AI Procurement’s Pricing Problem: Why “Market Rate” Means Nothing When the Market Is Inventing Itself

Procurement leaders argue AI vendors are pricing on confidence, not value. Ten companies, ten quotes for the same product, and no benchmark anyone trusts.

Put the same AI product in front of ten companies and you can walk away with ten completely different commercials. Each one will be delivered with a straight face, defended as the “market rate,” and structured to make the buyer feel late to the party. The problem is not that any one quote is wrong. The problem is that none of them are anchored. The market is inventing its pricing logic in public, and “market rate” has become the most convenient phrase in the seller’s vocabulary.

The debate gained traction after a LinkedIn post from a Senior Procurement Transformation Advisor flagged what she called a structural gap between AI hype and AI commercial reality. Her argument: the technology is moving fast, the pricing is still catching up, and treating both as equally mature is expensive. The post drew CPOs, third-party risk specialists, energy procurement leaders, AI vendors, and category strategists across banking, infrastructure, and tech. The agreement was strong. The disagreements were sharper, especially on what procurement should actually do about it.

The Confidence Premium

The most cited reframing came from Farah Boukillou, a procurement leader and former OCP Group executive. “AI is being sold on potential, but priced on confidence.” Her diagnosis explained the variability without needing a conspiracy. Procurement teams need to anchor conversations back to value realization, not vendor narratives.

Nuha Luqman, who works in procurement for energy ecosystems, captured the dynamic in one line. “’Market rate’ in AI right now often just means ‘what someone managed to get away with.’”

Nicolas Frendo, a financial executive at Acephalt, used a different metaphor to make the same point. “Pricing AI always feels like a weird game of telephone, every company hears a different number and nobody knows which came first. Seeing ‘market rate’ show up in a room honestly just means, yeah, no idea what the market actually is, someone had to improvise.”

The structural reality behind those observations is straightforward. When vendors lack benchmarks themselves, they price based on what each buyer signals they can absorb. Procurement teams without independent reference points end up funding the seller’s price discovery.

Commercial Maturity Lags Technical Maturity

The most consistent thread was that procurement teams keep confusing the two.

Azfar Wasim, a business and digital transformation consultant, put it directly. “Too many teams assume technical sophistication equals commercial maturity. It’s a very expensive assumption to make.”

Jahabar Ifthikar, a senior procurement and digital transformation leader on $1.8B+ infrastructure projects, framed the trap precisely. “AI may be advanced, but its commercial models are still immature and inconsistent, which creates a dangerous gap between perceived value and actual cost. What’s often sold as ‘market rate’ is, in reality, a moving target shaped more by narrative than fundamentals.” His verdict on the procurement role: “Not just negotiating price, but challenging assumptions behind usage, scalability, and value realization. If pricing isn’t transparently linked to outcomes or if it shifts risk disproportionately to the buyer, it’s not innovation, it’s poor commercial design.”

Abdulmajeed Al Sheraim, a procurement and contracting leader, identified the shift required. “The key shift here is moving from price validation to value validation. How does usage scale? Where does cost outpace value? What happens under different adoption scenarios? If those questions aren’t clear, then the risk isn’t just overpaying, it’s locking into a model that doesn’t hold as usage grows.”

Fatima Sayalı, a procurement officer in manufacturing, drew the discipline back to basics. “We apply TCO thinking to every category we manage. AI contracts deserve the same discipline. Lock-in risk, adoption assumptions, renewal terms, these are not ‘AI problems.’ They are procurement problems. The hype does not change the fundamentals.”

The Token Trap

Several commenters dug into the mechanics of where the cost risk actually sits.

Brian Mangano, founder of Balestra Group, drew a useful distinction. The pricing chaos fairly accurately describes LLMs and services that are just wrapping an LLM, but does not hold for AI use cases that are not purely a token exchange. The counterpoint that emerged was that even when buyers do not directly pay for tokens, they pay indirectly. Vendors absorb the token cost in their price, which means token volatility still shows up downstream, just less visibly. Tokens are not expensive by themselves, but they increase unpredictability whenever the use case consumes them aggressively.

Scott Cohen, a former Cloudflare, Spotify, and MongoDB procurement technologist, raised the reverse risk. “This does present an opportunity for purchasers to do token arbitrage. Essentially buy a flat fee AI feature in their platform and then use it so much they exceed the fee in token utilization.”

That arbitrage is happening in practice. The reverse risk for vendors is real. No one wants to launch a software company and watch it go bankrupt because of overusage of tokens. The instability runs both ways.

Pricing Models Are Multiplying

Clarice Camacho, a global energy procurement and risk management leader, mapped out the most common approaches now in market. Usage-based pricing offers scalability but kills predictability. Subscription is predictable but limits flexibility. Per-seat is ideal for teams but rarely tied to actual value. Tiered scales with customer needs and complexity. Outcome-based is powerful but hard to measure.

Her core principle: “The real question isn’t ‘which model is best?’ It’s: which model aligns with the value your users actually get?” The strongest AI companies, she noted, are blending models. Subscription plus usage caps. Freemium entry plus premium tiers. Custom enterprise pricing for scale. “If your pricing doesn’t match how value is delivered, you’ll either leave money on the table or lose customers.”

A Public-Utility Model for AI

The most concrete buyer-side proposal came from Matt Vander Weg, a supply chain operations executive. “AI pricing in Procurement needs to be treated like a public service. Think of it as your electric bill, water bill etc. You have an allotment of use that is fixed (estimated consumption) plus overage costs.”

His negotiation playbook was specific. Negotiate year one as a variable cost to set the company’s average mark. Year two, set the fixed rate at the previous six months’ consumption rate. From there, run the fixed rate on a rolling six or twelve month prior baseline with allowable adjustments up and down. He went further, suggesting that cable, electric, and energy companies should be the ones negotiating directly with OpenAI, Anthropic, xAI, and others to remove the burden of individual commercial entities negotiating ambiguous pricing models.

Build Versus Buy Gets More Attractive

A second strategic thread argued that pricing instability makes building in-house more attractive.

Gael De Martelaere, a Chief Procurement Officer, put it bluntly. “Building becomes more and more attractive versus buying. Buy basic capabilities and infrastructure and build your AI on top of it yourself. Most AI buys end up being single-use case apps that with the right inhouse capabilities can be fairly easily built.”

The counterpoint that emerged was practical. Building in-house only works when the resources exist. At minimum, that means one procurement professional who understands the processes and data freed from other duties for weeks to months depending on complexity, plus the right technical people. Without that, the build path becomes more expensive than the buy path.

Adoption, Not Just Price

The deepest critique came from Christian Engler, a strategic procurement leader in public sector and IT. “The pricing model is only half the story. Many companies negotiate AI commercials as if cost control alone solves the problem. In practice, weak adoption, unclear ownership and poor use-case discipline destroy ROI faster than pricing ever will. Commercial maturity and organisational maturity need to move together.”

Bob Thompson, a procurement transformation specialist, raised the prior question. “Does the tech I buy today really work and has it been properly implemented? Yes or no? Then ask yourself, will adding AI actually enhance or increase my ROI, or will it simply cost me more money?”

John Hatton, a CPO and FCIPS, kept the framing grounded. “Nobody’s talking about the costs, nor the short to medium term stability and viability of the AI natives.” Good practice advice on AI procurement, he noted, frankly applies to just about any procurement activity.

Ghanshyam Rao, a procurement and supply chain strategist, distilled the test. “If procurement cannot clearly measure value, usage risk, and lock-in exposure, the pricing model is not mature enough yet.”

Takeaways for Procurement Leaders

Three lessons run through the discussion. First, “market rate” is not a benchmark when the market is still inventing itself. Independent reference points, usage modeling, and TCO discipline matter more in AI than in any settled category.

Second, the pricing model has to match how value is actually delivered. Subscription, usage-based, per-seat, tiered, and outcome-based each carry different risk profiles. Blended models with usage caps are emerging as the most defensible structure.

Third, commercial maturity and organizational maturity have to move together. Weak adoption destroys ROI faster than any pricing model. The procurement team that locks in a great price on a tool nobody uses has not won anything.

How are you anchoring AI procurement decisions when there is no real benchmark to anchor to?

Continue the discussion on Chain.NET, the global supply chain community, at www.chain.net. Ask questions, join events, and access exclusive resources.