co…
Anthropic told investors this week that it is making 85% gross margins, a figure the columnist calls a cash co…
