AI · Retail · Strategy — Week 28

The Market Is Here

Three signals. One reality. SaaS is collapsing, Asia is building local, and the chatbot era is ending. Anyone still debating whether AI is “ready” is already late.

—

Signal 1: SaaS Is Collapsing

Gartner says [80% of enterprise applications shipped in Q1 2026](https://www.gartner.com/en/newsroom/press-releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task-specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025) embed at least one AI agent — up from 33% in 2024. Agents aren’t experimental. They’re default.

But only 31% of enterprises have an agent in production. 88% of pilots fail. The blockers? Governance (57%), evaluation gaps (64%), model reliability (51%). Not technology. Organizations.

Meanwhile, Gartner warns that [agentic AI puts $234B in enterprise application spend at risk by 2030](https://www.gartner.com/en/newsroom/press-releases/2026-07-01-gartner-says-us-dollars-234-billion-in-enterprise-application-software-spend-is-at-risk-from-agentic-artificial-intelligence). When agents, not humans, use software, seat-licensing breaks. Buyers stop paying for dashboards and start demanding outcomes. This isn’t a SaaSpocalypse — it’s a structural collapse. Per-seat SaaS is dying. Outcome-based, agent-licensed services are replacing it.

The 88% pilot failure isn’t a problem. It’s the size of the opportunity for organizations that can actually deploy.

SaaS is collapsing. The gap between capability and deployment is where the next wave of value lives.

—

Signal 2: Asia Goes Local

Europe litigates. Asia builds. The two largest digital economies in Southeast Asia are leading — sovereign infrastructure, hyperscaler local partnerships, and adoption curves that dwarf the rest of the world.

Indonesia: [$10.9B AI expansion](https://zexprwire.com/architecting-indonesias-sovereign-and-scalable-ai-future-inside-the-10-9-billion-tech-shift/), sovereign AI fund via Danantara, [Microsoft $1.7B commitment](https://news.microsoft.com/source/asia/2026/03/31/microsoft-deepens-thailand-partnership-with-more-than-us1-billion-investment-spanning-technology-trust-and-talent/). 92% worker GenAI usage — highest in the world. 18 million businesses on AI tools. [Nvidia-backed Firmus/DayOne building 170,000-GPU campus in Batam](https://www.morningstar.com/news/dow-jones/20260628918/nvidia-backed-firmus-dayone-to-build-indonesia-ai-data-center-campus).

Thailand: [B.Grimm + INET joint venture](https://www.bangkokpost.com/thailand/pr/3279465/bgrimm-and-inet-expand-into-digital-infrastructureasas-service-to-accelerate-thailands-ai-economy) for domestic digital infrastructure. [Digital Edge/B.Grimm $880M for 100MW AI campus](https://www.prnewswire.com/apac/news-releases/digital-edge-and-bgrimm-power-secures-thailands-largest-data-center-loan-of-us880-million-to-advance-ai-and-cloud-infrastructure-302763782.html) in the EEC. [Microsoft $1B+ (2026-2028)](https://news.microsoft.com/source/asia/2026/03/31/microsoft-deepens-thailand-partnership-with-more-than-us1-billion-investment-spanning-technology-trust-and-talent/). [AWS $5B over 15 years](https://www.lightreading.com/data-centers/thailands-digital-future-gets-1b-push-from-microsoft-for-cloud-and-ai). Google $1B, first cloud region open.

Both sovereign nations and global hyperscalers are pouring concrete locally. APAC is projected to host 40% of global data center capacity by 2030. The adoption base is already there — Indonesia’s 92% GenAI usage, Singapore’s 60.9% AI diffusion rate, 46% of Southeast Asian firms beyond pilot stage vs 35% globally.

Asia isn’t waiting for consensus. The infrastructure is being laid now, by the players who will set the terms. Late entries pay rent.

—

Signal 3: The Chatbot Era Is Ending

ChatGPT, Gemini, Claude — the text-in/text-out chatbot is dying. The people who built the current paradigm are already moving past it.

[Wharton’s Ethan Mollick](https://www.oneusefulthing.org/p/the-twilight-of-the-chatbots) said it plainly on June 30: “The twilight of the chatbots.” AI is shifting from co-intelligence — humans guiding chatbot output — to autonomous agents that perform extended tasks with minimal human intervention. The chatbot was a transitional form. It’s over.

The model monopoly is being eroded from two directions:

New paradigms from above: [Murati’s Thinking Machines Lab](https://cryptobriefing.com/mira-murati-thinking-machines-lab-bloomberg-tech/) — $50B valuation, ex-OpenAI CTO — builds “interaction models” that listen, see, and speak simultaneously. Full-duplex, 0.4s response. Not a faster chatbot. A model class where AI collaborates in real-time. [LeCun raised $1.03B for AMI Labs](https://en.wikipedia.org/wiki/Mira_Murati) building JEPA-based world models — explicitly not LLMs. Systems that understand physics and predict before acting. [Qwen’s AgentWorld](https://qwen.ai/blog?id=qwen-agentworld) brought world-model paradigms to open source: agents that simulate environments and rehearse before executing.

Open weights from below: [Qwen-AgentWorld](https://qwen.ai/blog?id=qwen-agentworld) outperformed Claude Opus 4.8 on agent benchmarks (58.71 vs 56.59) — open-sourced, free. [Zhipu GLM-5.2](https://securityboulevard.com/2026/06/zhipu-ai-reportedly-matches-claude-mythos-in-vulnerability-detection/) matches Anthropic’s restricted Claude Mythos on cybersecurity — at $0.17 per finding vs $1.00+. MIT licensed. Downloadable. Free. DeepSeek’s V4 Flash hit 79.0% on SWE-bench Verified with 1M context.

The closed-LLM chatbot — ChatGPT, Gemini, Claude — is being squeezed from both sides. New model classes that aren’t chatbots at all. And open weights that match frontier quality for free. A single-vendor closed-LLM strategy isn’t a competitive advantage anymore. It’s lock-in.

The deployment markets that absorb these next-gen paradigms first are Asian — where 92% of Indonesian workers already use AI daily. The paradigms are invented in San Francisco and Beijing. The demand that makes them commercial comes from Asia.

The chatbot is dead. The model monopoly is over. Stop optimizing prompts and start building for what comes next.

—

The Bottom Line

SaaS is collapsing — Gartner proved it. 80% of apps embed agents, $234B at risk, the deployment gap is the opportunity.

Asia goes local — sovereign infrastructure and hyperscaler partnerships in the region’s two largest digital economies. The adoption curves are already there.

The chatbot era is ending — Mollick declares it, Murati and LeCun build new paradigms, open weights from Qwen and GLM erode the monopoly from below.

SaaS is collapsing. Asia goes local. The chatbot is dead. Late entries pay rent.

—

*Axel Winter — [axelwinter.com](https://axelwinter.com) — “Building Digital and AI Business across Asia”*