AI Just Crossed Below $1 Per Million Tokens.

Published: September 2, 2026 | Category: AI News

The LLM Token Expenditure Index, Silicon Data’s flagship measure of daily AI token pricing, tracked on Bloomberg under the ticker SDLLMTK, that just fell to 97 cents per million tokens on Monday. AI Token prices are at record lows. That’s the index’s lowest reading since it was created late last year, and it’s now more than halved from its peak earlier this summer.

Sub-dollar AI. The basic unit of intelligence – a million tokens , now costs less than a cup of coffee.

Seventy-three percent of enterprises reported in 2026 that their actual AI costs exceeded original projections.

The Price Collapse Is Real And It’s Accelerating

Before the paradox, the data. Because the scale of the price drop is worth sitting with for a moment.

The price of processing AI tokens has fallen to roughly $1.16 to $1.18 per million tokens as of early August 2026. That’s the lowest average inference cost recorded this year, representing a 43% decline from $2.04 at the end of May. To put that trajectory in perspective: frontier AI intelligence is now priced at approximately 12% of its March 2023 levels. Some benchmarks show comparable AI capabilities costing over 280 times less than they did in early 2023.

And by Monday September 1st, it had crossed below $1 for the first time.

In late 2022, GPT-4 class inference cost $37.50 per million tokens. By August 2025, equivalent capability had fallen to $0.14 per million tokens. That’s a 280-fold reduction in two years. According to Epoch AI, inference costs are falling 5–10x per year, driven by both hardware improvements and algorithmic efficiency gains including quantization, speculative decoding, and distillation.

Three forces are driving this collapse right now, simultaneously:

Force 1: OpenAI’s Price War – OpenAI slashed pricing on its GPT-5.6 Luna models by 80% in late July 2026. After the cuts, input tokens cost $0.20 per million and output tokens run $1.20 per million.

Force 2: Chinese Open-Source Models – The rise of lower-cost open-source Chinese models, such as Moonshot’s Kimi K3 has pulled down market rates significantly. Deep Seek achieved price reductions of up to 99% earlier in 2026, making the American price war look almost restrained by comparison.

Force 3: Dynamic Pricing Structures – Other frontier labs have introduced dynamic pricing structures that allow access rates to move with demand, adding further downward pressure to the market rate for tokens.

None of these forces is slowing down. If anything, they’re feeding each other. Every price cut from a major provider forces the others to respond. Every capable open-source model that ships shifts the floor lower.

Who Gets Hurt: The Pressure on Frontier Labs

Before the implications for buyers, understand what this does to the companies on the other side of the transaction.

The drop puts particular pressure on frontier model companies. “Foundation model labs are the most directly exposed,” Charles-Henry Monchau, investing chief at Syz Group, wrote in a note cited by CNBC.

A sharp slide in prices can mean AI model users will need to shell out less, but it reduces pricing power for providers. A lower index price can condition consumers to expect lower rates for access to AI offerings, resulting in less pricing power for providers over time.

This matters for marketers because the companies building the AI tools you depend on are operating in an environment where their revenue per unit is collapsing while their infrastructure costs remain enormous. That creates pressure, on their business models, on their pricing strategies, and ultimately on which products they invest in and which they deprioritize.

The companies that survive this price war will be the ones with volume, massive adoption that makes up in scale what they’re losing in margin. Which means the race to grow user bases is intensifying exactly as prices fall.

The Paradox You Doesn’t Understand!

Here’s where it gets counterintuitive and where most internal AI budget conversations go wrong.

Despite the price of a single token dropping more than 90% since 2023, spending on large language models has doubled since late last year, according to the Silicon Data Token Expenditure Index. Apollo chief economist Torsten Slok described it as “Jevons paradox in action.”

The Jevons Paradox, named after 19th century economist William Stanley Jevons, describes a pattern that has shown up in every major technology transition in history. When coal-powered engines became more efficient, coal consumption went up, not down, because cheaper energy unlocked uses that weren’t viable before. The same happened with electricity. With bandwidth. With storage.

“As tokens get cheaper, companies don’t spend less but instead run more AI agents, automate more workflows and generate more code, pushing aggregate expenditure higher even as the unit cost of intelligence collapses,” Slok wrote.

The numbers make this concrete:

Per-token LLM prices dropped 1,000x in three years. During the same period, enterprise AI spending grew 320% — from $11.5 billion to $37 billion. The organizations spending the most on AI are overwhelmingly the ones that benefited most from falling prices.

Google reported processing 3.2 quadrillion tokens per month by mid-2026, roughly 7x its year-earlier rate. This is the Jevons Paradox in real time: the more efficient the resource, the more of it gets consumed.

3.2 quadrillion. Every month. From one company.

What’s Coming: The Trajectory Doesn’t Change

Predictions indicate continued declines in per-token costs as hardware improves, yet overall spending will rise with wider adoption across sectors such as healthcare and finance.

Goldman Sachs projects token consumption growing approximately 24x by 2030. Reasoning and agentic models are token-hungry, they consume far more tokens per task than chatbots, which is why “the model got cheaper” rarely lowers the bill.

The trajectory is clear and it isn’t changing: per-token prices will keep falling. Total token consumption will keep rising. Total AI spend will keep going up for organisations that are actually using AI to do more.

The strategic posture that wins in this environment is not “cut AI spend because tokens are cheap.” It’s “deploy aggressively into use cases that deliver real business value, measure tokens against outcomes, diversify your vendor exposure, and build consumption visibility into your financial reporting before the bill surprises you.”

The Bottom Line

A widely watched benchmark for AI token pricing just crossed below $1 for the first time, a threshold that puts meaningful pressure on the business models of top AI companies while representing a genuine opportunity for every business that uses AI as a production tool.

The businesses that understand the Jevons Paradox will use this moment to unlock use cases they’ve been holding back on cost grounds. The ones that don’t will keep wondering why their AI bill is going up in a market where tokens have never been cheaper.

Sources:

Share you part of the story on how you have successfully fixed the Value Inflation or If You liked our content, pls Comment Below!