We live in the Post ChatGPT Era

Generative AI is no longer about one killer product or chatbots. Some thoughts on AI in 2026.

FT. In 2025 we learned Sam Altman wasn’t what he seemed. Nor is the trajectory of OpenAI a sure thing.

For three years ChatGPT dominated how we experienced Generative AI.

But that season in late 2025 is coming to an end. New winners are emerging like Anthropic, Google, Cursor, Alibaba, Loveable and others. As I slow down a little in the next six or seven weeks, I’ll be pondering the AI and emerging tech space as a whole. What is the big picture that is emerging?

Models like Gemini 3 and DeepSeek-R2 (not yet released) will signal a shift into new directions, a 2026 world where Qwen models proliferate and Anthropic begins to scale faster. OpenAI is becoming less the major driver and ChatGPT is no longer the primary catalyst to AI adoption. 2026 heralds a fundamentally new world, where BigAI matures with the backing of BigTech.

This shift marks a super-charging of inference, corporate bonds and AI Infrastructure partnership deals. The predictive decisions of AI thinking or research models and reasoning models are a lot more compute intensive in their demand for compute even before Agentic AI becomes ubiquitous. Google estimates that 60% of AI’s energy now comes from inference. Meta says 60 to 70%. AWS, 80-90% of its ML compute demand.

This week we found out that Microsoft and Nvidia are teaming up to invest $15 Billion into Anthropic into yet another circular vendor financing deal. To meet capacity and future capacity BigTech and BigAI are having to resort to elaborate partnerships, schemes and accounting tactics.

The era of ChatGPT Dominance is Over

The Anthropic deal with Nvidia and Microsoft and the Gemini 3 launch tells me this is becoming less and less about OpenAI. As if that was not abundantly clear with Alibaba Qwen and DeepSeek racing with complementary open-weight models that are gradually approaching the capability of their close-source models cousins in the West.

Curious to see Anthropic mimicking the partnerships deals like OpenAI has done:

“For the first time, NVIDIA and Anthropic are establishing a deep technology partnership to support Anthropic’s future growth.” - Anthropic

The demand for compute will keep rising due to:

• Higher inference demands:

• More global AI adoption, with the proliferation of:

1. Reasoning models (Deep Thinking tasks)

2. Agentic AI adoption

3. Synthetic video like Veo, Sora and others.

4. Agentic commerce and Browser automation

5. A glut of AI startups being funded by American giants of Venture Capital

6. An explosion of the number of LLMs over the last three years that seems to be accelerating

7. Enterprise and commercial adoption of Gen AI products globally

8. BigTech will continue to take out corporate bonds (debt) since their cash balances are so healthy. Recently Meta, Google and Amazon have with more to come to fund costly AI Infrastructure and the energy infra required to power it.

Synthetic Video Based Inference will Skyrocket in 2026

If 2025 taught us anything it’s also that video and music based LLMs and inference is just getting started and is here to stay. This will become rather inference intensive if you consider the shear volume.

a16z

Learn more about Google Antigravity, an agent-first coding tool.

How impressive is Gemini 3?