The Great Sorting: Capital, Hype and the Infrastructure Chase

The $965bn Anthropic Series H was a Rorschach test.

You either saw a typo or a market signal.

I saw the latter. It told me we are in a cycle where narrative has decoupled from capital reality. The job is not to catch the falling knife of overvalued private rounds. The job is to find the capital flows that are real, verifiable and already in motion.

This week's threads gave me the map.

Here is how I am playing it.

The Accumulate Zone

Energy Infrastructure

The 44.9 Bcf/d pipeline data is the most concrete signal in the entire dataset.

Data centres need power. AI models need data centres. The grid cannot handle the load without new gas-fired generation and the pipes to feed it.

I am focused on Energy Transfer and Kinder Morgan.

The logic is simple. These assets sit in the physical layer of the AI stack. They cannot be disrupted by software. They cannot be replaced by open source. They have 40-year asset lives and regulatory defences against new competition.

The risk is interest rate sensitivity. MLPs trade like bonds. A hawkish Fed hurts the thesis near term. I size the position knowing this and wait for dips on macro fear.

Computing Infrastructure

Nvidia is the landlord of the AI age.

The Blackwell STAC-AI benchmark for finance was the clearest competitive moat I have seen in months. It signals something specific. The inference market is not commoditising yet. Nvidia owns it.

I hold Nvidia as a core position.

The risk is customer concentration. If the hyperscalers design their own chips and the performance gap narrows, the multiple compresses. I watch Groq and Cerebras. I do not fade the moat until I see the benchmarks.

The Avoid Zone

Private AI Valuations

The Anthropic number is not investable.

Even if it was $965m, the behaviour around it tells me the private market has lost its filter. LPs are writing cheques based on FOMO not fundamentals. When the mark-to-market comes, these rounds will be a liability.

I avoid the narrative and wait for the public filing.

Crypto Sports Infrastructure

Chiliz and the Solana sports thesis have a compelling story.

"Sports industry on-chain. $400bn total addressable market. Fan engagement."

The capital signal says otherwise.

The flows into these chains are retail. The indices Glassnode publishes are an attempt to dress speculation as institutional behaviour. There is no major league balance sheet backing the infrastructure yet. The regulatory tail risk is high.

I fade this until hard institutional dollars arrive.

The Watchlist

Biology and AI

The ESM models from CZI and the 1.1 billion protein structures are the strongest intelligence signal in the dataset.

It is also the hardest to trade.

There is no clean public equity linked directly to this. The beneficiaries are compute providers and big pharma. The path from breakthrough to commercial product is ten to fifteen years.

I keep a watching brief. I do not overweight.

If I had to play it, I would buy Nvidia and call it a biology bet. The bitter lesson of AI is that scale wins. Nvidia sells the scale.

Where I Might Be Wrong

The Energy Thesis

If AI efficiency gains outpace demand growth, the pipeline narrative breaks.

Blackwell itself is more efficient than Hopper. If the curve bends down across the industry, the energy demand thesis for 2030 is overstated.

The Nvidia Thesis

If inference commoditises to ASICs or open source hardware, the moat shrinks.

I respect Groq and Cerebras. I do not yet see them in the benchmarks that matter for enterprise deployment.

The Biology Thesis

If a biotech licensee of ESM goes public and 10x, I will have missed the equity upside.

That is the trade-off. I choose verifiable capital flows over scientific optionality.

The alpha this week is in the structure.

Follow the hard dollars into the ground and the rack. Let the narrative stocks trade on sentiment.

I am positioned for the infrastructure supercycle and I am waiting for the entry points to get better.