OpenAI’s latest release, GPT-5, landed with more of a shrug than a standing ovation. A recent Financial Times piece captured this mood perfectly: progress now feels incremental. Benchmarks improve slightly, personalities evolve, but the “wow” seems to have shifted from model performance to the gossip around who’s joining which lab or how much Meta is paying for talent. This isn’t a slowdown, it’s a signal. AI is moving into a new phase where the real action is no longer in bigger models, but in what we build on top of them: applications, devices, and services.
We’ve seen this script before. In the early days of the internet, all the excitement (and market value) sat in routers, telcos, and fibre lines. Cisco once wore the crown of the world’s most valuable company.
But infrastructure commoditised quickly, and the real breakthroughs came higher up the stack: search engines, social platforms, and smartphones. AI is on the same trajectory. The foundation models will become the equivalent of digital highways, and the real value will emerge in the “cities” built upon them.
The ChatGPT “meh” moment marks that turning point. Four shifts show us why:
1. Scaling Laws Are Hitting Limits
Since GPT-3, the formula was simple: throw more compute and data at the problem. But the “low-hanging data” is gone, compute and energy are scarce, and the returns from bigger models are diminishing.
Labs now obsess over costs, latency, and reliability, very much the language of infrastructure, not consumer magic.
2. From ‘A Model’ To ‘A Moment’
People don’t wake up asking for a foundation model; they want outcomes in real time. That’s why companies like Perplexity, Cursor, Copilots, and Gemma are thriving. They turn complex models into simple, useful moments: instant answers, fast code, beautiful designs.
The next wave will be even more personal: AI tutors, financial planners, travel concierges. The excitement shifts to how AI makes life better in the moment.
3. AI Breaking Out Of The Browser
The next leap is hardware. On-device AI from Apple Intelligence to Gemini Nano is making assistants faster, private, and offline-capable. Meanwhile, new form factors will redefine “ambient AI”: smart glasses that track context, wearables that summarise your day, and cars that act as copilots rather than dashboards.
Just as the iPhone transformed the internet era, new devices will reshape the AI era.
4. Enterprise Services: India’s Sweet Spot
If models are infrastructure, services are deployment. Global labs are already embedding engineers into clients to wire AI into workflows. That’s familiar ground for Indian IT majors, but the model is shifting from billing hours to delivering outcomes, from generic consulting to specialised copilots in domains like sales, compliance, and contracts. Agentic automation is the next step in systems that don’t just draft, but act. This is India’s chance to lead.
So, is the “model era” over? Not quite. Research will continue on video-native models, memory, and tool use, but the gravitational pull is clearly upward. As labs begin to look more like infrastructure providers, the real race is no longer about size, but usefulness.
The internet didn’t slow down when routers commoditised; it took off. AI is about to enter its own “apps, devices, and services” decade. The winners won’t be those boasting about which model they use, but those who turn models into moments, moments into trust, and trust into growth.
The model wars are fading. The moment wars are just beginning.
(The author is the CEO of AI&Beyond)
Disclaimer: The opinions, beliefs, and views expressed by the various authors and forum participants on this website are personal and do not reflect the opinions, beliefs, and views of ABP Network Pvt. Ltd.
 "‘Canada had no evidence’: Former India envoy slams Trudeau for politically motivated Nijjar case")
