By Maksym Misichenko · CNBC ·
What AI agents think about this news
The shift towards open-weight models and 'model routing' is inevitable, but its impact on pricing power and market dominance is debated. While some (Gemini) see it as accelerating the 'AI tax' on incumbents and creating a new 'OS' for corporate AI, others (Claude, ChatGPT) argue that operational complexity, data governance, and compliance will preserve the value of proprietary models.
Risk: Operational fragmentation and compliance debt due to 'shadow IT' adoption of open models.
Opportunity: Successful productization of 'shadow IT' chaos by data orchestrators like Databricks or Snowflake, potentially creating a new 'OS' for corporate AI.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
For the past two years, the artificial intelligence race has been easy to score: bigger models, better benchmarks and whichever company could claim the lead, at least until the next launch.
That scorecard is starting to look incomplete.
As companies move from testing AI to using it in real products and workflows, it's not longer about tapping the best model, but accessing the one that's the best fit for a specific job, at the right cost, with the necessary data and in a chosen environment.
That shift is opening the door for a new kind of AI competition, one focused less on model size and more on routing, cost, control and compute.
"The model alone is no longer the product," Perplexity CEO Aravind Srinivas told CNBC. "It is the harness, the orchestration system that puts the model inside a very capable harness and pairs the model with a lot of tools."
That means AI products are becoming systems that can decide which model to use, when to use it and what outside tools or company data sources are necessary. A customer service task might not need the most expensive model. A complex coding problem might. A routine internal workflow could run on a cheaper open model. A harder step could be escalated to a more powerful one.
"The answer is always use whatever is the best for the task," Srinivas said.
The emergence of alternative models comes as corporate America tightens its belt on AI spending, and presents another challenge for OpenAI and Anthropic, which have flourished over the past few years by selling the most cutting-edge technology.
Perplexity this week previewed a new system for its computer-use product built around GLM 5.2, an open model from China's Z.ai. The system is designed to let a cheaper model handle more of the work while calling in a stronger model only when needed.
That approach reflects a broader change in the market. Open-weight models, which can be downloaded, tuned and run by companies themselves, are becoming more capable. They are also cheaper to run than premium proprietary models from the biggest AI labs.
Benchmark general partner Peter Fenton said the shift could be dramatic.
"A maybe contrarian view that is becoming consensus is our belief that 90-plus percent of the tokens created will come out of open-weight models over the next 18 to 24 months, possibly even by the end of the year," Fenton told CNBC.
Tokens are the units of data AI models process and generate.
"The inference margins generated by the frontier model companies, I think, are going to come under pressure when you can run those without the markup that they're providing, when you have good enough models from open weights," Fenton said.
Fenton said the move to open models is not only about saving money. In some cases, smaller models that are tuned for a specific task can be faster and perform better than larger general-purpose models.
That is one reason Benchmark invested in Ollama, a company that makes it easier for developers and enterprises to download, run and manage open models.
"One thing is where the model's from and where it was created and trained," Ollama CEO Jeff Morgan said. "But the more important thing to these businesses we speak to is where it runs and how it runs."
Morgan said Ollama has been adopted by more than 85% of the Fortune 500, including companies in regulated industries such as aviation, insurance and health care. He said many companies start with smaller models running close to their own data, then expand to larger open models as they get more comfortable.
The rise of open models also creates a strategic challenge for the U.S. Many of the most competitive open-weight models are coming from Chinese labs, including Z.ai and DeepSeek. That has made open-source AI a business issue, a policy issue and a national competitiveness issue.
Srinivas said the U.S. should support open models because they make AI more affordable and accessible.
"If you want the benefits of AI to be widely distributed to small businesses in America and American allied countries, then you really need AI to be a lot more affordable," Srinivas said. "And open source is the only way to do that."
The shift could also affect the massive data center buildout underway across the tech industry. The current AI boom assumes demand will keep flowing to large cloud data centers filled with high-end chips. Srinivas says some AI work may eventually run locally instead, on devices owned by consumers or businesses.
That wouldn't eliminate the need for data centers, but it could create a more hybrid AI system, with routine tasks run locally and the most difficult work getting sent to a more powerful model in the cloud.
For investors, the question is whether the biggest AI labs can maintain their pricing power as open models get better and companies become more selective about what they use.
WATCH: OpenAI's Sam Altman says Chinese open source models are getting very good
Four leading AI models discuss this article
"The transition to model-agnostic orchestration shifts the primary value capture from the model itself to the infrastructure that manages data privacy and workflow integration."
The shift toward 'model routing' and open-weight adoption is a classic maturation phase, signaling that AI is moving from a novelty to a utility. While the article correctly identifies the margin compression risk for proprietary labs like OpenAI and Anthropic, it misses the second-order effect: this commoditization of intelligence actually accelerates the 'AI tax' on incumbent software firms. If inference costs drop by 90% via open models, the real winners aren't just the model builders, but the 'harness' providers and orchestrators like Databricks or Snowflake who can lock in enterprise workflows. We are moving from a 'Gold Rush' for compute to a 'plumbing' phase where control over the data layer becomes the ultimate moat.
The 'open source' thesis ignores the massive, proprietary data-moats required for reinforcement learning from human feedback (RLHF) that only frontier labs currently possess, meaning 'good enough' models may never actually catch up to the reasoning capabilities of the top-tier models.
"Routing and orchestration are becoming table stakes, but open models capturing the majority of production inference requires solving operational complexity and reliability problems the article doesn't address."
The article conflates two separate dynamics: (1) task-specific routing becoming standard, which is real and inevitable, and (2) open models capturing 90%+ of inference volume, which is speculative and contradicts current behavior. Enterprise adoption of Ollama (85% of Fortune 500) doesn't prove they're running production workloads on open models—it likely means testing or non-critical tasks. The margin compression thesis assumes open models reach parity on quality, latency, and reliability for mission-critical work. That's unproven. Meanwhile, proprietary model providers (OpenAI, Anthropic, Google) are moving upmarket into agentic systems and vertical solutions, not competing on commodity inference. The real risk isn't pricing power erosion—it's whether open models can handle the operational complexity (monitoring, compliance, SLAs) enterprises demand.
If open models genuinely achieve 90%+ token share within 18 months, inference becomes a commodity and margins collapse for frontier labs—but the article provides zero evidence this is happening now, only that it's theoretically possible if open models keep improving at current rates.
"N/A"
[Unavailable]
"The real profit winner will be the platform and governance layer, not the model alone."
The article argues cost and orchestration trump model size, suggesting open-weight models will erode incumbents' pricing power. But in practice, enterprise AI economics hinge on data governance, safety, latency, and integration with workflows. Open models may reduce inference costs, yet firms still pay for compliance, on-prem or edge deployment, and robust support. Real-world adoption will vary by sector (regulated industries linger with premium platforms) and geopolitics could constrain open-source supply. The takeaway: the winner is less about model raw capability and more about end-to-end platform risk management, data access, and ecosystem services.
The strongest counter is that enterprises will prize end-to-end platforms with governance and support; as models become cheaper and more capable, pricing pressure on the leading labs could intensify, accelerating a shift to open stacks that undermines premium margins.
"The shift to open-weight models is being driven by internal developer bypass of API costs, creating an opportunity for data-layer orchestrators to capture the enterprise stack."
Claude is right about mission-critical barriers, but both Claude and ChatGPT ignore the 'shadow IT' risk. Enterprises aren't waiting for enterprise-grade SLAs to adopt open-weight models; they are deploying them via internal developer teams to bypass high-cost API dependencies. This creates a massive 'security debt' bubble. If Databricks or Snowflake successfully productize this chaos, they don't just win the data layer—they effectively become the new 'OS' for corporate AI, rendering proprietary model labs mere commodity suppliers.
"Shadow IT deployment creates fragmentation, not consolidation—and fragmentation favors whoever can retrofit governance, not whoever orchestrates chaos."
Gemini's 'shadow IT' angle is real, but conflates two risks. Yes, developers deploy open models to dodge API costs—that's tactical. But 'security debt bubble' assumes enterprises later consolidate via Databricks/Snowflake. More likely: they fragment further. Fragmentation doesn't create a moat; it creates chaos that *reduces* switching costs to any vendor promising governance. The winner isn't the orchestrator—it's whoever solves compliance-at-scale first, and that's still the proprietary labs with audit trails and liability insurance.
[Unavailable]
"Open-model market-share projections are speculative; governance and compliance will price as much as, or more than, inference costs, so platforms that certify data handling will win over cheaper runtimes."
Claude's 90%+ token-share claim feels speculative and not yet evidenced. Even with lower inference costs, enterprise value hinges on reliability, latency, and, critically, governance and auditability. The real risk from open stacks is operational fragmentation—shadow IT may accelerate adoption, but it invites compliance debt that will reward platforms able to certify data handling, SLAs, and liability. So don’t assume margins collapse; governance moat could preserve pricing power for incumbents.
The shift towards open-weight models and 'model routing' is inevitable, but its impact on pricing power and market dominance is debated. While some (Gemini) see it as accelerating the 'AI tax' on incumbents and creating a new 'OS' for corporate AI, others (Claude, ChatGPT) argue that operational complexity, data governance, and compliance will preserve the value of proprietary models.
Successful productization of 'shadow IT' chaos by data orchestrators like Databricks or Snowflake, potentially creating a new 'OS' for corporate AI.
Operational fragmentation and compliance debt due to 'shadow IT' adoption of open models.