Recently, I witnessed a frenzy of AI announcements that left Wall Street at two extremes: some celebrating the future of productivity, others breaking into a cold sweat over these astronomical valuations.

Within days, virtually all major players dropped a bombshell. Google DeepMind unveiled Gemini 3.1 Pro with a context window of 1 million tokens—exactly the kind of upgrade you’ve been seeing coming for a while, yet still find impressive. Anthropic didn’t fall behind with Claude Sonnet 4.6, focusing on coding and reasoning in long contexts. Meanwhile, Alibaba rolled out Qwen 3.5—397 billion parameters of pure open model.

But what really caught my eye was the spending frenzy on infrastructure. Google, Amazon, Meta, and Microsoft committed something on the order of US$ 650 bilhões for AI infrastructure in 2026. Let me repeat: 650 billion. Is that a sign of confidence or speculation? Hard to say. OpenAI entered this arms race with a US$ 10 billion agreement with Cerebras Systems for wafer-scale chips.

In China, ByteDance released Seedance 2.0—a video generation model that produces realistic clips from text. Of course, it came with controversies over synthetic media, but it shows that innovation and controversy now live side by side.

What I found interesting is that edge computing also gained prominence. Ambiq expanded operations in Singapore, focusing on ultra-low-power AI for wearables and industrial systems. In times of rising energy demand, efficiency has become a competitive weapon.

On the regulatory front, the UK announced plans for free AI training for 10 million adults by 2030. The EU moved forward with a transparency code under the AI Act, specifying how to label AI-generated content. Basically, regulators are running to catch up with innovation.

What impressed me most was seeing AI move out of the labs and into real-world operations. Reuters reported a 10% improvement in writing with AI tools. In biotech, 73% of companies have already adopted AI tools for protein prediction. Lowe’s rolled out voice agents nationwide. Samsung partnered with Gracenote to improve search on smart TVs.

All this frenzy raises the question that won’t go away: is it abundance or a bubble? Wall Street is split. Optimists see a renaissance in productivity through automation. Skeptics see capex expanding and valuations extremely high, with uncertain monetization. For society at large, the risks are even greater—job displacement, misinformation, and opaque systems operating beyond public understanding.

But one thing is clear: the race for AI is accelerating, and no one is standing still. Not regulators, not investors, and not tech companies. It’s a frenzy that will likely define the next decade.