NeoCloud Power Score: Bitcoin Miners and AI Clouds Ranked — August 2026

Published by Block 21 Capital · 12 August 2026

CoreWeave reported second quarter results last night: revenue of $2.58 billion, up 112% year over year, a revenue backlog of $104.2 billion, and contracted power of 4.2 gigawatts as of yesterday. It is the company the neocloud category is named after.

It does not top this ranking. The reason it does not is the most useful thing in the table.

The NeoCloud Power Score grades 22 public operators — bitcoin miners converting capacity to AI hosting, plus the two pure-play AI clouds — on seven weighted components, and returns a score from 1 to 100. It answers one question: who can actually convert power into AI compute revenue?

The rankings

#	Operator	Ticker	NPS	MW Contracted	MW Active	Energized	$/Watt
1	Hut 8	HUT	80.5	8,375	1,020	12.2%	$1.37
2	CoreWeave	CRWV	77.4	4,200	1,500	35.7%	$10.62
3	Bitdeer	BTDR	75.7	3,013	1,700	56.4%	$1.07
4	Iris Energy	IREN	73.6	4,500	810	18.0%	$4.38
5	Cipher Mining	CIFR	69.7	5,100	700	13.7%	$1.98
6	Nebius	NBIS	67.9	5,000	600	12.0%	$15.75
7	Marathon Digital	MARA	66.2	4,800	1,282	26.7%	$2.04
8	TeraWulf	WULF	60.6	5,000	522	10.4%	$3.99
9	Core Scientific	CORZ	59.3	4,500	445	9.9%	$1.62
10	HIVE Digital	HIVE	59.0	900	444	49.3%	$1.41
11	CleanSpark	CLSK	56.8	2,300	1,000	43.5%	$2.03
12	Keel Infrastructure	KEEL	51.1	2,200	648	29.4%	$1.18
13	Riot Platforms	RIOT	49.6	2,000	700	35.0%	$5.64
14	Galaxy Digital	GLXY	48.9	1,600	200	12.5%	$3.79
15	BitFuFu	FUFU	41.0	500	478	95.6%	$0.78
16	Applied Digital	APLD	39.3	4,000	286	7.1%	$8.49
17	Bit Digital	BTBT	37.8	100	32	32.0%	$7.02
18	Cango	CANG	37.0	400	341	85.2%	$0.67
19	Soluna Holdings	SLNH	35.2	4,300	170	4.0%	$0.30
20	Canaan	CAN	31.7	250	250	100.0%	$1.49
21	WhiteFiber	WYFI	27.9	116	17	14.7%	$8.19
22	VivoPower	VIVO	22.5	358	42	11.7%	$0.66

Full component detail and underlying metrics sit on the AI & HPC tracker.

Why CoreWeave is second

CoreWeave ranks first in the cohort on contracted AI revenue. It ranks first on GPU deployment. It ranks second on energized megawatts, behind only Bitdeer. On the components that measure operating an AI cloud, it is at or near the top of every one.

It ranks ninth on secured power pipeline.

At 4,200 MW contracted, CoreWeave sits behind Hut 8's 8,375, Cipher's 5,100, TeraWulf's and Nebius's 5,000, Marathon's 4,800, Iris Energy's and Core Scientific's 4,500, and Soluna's 4,300. Eight companies in this table control more future power than the largest AI cloud in it.

That asymmetry is not an oddity in the data. It is the entire structure of this market, and it explains the deal flow. CoreWeave has twice tried to buy Core Scientific outright — roughly $1 billion in cash in 2024, then a $9 billion all-stock agreement in July 2025 — and was rejected both times, the second by Core Scientific shareholders in October 2025. It secures capacity through leases instead, including Galaxy Digital's Helios campus.

The miners have the land, the interconnects and the substations. CoreWeave has the customers, the GPUs and the balance sheet. Neither side has both, and the gap between them is where the deals are being written.

The pipeline trap

Hut 8 leads at 80.5 on a pipeline no one matches — 8,375 MW, roughly double CoreWeave's contracted position.

But 12.2% of it is energized. Hut 8 ranks 16th of 22 on conversion rate. It has assembled the best land bank in the industry and has not yet proven it can turn it on at scale. That distinction disappears in any ranking sorted by announced megawatts, which is how most coverage of this sector is written.

Bitdeer at 75.7 is the mirror image: a smaller pipeline at 3,013 MW, but 1,700 MW already live — the highest active capacity of any operator here, CoreWeave included — at a 56.4% energization rate.

Soluna makes the point from the other end: 4,300 MW secured, 4.0% energized, score of 35.2. Announced capacity without execution is a press release.

Energized is necessary, not sufficient

Riot Platforms is the instructive case mid-table. At 35.0% energization it has one of the better conversion rates in the cohort, ahead of Hut 8, Cipher, TeraWulf and Core Scientific. It has done the hard part.

It scores 49.6 and ranks 13th, because energized megawatts running ASICs are not energized megawatts running GPUs. Riot has no disclosed GPU deployment and no announced hyperscaler agreement. The power is real. The pivot is not yet.

The same reading applies to CleanSpark at 43.5% and Canaan at 100%. High conversion on capacity still committed to hashing does not become AI revenue on its own.

What a watt costs

Market cap divided by contracted megawatts is a crude lens and a revealing one.

Nebius trades at $15.75 per watt and CoreWeave at $10.62. Bitdeer trades at $1.07, Keel at $1.18, Hut 8 at $1.37.

The obvious reading is that the miners are cheap. The honest reading is that the market is pricing execution risk, and it is probably right to. Hut 8 is cheap per watt in part because 88% of its pipeline is not yet delivering anything. CoreWeave is expensive per watt because $104 billion of backlog is already contracted against its capacity.

This is the metric's real limitation: it compares a land bank to an operating business and reports the difference as valuation. Use it to ask why the gap exists, not to conclude that the cheap names are underpriced.

Methodology

Seven components: contracted AI revenue (22%), energized capacity (20%), secured pipeline (18%), conversion rate (12%), GPU deployment (12%), capital efficiency (10%), and self-funding capacity (6%). Each is a percentile rank within the 22-operator cohort; the weighted sum is the score.

Three caveats worth stating plainly. Scores are cohort-relative, so adding or removing an operator shifts every score slightly. Contracted AI revenue is scored on a 0–10 judgment scale from disclosed agreements and is the most subjective input. And CoreWeave does not disclose a current GPU count — because the component is a rank rather than a magnitude, its first-place position is not in doubt, but no precise figure is implied.

CoreWeave and Nebius are included as full cohort members rather than benchmarks. Neither is a bitcoin miner.

Block 21 Capital publishes independent research. This is not investment advice and no position in any security mentioned should be inferred. Data is drawn from public filings, earnings materials, company announcements and market data providers, and may contain errors. Scores are a snapshot as of the publication date.