The AI Supercycle:
Six Months In

Six Months of Capital, Not Six Months of Hype

Every prior technology supercycle eventually had a moment where the spending stopped looking like enthusiasm and started looking like infrastructure. For AI, that moment is this year. US venture funding hit $412.7 billion in the first half of 2026 — and 86% of every dollar deployed went to AI companies specifically, according to PitchBook's own H1 data. That is not a sector rotation. That is close to the entire venture capital industry re-pricing itself around one thesis.

The capital is no longer confined to model labs either. Google DeepMind is now paying for equity stakes in production studios. Nvidia and SK hynix have signed a multiyear partnership specifically to secure the next generation of AI memory supply. Andrej Karpathy — one of the most closely watched individual researchers in the field — reportedly joined Anthropic this month, extending what is already the most aggressive AI hiring run of 2026. None of this is speculative positioning for a future that might arrive. It is present-tense infrastructure being built by people who have already decided the future arrived.

The physical side of the thesis moved just as fast. BMW's upgraded humanoid robot is now walking the factory floor at its Spartanburg plant — not a demo, an actual production environment. Agility Robotics is going public via a SPAC merger. Unitree unveiled a $650,000 transforming "mecha" robot the same week its Shanghai listing cleared approval. Robotics has stopped being the "someday" half of the AI Supercycle and started being the half you can watch happen on a factory floor.

Seven Moves, Seven Layers of the Same Stack

The table below tracks the specific moves that turned "AI Supercycle" from a thesis into a checkable record this year:

← SCROLL TO SEE FULL TABLE ON MOBILE →

India Isn't Watching This Supercycle — It's Building Its Own Layer of It

India's position in the AI Supercycle this year isn't as a customer of someone else's infrastructure — it's as a builder of a parallel, sovereign one. The IndiaAI Mission has scaled its shared national compute facility to more than 45,000 GPUs, up from an initial 38,000, with another 20,000 already in the pipeline, all offered to startups and researchers at a subsidised rate. That compute backbone now supports 15 home-grown language models and an open repository — AI Kosh — holding more than 12,500 datasets and 300-plus AI models.

On the hardware side, the India Semiconductor Mission has approved 12 new manufacturing projects worth roughly ₹1.64 lakh crore (~$19.7 billion) — one fabrication unit, two compound-semiconductor fabs, and nine testing-and-packaging facilities. Micron's assembly and test facility in Sanand and Kaynes Semicon's plant in the same city are already operating, not just announced. February's India AI Impact Summit in New Delhi drew delegations from over 100 countries and catalysed more than $200 billion in AI-related investment commitments, according to government figures.

The honest caveat: outlay and commitment are not the same thing as delivered capacity, and India's own IT minister has framed the semiconductor buildout as a multi-year, "enduring capability" project rather than a near-term output story. But the direction is unambiguous — India is positioning to be a producer inside this supercycle's hardware layer, not just a consumer of the compute someone else built.

86% of Every Venture Dollar — Is the Concentration the Story Too

The number that makes the AI Supercycle real — 86% of H1 2026 US venture capital flowing to AI — is the same number that should make a long-horizon investor pause. A venture ecosystem where nearly nine of every ten dollars chase one thesis is not a diversified allocation of capital toward a broad technological transition. It's closer to the entire industry making one enormous, correlated bet.

That is not, by itself, a reason to think the thesis is wrong — infrastructure supercycles have always looked like capital concentration while they're happening; that is partly what makes them supercycles. But it is a reason to be honest about what happens if any single layer of the stack — model economics, chip supply, energy availability, or regulatory response — turns out to be weaker than currently priced. A correlated bet pays out big when it's right and corrects hard, in unison, when any part of the chain is wrong. The concentration is the upside case and the risk case simultaneously — it just depends on which layer breaks first, if any does.

Three Paths for the Second Half of 2026

What This Means for Your Decisions

Six months into the year the $1Q Thesis named as the AI Supercycle's proving ground, the honest read is that the thesis is being confirmed faster and more broadly than a single-sector story would suggest — this is now touching venture capital, public-market infrastructure, sovereign policy, and physical factory floors simultaneously. That breadth is itself informative: a thesis confirmed in one narrow place is a bet; a thesis showing up independently across capital markets, national industrial policy, and factory automation is closer to a structural shift already underway.

For India specifically: the IndiaAI Mission and Semiconductor Mission are not adjacent to this story — they are India's own entry point into it. The GPU capacity, the fab investments, and the 4x year-on-year jump in domestic AI startup funding are worth tracking as a genuine second front to the global capital story, not a footnote to it.

The most important thing to watch is not any single deal or funding round — it is whether the 86% venture concentration in AI starts broadening into adjacent infrastructure (energy, materials, logistics) or stays narrow. Broadening is the healthy version of this supercycle. Staying narrow is the version most likely to end in a sharp correction when sentiment eventually turns.