Tokens are the unit of consumption behind every AI request, and they are the real cost driver that flat subscription pricing currently hides. As enterprise usage grows, pricing is likely to move from all-you-can-eat subscriptions towards pay-per-use, and possibly towards tiered token types for different classes of work. That makes AI a variable cost line, which requires token budgeting frameworks — departmental limits, usage tracking and optimization KPIs — rather than an annual license assumption.
How does your organization assess AI investments?
I’d wager it’s through a lens of ‘all-you-can-eat’ subscriptions for language models, enterprise software and automation tools.
But there’s a seismic economic shift happening just beneath the surface that’s going to fundamentally reshape how you think about, and pay for, AI investments.
And that shift starts with ‘tokens’. Your real currency in the AI era.
What are tokens?
Think of tokens as the fundamental building blocks of AI processing. Every time you ask AI to write an email, analyze data, or answer a question, your request gets broken into these discrete data packets for processing.
AI applications are extremely token-hungry, and their appetite is growing exponentially every day.
To give an idea of the scale at play: OpenAI alone processes over 100 billion tokens daily, with global consumption potentially reaching trillions. For perspective, all indexed web text ever recorded in history equals about 2 quadrillion tokens. We’re processing massive volumes of recorded human knowledge every single day.
Why does this matter for your business?
The monthly AI subscription fees you’re paying hide the massive infrastructure costs of processing tokens. As AI use cases become more sophisticated and enterprise-level usage dominates, you can bet that token consumption will become the real driver of costs.
Enter AI factories and massive government spending
We’ve long accepted that traditional datacenters weren’t built for AI’s intensive demands. NVIDIA’s answer to this infrastructure gap fits the bill: build mega-scale AI factories. These are purpose-built GPU-fueled facilities designed specifically for AI processing. If you think of datacenters as digital storage warehouses, then AI factories are sophisticated industrial-scale production lines that crunch tokens to manufacture intelligence at scale.
AI factories aren’t just Big Tech initiatives; they’re becoming a national infrastructure priority. Governments worldwide are investing hundreds of billions in AI infrastructure, signaling this is as critical as highways, power grids, and internet backbone.
The competitive landscape is heating up too. While NVIDIA leads infrastructure, companies like AMD, Google TPUs, and other emerging players are driving innovation in token efficiency, helping to drive down the cost-per-token even as token consumption skyrockets.
The imminent price revolution
This massive infrastructure boom comes with a price tag that will inevitably reach the end user. Expect a shift from ‘all-you-can-eat’ subscriptions to ‘pay-per-use’ pricing. Future AI services will likely charge based on how many tokens you consume, just like how cloud computing charges for data storage or processing power.
Will all AI tokens be the same?
Here’s where I think it gets really interesting. Although we’re not there yet, I expect to see different types of tokens for different performance needs, e.g.:
- Basic tokens for simple tasks like drafting emails
- Premium tokens for complex analysis that requires ‘deep thinking’
- Research tokens that can access specialized databases and cross-reference multiple sources
AI factories produce the computational capacity to generate tokens, while AI service providers will charge based on the number and type of tokens you consume.
What does this mean for business planning?
Flexible pricing means AI expenses will fluctuate based on consumption, and enterprises that use AI efficiently will have significant cost advantages. This variable cost structure will require novel approaches to budgeting and financial planning.
But, you can already start implementing token budgeting frameworks today with measures like setting departmental token limits, tracking usage patterns, and establishing optimization KPIs.
Looking ahead: key considerations
As we navigate this transformation, I’d consider:
- How usage-based AI pricing will affect budget planning,
- How to optimize AI usage for maximum value and cost control,
- How AI infrastructure efficiency could serve as a real competitive advantage,
- What risks come with variable pricing models, from budget volatility to vendor lock-ins.
The bottom line
We’re architecting the Industry 4.0 revolution in real-time, and the truth is, we have more questions than answers. I find that both humbling and exciting.
As we explore this new landscape, we’re investing in fundamentally new ways of creating, sharing, and pricing intelligence itself.
Organizations that recognize this shift early, implement token management frameworks, and diversify their AI infrastructure partnerships will have the greatest competitive advantage.
Kamran Habibollah is the founder of Third Horizon Capital Advisory, which provides CFO-grade strategic finance insights to founders, CEOs and boards. He spent 20+ years in global technology enterprises running multi-billion-dollar P&Ls.
Kamran Habibollah
Founder & Principal, Third HorizonTwenty years in technology and telecommunications finance, including senior finance leadership at Cisco, across the Middle East, Africa and Europe. He advises founders, CEOs and boards on capital strategy, transactions and investor relations from Dubai.
More about Kamran →