Silicon Data's LLM Token Expenditure Index (SDLLMTK) has fallen roughly 20% from its May peak, drawing attention as a gauge in which lower AI usage prices may signal trouble ahead for AI-related stocks. On July 3, 2026, Bloomberg reported that the AI trade is losing one of its key signals.
July 2026 · LLM Token Price Index
The AI Trade Loses a Key Signal as Token Prices Slide
The SDLLMTK index — tracking prices buyers actually pay — has fallen about 20% from its May 2026 peak, a warning sign of weakening pricing power across the AI sector.
−20%
Fall from the May 2026 index peak
90%+
Of global LLM inference spend covered
400+
Models tracked (20+ in the basket)
$700B+
AI capex now straining inference budgets
Reference price per 1,000,000 tokens
As of April 2026 · frontier vs open-weight segment
Frontier tokens cost roughly 5× more than open-weight — a gap fueling the shift to cheaper, efficient models.
Why prices are falling →
HARDWARE
New GPUs (Nvidia Blackwell) boost inference efficiency
→
COMPETITION
Intensifying price war + a shift to cheaper models
→
RESULT
Lower blended per-token prices, weaker pricing power
The free-for-all phase is ending
Companies are capping employee AI usage as costs bite
Uber
Burned through its AI budget in four months; set a $1,500 monthly cap per coding tool
Walmart
Ended unlimited access to its "Code Puppy" assistant and introduced token caps
Amazon · Meta · Cisco
Imposed usage limits or steered staff toward cheaper models
▲ Upside for users
Cheaper access to high-capability models. Demand elasticity means total spend could still rise even as unit prices fall — the sector isn't necessarily shrinking.
▼ Warning for investors
Falling effective prices signal weakening pricing power. Corporate caps and the flat-rate-to-usage-based shift mark a turning point for AI economics.
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…