AI Model Providers Are Moving Up The Stack
Reshaping Ecosystems and Sovereignty
While trying to get my head around Anthropic Skills after reading their tag line, Don’t Build Agents, Build Skills Instead, I again released something…
AI Model providers are moving up the stack…here are three reasons I say this…
Recently I asked the question, will the model eat your stack?
I also considered how the subsumption window of AI is moving along and subsuming products which are really just a thin wrapper. And the little IP they have, is easily subsumed by the broadening standard offerings from model providers.
There are also three AI Agent architectures emerging, and the architecture showed-cased by Anthropic recently fits in with the last architecture I list in the article.
And with Skills, Anthropic really doubles down on the third and last architecture. What this approach is working in its favour is powerful reasoning and coding capabilities of their models.
Some background…
Traditional foundation model providers — OpenAI, Anthropic, Google — are expanding beyond raw models into integrated AI Agents and applications.
This vertical integration commoditises base models, captures higher-value orchestration layers and disrupts consumer and enterprise AI landscapes.
Other tech providers face intensified competition, but…model and data sovereignty are challenges. Whenever organisations want to have full control over the flow of their data, or have their own private instances of models, then this approach becomes problematic.
The Upward Shift in AI Stacks
So as I have alluded, providers now prioritise agentic systems that autonomously handle tasks, integrating tools and workflows.
In late 2025, OpenAI, Anthropic, Google, and Microsoft co-founded the Agentic AI Foundation under the Linux Foundation to standardise open protocols for AI Agents.
Key contributions include Anthropic’s Model Context Protocol (MCP) for agent-tool connectivity, OpenAI’s AGENTS.md and Computer-Using Agent (CUA), and Google’s Agent-to-Agent (A2A) standards.
Products like OpenAI’s GPT Atlas agentic browser and Anthropic’s Claude Code exemplify this move, enabling rapid app development and enterprise automation.
Transformations in Consumer AI
Hence consumer AI is evolving from passive tools to proactive AI Agents embedded in devices and apps.
Power shifts to hardware-software integrators enabling local processing for privacy and speed.
Providers branch out with features like OpenAI’s Agent Mode and Google’s Gemini integrations, reducing reliance on third-party apps.
This commoditises generic models, pushing consumers toward seamless experiences in ambient computing and personalised services.
Advancements in Enterprise AI
Enterprise adoption surges, with 2026 focusing on scalable agentic workflows across compliance, revenue operations, and supply chains.
Providers offer end-to-end solutions, such as agent interoperability via MCP, enabling P&L ownership by AI systems.
This industrialisation moves beyond experimentation, with agents handling domain-specific tasks in marketing, legal and engineering.
Enterprises gain efficiency but must redesign processes, creating demand for specialised integrators.
Branching Out From Models to Full Ecosystems
Model providers vertically integrate to own the stack, from training to deployment.
This includes custom tuning for agents, as seen in Claude Code’s coding prowess.
Open-source initiatives like the Agentic AI Foundation accelerate this, with over 60,000 projects adopting MCP since August 2025.
Implications for Other Technology Providers
This expansion challenges startups and cloud giants.
Foundation providers clone successful apps, turning innovations into roadmaps and acquiring threats cheaply.
Cloud providers like AWS partner for scale but compete in infrastructure, risking fragmentation.
Incumbents face disruption across layers — models, apps, devices — potentially eroding valuations. Outsourcing models shift, with AI automating 40–70% of tasks, pressuring suppliers to deliver transparent, audit-ready systems.
New opportunities arise for domain-specific agents and services firms implementing workflows.
The Rise of Model Sovereignty
Sovereignty becomes central, with 72% of leaders citing it as the top 2026 challenge amid regulatory tightening.
Pursue independence from dominant providers, focusing on data control and local AI stacks.
Trends include sovereign AI adoption scenarios
Enterprises adopt independent data planes to avoid lock-in, ensuring compliance across the AI stack. This counters provider dominance, promoting open standards and edge-based physical AI.
This ascent redefines AI value capture.
Providers dominate higher layers, but sovereignty and specialisation offer counterbalances for diverse stakeholders.
Chief Evangelist @ Kore.ai | I’m passionate about exploring the intersection of AI and language. Language Models, AI Agents, Agentic Apps, Dev Frameworks & Data-Driven Tools shaping tomorrow.
COBUS GREYLING
Where AI Meets Language | Language Models, AI Agents, Agentic