While attending Telecoms World Asia in Bangkok last week, I had several interesting conversations with colleagues regarding my “Value Threading” methodology. While we were on the mark regarding “ecosystem monetization” and building customer value, the discussion inevitably turned to the valuations of AI Platforms.
A recent Wall Street Journal article (“AI Investors Want More Making It and Less Faking It”) questions the market value of the many companies bringing infrastructure, Large Language Models (LLMs), and SaaS applications to market. The question everyone is asking is: Where is the actual value?
With Oracle reaching new market cap heights, the excitement is noteworthy. However, the reality is that the current circular investment model is disconnected from the actual business value achieved, even if market caps currently reflect that optimism.
The Pricing Paradox
Let’s leave the market-cap speculation to the analysts and focus on the value being delivered by seemingly every global enterprise investing in AI. What do their value models look like? On a scratch sheet, it is clear that the potential business value is massive. In my view, this is why most have invested in AI platforms with very little concrete value modeling.
What I find most amazing is how inexpensive LLMs are right now. I completed a preliminary study last week and was stunned by the low barrier to entry. Why are LLM models in a clear “race to the bottom” on pricing? Are the capital markets literally investing in price reduction?
My recent update to Gemini is a perfect example. I pay ~$240/year for a subscription that now provides access to the new Gemini model. Tack on an additional $99/year for cloud storage and Google tools, and I am spending ~$350/year for the best productivity suite I have ever used.
To put this in perspective: If I owned a company of 1,000 employees (inferring a $250M to $500M tech company), I could spend $350,000 to equip every single employee with this full productivity stack. If I were to standardize on a global CRM or ERP platform, the cost would be 10X that amount.
So who is funding this discount? You guessed it: the capital markets. They are literally buying customers to capture market share.
The Shift to Consumption Pricing
Digging deeper, I looked at the cost of a “token” as it relates to—or replaces—per-user pricing. Here again, prices continue to decline at a staggering rate. In the Epoch AI article “LLM inference prices have fallen rapidly but unequally across tasks,” the cost per million tokens is dropping amazingly fast.
I wanted to do some math. What would it look like if everyone in the United States had a license to Google’s Gemini Pro, as I do?
Here is the math:
- US Population (approx. 2025): ~347 million people
- Annual License Cost: $240 ($19.99/month × 12)
- Calculation: 347,000,000 people × $240/year = $83,280,000,000
Total: ~$83.3 Billion
For context on how small that number is:
- US Defense Budget: ~$850 Billion
- US Healthcare Spending: ~$4.5 Trillion
- Total US GDP: ~$29 Trillion
So, where is the future revenue coming from?
The Agentic Explosion and the Unfunded Liability
The “Agentic World” is projected to drive nearly all the growth in this sector. Projections suggest that upwards of 99% of all future AI transactions will be agentic—non-human-based transactions.
This is the volume everyone is counting on. Having your AI platforms ingest massive amounts of data, continually scan systems, communicate with customers, or train robots is where nearly all future token volume will be generated.
In effect, any global enterprise adopting these tools has already embarked on a massive, unfunded liability.
This is what the capital markets are funding today: the infrastructure for an infinite future of consumption. Planning for this now is a massive opportunity to control this liability before it spirals out of control.
Three Steps to Take Now
To get ahead of this curve, there are three key “Value Threading” initiatives you should start immediately:
- Deploy a Purpose-Built “Traffic Cop” – Create an AI API Management Plane that is agnostic to ANY cloud platform or LLM. This acts as a management layer across your cloud or LLM solutions, giving you the routing flexibility to manage costs, throughput, and security. You need a central point of control for all incoming traffic.
- Implement an Agnostic Semantic Cache – As part of these incoming calls, implement a Semantic Cache. This creates a “memory” for your system; if a question has been asked before, the system retrieves the answer from the cache rather than making a new API call. This improves performance and drastically reduces token consumption—a direct cost savings today that scales into the future.
- Institute Consumption-Based Chargebacks – Finally, to track and prioritize costs, implement usage-based “billing” at the API gateway. This creates internal billing for every department or cost center using the AI platform. This transforms your IT solution into a charged service and, most importantly, allows you to track the costing of this growing liability.
Is now the time to put these foundational business process improvements into motion? Absolutely. It is the only way to get ahead of the massive unfunded liability in your future.
Deploying this now, before you have the scale, will be a foundation for years to come.
Over the next few months, I will be sharing more about the cost-savings opportunity here, as well as the projected benefit of controlling that unfunded IT future liability.
