Accenture reported earnings and the stock fell 18% in a single session. Cognizant, Wipro, Capgemini and IBM fell with it. The largest IT-services companies in the world repriced together, in a day.
Two forces drove it, and they hit the same business model at once. Demand softened. And the market decided AI is starting to do the work these firms used to bill for. The second force is the one that matters, because it is structural, not cyclical, and it reaches a long way down-market, into the SMB-to-Corporate accounts a channel partner serves.
It is worth being clear about what is causing this, because the cause is not anyone acting against the channel. The platforms are driving the cost of intelligence down for customers, which is exactly what their customers and their shareholders want. The commoditization is a feature of the AI era working as intended. The casualty is incidental: the channel margin that was quietly built on intelligence being scarce. Understanding the repricing as a natural consequence, rather than an attack, points to the right response.
Why Scarcity Was the Whole Game
For decades, a channel partner's margin rested on a single fact: intelligence was scarce. Expertise was expensive, slow to scale, and hard to replicate, so configuration, advisory, and the knowledge markup justified the rate card. AI does not eliminate intelligence. It eliminates exclusivity over it. And exclusivity is what set the price. When the scarcity collapses, the margin built on top of it collapses with it.
The Price Signal
Look at where model pricing is going. Open-weight models now deliver coding and reasoning performance approaching the frontier at a fraction of the cost. Leading proprietary models can run ten to twenty times more per token. The lesson customers are drawing is not "use the cheap model for everything." It is more dangerous for a channel: route each task to the cheapest model that can handle it, and reserve the expensive frontier model for the small share of work that genuinely requires it. When the underlying intelligence is a commodity routed by price, the markup that used to sit on top of it has nowhere to hide.
Not every layer of a partner's value moves the same direction. The split is the whole point.
| Layer of partner value | Direction under AI | Why |
|---|---|---|
| Product / resale margin | Compresses | Platforms bundle the same capability at near-zero marginal cost |
| Generic advisory / knowledge markup | Compresses | A foundation model answers the generic question for free |
| Configuration and implementation | Compresses | Falling code-production cost makes the work cheap to reproduce |
| Outcome ownership / accountability | Holds | Someone must still stand behind the result if it fails |
| Domain depth in a vertical | Holds | Specific process logic is costly to encode and rarely worth a platform's time |
Why This Hits SMB-to-Corporate Specifically
It is tempting to read the Accenture story as an enterprise problem (large firms billing large hours for work AI now does). But the commoditization argument is sharpest in the SMB-to-Corporate market. These customers were never paying for frontier-grade sophistication; they were paying for access to expertise they could not hire. The moment that expertise is available at the price of an API call, the markup a partner placed on it has nowhere to hide. If producing the work approaches zero marginal cost, anyone can reproduce it, including the customer.
The Same Fact, Two Signs
The inference-cost story cuts both ways depending on where a customer sits. A partner whose customers run on the lower-cost compute stack feels a tailwind; one whose customers sit on the more expensive side feels a headwind it will be negotiating against at every renewal. The realignment between the Google, AWS and Microsoft ecosystems, traced in our 2026 Channel Forecast, will keep moving those costs around for years. The point for a channel partner is simple: you do not set the price, you absorb the friction of being on the wrong side of it. Building a practice around which model or which cloud wins is a bet on something you do not control.
The Way Out
The escape is not a better model or a cheaper one. It is to stop selling intelligence and start owning the consequence of it: the outcome the customer cannot produce alone and cannot remove you from without removing the workflow itself. In the SMB-to-Corporate market this is more available than it looks. These customers rarely have the internal team to own an AI outcome themselves, which is exactly why a partner who will stand behind it becomes harder to replace as the underlying intelligence gets cheaper. Falling model costs become a tailwind instead of a threat, because the cheaper the intelligence, the more the value concentrates in who is accountable for the result.
This is also where partner and platform interests line up. A platform that has worked to make intelligence cheap still needs that intelligence to succeed inside real customer operations, and in the SMB-to-Corporate market it cannot do that account by account on its own. A partner who guarantees the result drives adoption of the platform's now-inexpensive capability into accounts that would otherwise stall. The repricing does not pit the channel against the platform; it relocates where each one earns.
The Common Thread
Commoditization is coming for the markup. It is not coming for accountability. When the intelligence in an offer is racing toward free, accountability for the result is what a customer is still paying for, and it is the layer the platforms are content to leave to a channel. A partner that sells knowledge is selling something that gets cheaper every quarter. One that sells accountability is selling the thing that gets more valuable as everything around it gets cheaper.
Part of a series
This post is one part of the Insights series, our post-by-post working through of the thesis the 2026 Channel Forecast sets out in full.
Sources
- Accenture earnings and the IT-services sector selloff (Cognizant, Wipro, Capgemini, IBM): Reuters, Bloomberg, CNBC, Financial Times.
- McKinsey and consulting-sector workforce reductions: The Wall Street Journal, Financial Times.
- Open-weight model price-performance and task-routing economics: The Information, TechCrunch, Reuters.
