The rapid evolution of artificial intelligence has transformed how businesses operate and interact with technology. However, as companies like Microsoft, Google, and Anthropic pour billions into the development of Large Language Models (LLMs), the question of how to effectively monetize these innovations has become increasingly complex. With many users enjoying free access to AI tools like ChatGPT, the struggle to establish a sustainable pricing model for these services is raising eyebrows across the industry.
The Cost of Innovation
For those who have dabbled in the world of AI, using platforms like ChatGPT or Anthropic’s Claude feels like a steal. These tech giants have invested staggering sums—reportedly hundreds of billions—into creating sophisticated LLMs. While the free versions of these tools are a boon for users seeking assistance with tasks from coding to holiday planning, companies have a vested interest in recouping their investments. Thus, they offer premium versions packed with additional features tailored for more specialised tasks.
Yet, as Simon Gooch from Saviynt notes, “Trying to tie someone into a cost model for the next 12 months, two years, three years, it doesn’t make any sense, honestly, because we don’t know.” This uncertainty stems from the volatile economics surrounding tokens, the fundamental units that power LLMs and agentic AI.
The Token Dilemma
When users interact with an LLM, their prompts are broken down into tokens—essentially chunks of data that the model processes to generate a response. However, the unpredictability of this process complicates how businesses can estimate costs. Variations in prompts can lead to differing outputs, and even the same input may yield inconsistent results across different models. Adding to this complexity is the rise of agentic systems, where multiple AI agents work in tandem, further escalating both token usage and unpredictability.
Despite the cost of individual tokens having decreased significantly in recent years, a report from Goldman Sachs indicates that the overall consumption of tokens is skyrocketing. The bank predicts that by 2030, external token usage will soar to an astonishing 120 quadrillion tokens per month as businesses increasingly adopt AI. However, many organisations, including giants like Microsoft, are grappling with the reality of managing these escalating costs, often discovering their token consumption only when they receive their monthly bills.
The Challenges of Cost Management
Will Venters, an associate professor at the London School of Economics, highlights the difficulties companies face as they experiment with AI. “People are finding it really hard to manage that cost… it’s a non-deterministic output, so it’s a non-deterministic value,” he explains. As organisations integrate AI into their operations, employees can inadvertently consume tokens at an alarming rate.
Smaller companies have found ways to navigate these challenges, with some cleverly using flat-fee personal accounts to keep costs down. Oliver King-Smith, founder of engineering software firm smartR AI, warns that this workaround may not last. As the big players come under pressure from shareholders to turn a profit, tighter restrictions on these accounts are imminent.
Companies also need to sharpen their focus when choosing AI models. Rob Steele, CFO at the UK accounting software firm iplicit, emphasises the importance of precision in prompts: “You wouldn’t send someone in your family out to get the weekly shop without detailed instructions.” As AI becomes more embedded in products that could serve thousands, the potential for costs to spiral out of control increases dramatically.
Future Pricing Models
As businesses look to the future, the unpredictability of token costs presents significant challenges. Venters notes that while the value derived from AI may justify token use, companies still need to find ways to pass these costs on to their customers. Bill Peterson, senior director of product marketing at Sumo Logic, admits that finding an effective pricing strategy is still very much a work in progress. Options under consideration include raising prices across the board, charging based on results, or offering bundles of services.
However, any pricing model will likely be subject to rapid changes as LLM providers adjust their own pricing structures. “You get into variable pricing, and it’s changing every couple of months,” Peterson explains. “Customers don’t like that. That’s not how anybody builds a budget.”
Why it Matters
In a world increasingly driven by artificial intelligence, understanding the complexities of pricing and tokenomics is crucial for businesses and consumers alike. As AI technologies continue to permeate various sectors, the ability to manage costs effectively will not only influence corporate profitability but also shape the future landscape of AI services. The challenge lies in striking a balance between innovation and sustainability—ensuring that the benefits of these remarkable technologies can be enjoyed without breaking the bank.