CIFR research

Position in the Physical Layer of AI thesis

CIFR is the Datacenters pivot. NBIS is the anchor (clean neocloud growth); CIFR is the BTC-mining-to-AI-hosting pivot. Cipher Mining built out gigawatt-scale BTC mining infrastructure on cheap Texas power; now the company is converting some of that capacity to AI-hosting contracts. The thesis: the buildout already happened, the power contracts are signed, the only question is whether AI hosting margins beat BTC mining margins through the cycle.

Quinn's book opened CIFR after the manifesto v2 sleeve definition. Position sized at <2% of book per portfolio snapshot.

Recent catalysts (last 30 to 60 days)

First HPC hosting contract signed. Disclosed Q1 2026: multi-year HPC hosting agreement with a "leading AI infrastructure customer." Customer name not public; contract value disclosed in 10-Q footnote.
Texas power capacity expansion. Additional 250MW of power capacity announced for the Black Pearl facility, with grid interconnect commitments through 2027.
BTC halving aftermath. The April 2024 halving + April 2028 halving brackets define the BTC mining margin compression cycle. CIFR's AI pivot is a hedge against that compression.
Q1 print: BTC profitability + AI optionality. BTC mining still drives revenue; AI is the strategic narrative.

The thesis (what has to be true)

1. AI hosting margins beat BTC mining margins on Texas power through 2027. If BTC pricing rallies hard, the AI pivot stalls. If BTC compresses, AI hosting is the path. 2. The HPC customer contract converts to a multi-customer base. Single-customer concentration is the immediate risk. 3. Texas ERCOT power pricing stays favorable. Sub-$30/MWh average price is the floor for CIFR's economics. 4. The grid interconnect commitments hold. Texas has had interconnect delays elsewhere; CIFR's pipeline depends on the queue resolving.

Kill vectors (what would break the thesis)

BTC bull market. If BTC rallies past $200K and stays there, mining margins make the AI pivot less attractive financially.
Customer concentration shock. If the lead HPC customer cancels or doesn't renew, the AI optionality story loses the proof point.
ERCOT pricing shift. Texas power price spikes during summer 2024-2026 have hit mining margins; another structural shift hurts.
Capital event. CIFR has raised capital historically; dilutive secondary at low price is a discipline break.

Layer context

Datacenters layer per current metadata, alongside NBIS (anchor, clean neocloud) and IREN (BTC-mining-to-AI pivot, similar profile). CIFR is the most-explicitly-pivoting of the BTC-AI hybrids in the book.

Position discipline

Pivot sleeve rules.
Trim trigger: +100 percent from entry.
4-quarter thesis-fail rule applies. If HPC contract doesn't expand within 2 quarters or BTC margins overwhelm AI optionality, the clock starts.

Moat 5. Asym 6. The moat is the lowest of the Datacenter names because the BTC-to-AI pivot is not exclusive to CIFR (IREN, MARA, others are doing similar). The asym is moderate because the power capacity is real, but the customer concentration is concentrated.

Real money. Real position. Real receipts.