Nefarious Trading Est 2021
⏱ 14 min read Research · Vol. 01 No. 35 · June 2026
NBIS#1 · 8.6 · LEADER ▲ CRWV#2 · 7.4 · LEADER ▲ IREN#3 · 6.6 · STRONG ▲ APLD#4 · 6.0 · STRONG ▲ WYFI#11 · 3.6 · LOTTERY TICKET ▼ MARA#12 · 3.4 · AVOID ▼ KEEL (ex-BITF)#13 · 2.6 · AVOID ▼ NBIS#1 · 8.6 · LEADER ▲ CRWV#2 · 7.4 · LEADER ▲ IREN#3 · 6.6 · STRONG ▲ APLD#4 · 6.0 · STRONG ▲ WYFI#11 · 3.6 · LOTTERY TICKET ▼ MARA#12 · 3.4 · AVOID ▼ KEEL (ex-BITF)#13 · 2.6 · AVOID ▼
Special Report · AI Data-Center Power Rankings · Neoclouds vs Miner-Pivots
The Neocloud
Power Rankings
NBIS · CRWV · WYFI vs 10 competitors · rated & ranked
13
names scored on 5 dimensions · 1 super-cycle · 2 real leaders

You picked NBIS, CRWV and WYFI as the three leaders in AI data centers. Two of them are. The third is a $1.5B lottery ticket that ranks 11th of 13 — and three of the "competitors" are better risk-adjusted plays.

TL;DR — This is a two-tier cohort sitting on top of the AI-infrastructure capex super-cycle. Tier one is the purpose-built neoclouds — NBIS and CRWV — with proven demand, operating scale, and real moats (owned stacks, Nvidia alignment, signed hyperscaler take-or-pay backlogs). Everyone else is a crypto-miner-to-AI "power landlord" pivot, where the only thing that matters is whether contracts are signed and capacity is energized — versus merely announced. On the verdict for your three picks: NBIS is the single best name in the entire group (ARR $1.92B, +674% YoY; ~45% core EBITDA margin; $9.3B cash; $46B Microsoft+Meta backlog; Nvidia equity). CRWV is a legitimate leader but the riskiest one (~$25B debt vs $2.2B cash). WYFI (WhiteFiber) is the lottery ticket — ~$1.5B cap, ~$79M revenue, and an "Initial Customer" that was 70.7% of FY25 revenue and has paused services, so it ranks 11th of 13 on quality. But that's exactly what a lottery ticket is: tiny, fragile, and asymmetric — on a ~$1.5B base, if NC-1 energizes and the signed $865M Nscale deal ramps, the upside is violent relative to its size. So treat your trio as anchor (NBIS) + high-conviction leader (CRWV) + lottery ticket (WYFI) — and size the lottery ticket like one. If you'd rather not run that risk, IREN is the safer third name (EBITDA-positive, signed $9.7B Microsoft + $3.4B Nvidia).
13 names researched and adversarially fact-checked against Q1 2026 filings. Scores & rankings are John's composite of three analyst lenses. Highly volatile, mostly loss-making cohort — analysis, not a recommendation.
§ The Verdict on Your Picks

Two leaders and a lottery ticket. NBIS ✓ anchor · CRWV ✓ leader · WYFI ◆ moonshot.

NBIS #1
The genuine category leader · 8.6/10
CRWV #2
Real leader, highest leverage · 7.4/10
WYFI #11
The lottery ticket · 3.6/10

NBIS and CRWV earn their leader status. Nebius tops both the growth and AI-moat lenses with verified financials, an owned software + data-center stack, Nvidia's ~$2B equity and "Exemplar Cloud" status, and $46B of Microsoft/Meta take-or-pay backlog. CoreWeave is the premier operating neocloud — every leading AI lab as a customer (OpenAI, Anthropic, Google, Meta) plus Microsoft, a $99B backlog, and >1 GW already live — but its leadership is bought with extreme leverage (~$25B debt, ~$536M/quarter interest) and compressing margins, making it the most fragile of the top names.

WYFI is the lottery ticket — play it as one. WhiteFiber is a real, recently-IPO'd company (spun out of Bit Digital, August 2025), but it's ~40x smaller than NBIS by market cap (~$1.5B vs ~$70B), generates only ~$79M revenue, is deeply unprofitable (~$169M/quarter capex burn against ~$76M cash), and carries an existential flaw: its "Initial Customer," 70.7% of FY2025 revenue, has paused services. That's why it ranks 11th of 13 on quality, and why IREN, APLD, CORZ and RIOT all outrank it. But none of that disqualifies it as a deliberate small speculation: on a ~$1.5B base, a signed $865M Nscale colocation deal and the NC-1 build scaling 54→99→200 MW give it genuinely asymmetric upside. Own it as a tiny, sized-to-lose moonshot — not as a third "leader." If you don't want that risk at all, IREN is the safer slot.

In Plain English

Two of your three picks are spot-on: Nebius is the best company in the whole group, and CoreWeave is a true leader (just heavily in debt). WhiteFiber is different — it's tiny, losing money fast, and its biggest customer just paused. That makes it a lottery ticket: keep it, but bet only what you can afford to lose, because the payoff (if it works) is huge relative to its small size. Prefer no gamble? IREN is the safer third name.

§ The Framework

One sector, two completely different businesses. Purpose-built neoclouds vs power landlords renting to AI.

Tier 1 · The neoclouds
Built for AI from day one — NBIS, CRWV
They own the full stack: GPU-dense data centers plus the cloud software, scheduling and inference platforms sold as a service. Real revenue today, real moats, Nvidia alignment, and signed hyperscaler/AI-lab demand. This is a software-and-infrastructure business, not a real-estate one.
Tier 2 · The miner-pivots
Bitcoin miners turning power into AI rent
IREN, APLD, CORZ, RIOT, WULF, CIFR, HUT, GLXY, MARA, KEEL. Their entire thesis rests on one scarce asset: secured, near-term-energizable grid power, converted into leased data-center capacity. They're effectively landlords — the building and the megawatts, not the software.
The only metric that matters in Tier 2
Signed & energized — vs announced & hoped
Within the pivots, the dividing line is whether leases are signed (ideally hyperscaler- or Google-backstopped) and capacity is energized — versus a pipeline of "planned GW" that's years from carrying a paying tenant. That single distinction sorts the Strong names from the Avoids.
In Plain English

There are really two kinds of company here. Nebius and CoreWeave build and run AI cloud platforms — that's a tech business. Everyone else is an ex-Bitcoin miner that has cheap electricity and is racing to build warehouses full of GPUs to rent to AI companies — that's a landlord business. For the landlords, the only thing that counts is: have they actually signed paying tenants and switched the power on, or is it still just plans?

§ The Power Rankings

All 13, ranked. Composite score = risk-adjusted quality across growth, balance sheet, AI moat, valuation and momentum.

#NameTickerScoreTierThe one-liner
1NebiusNBIS8.6LeaderThe genuine leader — explosive verified ARR, ~45% core margins, $9.3B cash, Nvidia equity, $46B MSFT+Meta backlog.
2CoreWeaveCRWV7.4LeaderPremier operating neocloud, all four AI labs, $99B backlog — but ~$25B debt makes it the riskiest leader.
3IRENIREN6.6StrongBest miner-pivot: EBITDA-positive, owned power, signed $9.7B MSFT + $3.4B Nvidia, capacity energizing now.
4Applied DigitalAPLD6.0Strong$31B / 1.2 GW of 15-yr take-or-pay leases, CoreWeave-anchored; back-end-loaded but genuine.
5Core ScientificCORZ5.4StrongColocation revenue inflecting (243→590 MW), but ~76% single-customer (CoreWeave) dependence.
6Riot PlatformsRIOT5.3SpeculativeCleanest balance sheet in the group, 1.7 GW power, AMD anchor — but AI revenue still tiny.
7TeraWulfWULF5.0SpeculativeHPC lease income live (Core42), ~$13B Google-backstopped Fluidstack backlog — heavy concentration.
8Cipher DigitalCIFR4.9SpeculativeAWS 300 MW / $5.5B + Fluidstack, $11B+ backlog — but AI rent only starts ~Aug 2026.
9Hut 8HUT4.8Speculative$16.8B contracted backlog (Anthropic/Beacon Point) — but still ~93% Bitcoin, capacity all pre-2027.
10Galaxy DigitalGLXY4.4SpeculativeHelios 1.6 GW + 526 MW CoreWeave annuity is credible — but core is a volatile crypto-trading firm.
11WhiteFiberWYFI3.6SpeculativeNOT a leader: smallest scale, "Initial Customer" 70.7% of revenue PAUSED, weakest balance sheet, parent-dependent.
12MARA HoldingsMARA3.4AvoidRevenue −18% YoY, ~$1.3B quarterly loss, AI pivot with ZERO signed tenant, first capacity ~mid-2028.
13Keel (ex-Bitfarms)KEEL2.6AvoidWeakest of all: shrinking revenue, negative EBITDA, ZERO signed AI leases, no DC revenue until 2027.
⚠ Two ticker changes you need to know

BITF no longer exists as you knew it. Bitfarms legally re-domiciled and rebranded to Keel Infrastructure Corp. around April 1, 2026; shares began trading as KEEL (~$6.69) on April 6. Any stale "BITF ~$1.98" quote is wrong — that's the old line.

CIFR is now Cipher Digital Inc. (renamed from Cipher Mining), and MARA is the former Marathon Digital. Same companies, new names.

In Plain English

Here's the full leaderboard. The top two (your NBIS and CRWV) are the real deal. Spots 3–5 are the strongest of the ex-miners. The bottom four — including your WYFI — are either tiny, unproven, or have no signed AI customers at all. Also heads-up: "BITF" is now called KEEL, and CIFR is now Cipher Digital.

§ The Scorecard

Where the top names win and lose. Subscores, 0–10 — no one is strong everywhere.

TickerGrowthBalance SheetAI MoatValuationMomentumComposite
NBIS979388.6
CRWV828457.4
IREN747576.6
APLD657566.0
RIOT584565.3
WYFI424453.6

The scorecard tells the real story. NBIS is the only name without a glaring weakness except valuation (P/S ~75x — you pay up for the best). CRWV's balance sheet is a 2 — that's the ~$25B debt. RIOT scores an 8 on balance sheet (net debt/equity ~21%, the cleanest in the cohort) but only a 4 on AI moat because its AI revenue is still tiny. WYFI is a 2 on balance sheet and 4s elsewhere — there's no dimension where it leads. Valuation is the one place the leaders score low: quality this scarce isn't cheap.

In Plain English

No company is great at everything. Nebius is strong across the board but expensive. CoreWeave is brilliant but drowning in debt. Riot has the safest finances but barely any AI business yet. WhiteFiber has no real strength anywhere. The catch with the best names: they're not cheap — you pay a premium for quality.

§ Your 3 Picks, Up Close

The names you backed. Two leaders to anchor — and a lottery ticket to size small.

Nebius$NBIS · #1
▲ 8.6 / 10 · LEADER
Keep — this is the best name in the group. A purpose-built neocloud (ex-Yandex international), ~$70B cap, ~$254. ARR $1.92B (+674% YoY), ~45% core AI-cloud EBITDA margin, $9.3B cash. Microsoft (~$17.4B) + Meta (~$27B) take-or-pay backlog and Nvidia's ~$2B equity. The catch: P/S ~75x, and most of its >3.5 GW contracted power is still unbuilt — size for volatility.
ARR $1.92B$9.3B cashNvidia equity~75x P/S
CoreWeave$CRWV · #2
▲ 7.4 / 10 · LEADER
Keep, but eyes open on the debt. The largest operating neocloud — $2.08B/quarter revenue (+112%), >1 GW live, a $99B backlog, and every major AI lab plus Microsoft as a customer. The risk is real: ~$24.9B debt vs $2.2B cash, ~$536M/quarter interest, $31–35B capex, and compressing margins. High-conviction, high-volatility — not a sleep-at-night holding.
$99B backlog~$25B debtall 4 AI labs+112% rev
WhiteFiber$WYFI · #11 · the lottery ticket
◆ 3.6 / 10 · MOONSHOT
Keep it — but as a lottery ticket, not a leader. The bear case is real: ~40x smaller than NBIS (~$1.5B cap, ~$79M revenue), unprofitable with ~$169M/quarter capex against ~$76M cash — the weakest balance sheet in the group — and its "Initial Customer" (70.7% of FY25 revenue) has paused. The bull case is asymmetry: a ~$1.5B float, a signed $865M Nscale colocation deal, NC-1 (Madison, NC) scaling 54→99→200 MW via Duke Energy, and parent Bit Digital's backing. If it executes, the upside on this base dwarfs the large-caps. Size it sized-to-lose.
~$1.5B float$865M Nscaletop customer pausedasymmetric upside
The safer alternativeIREN · #3
▲ 6.6 / 10 · STRONG
If the lottery-ticket risk isn't for you, this is the swap. IREN is the best miner-pivot: already EBITDA-positive (~41%), TTM revenue ~$757M with AI cloud ramping (+94% QoQ), owned power (Sweetwater 1.4 GW energized April 2026), and signed $9.7B Microsoft + $3.4B Nvidia contracts (plus Nvidia warrants). Less explosive upside than the WYFI moonshot, but the execution is real, not promised.
EBITDA+$9.7B MSFT signed$3.4B Nvidia1.4 GW live
In Plain English

Nebius: keep it, it's the best of the bunch (just pricey). CoreWeave: keep it, but respect the huge debt load. WhiteFiber: keep it too — but treat it as a lottery ticket, a tiny bet you can afford to lose, because the payoff if it works is huge relative to its small size. If you'd rather not gamble, IREN is the safer third pick and already makes money.

§ The Leopold Lens

Why hold a long-shot at all? Because the smartest money in AI owns this exact cohort — and a lottery ticket is how you get convexity on it.

The cleanest frame for this whole basket is Leopold Aschenbrenner's "Situational Awareness" thesis. The ex-OpenAI researcher's fund has gone from ~$225M to ~$20B (+1,000% since 2024, ~+270% in 2026) on one idea: AGI is a near-term, physically constrained event, so the trade is to own the bottlenecks of the buildout — power, data centers, compute, memory, optics — and short what AI eats (he's actually short the chipmakers and owns no Nvidia). Trillions flow into GPUs, data centers and electricity before the decade is out; the binding constraint isn't algorithms, it's megawatts and concrete.

~39%
of Leopold's book is NBIS — his #1 position
NBIS·CRWV·IREN
His "neocloud trifecta" — your 2 leaders + the swap
+1,000%
Fund return since 2024 on this exact theme

Why this matters for your picks: your two leaders aren't a contrarian call — they're the consensus of the most successful AI-infrastructure fund on the Street. Nebius is Leopold's single largest holding; CoreWeave and IREN sit right beside it, and he also owns the Tier-2 power landlords (Core Scientific, Riot, plus crypto miners and Bloom Energy). The leaders are validated by the smartest money in the room.

So why WYFI — the long-shot? It's the same trade, expressed as a convex call option.

Here's the key move in Leopold's playbook that most people miss: alongside the big neocloud equity stakes, his second-largest position has been call options on a small GPU-cloud startup. He pairs core holdings with asymmetric, convex bets on small names that get violently re-rated if the buildout runs as hot as he thinks. WYFI is the retail-accessible version of that bet. You can't buy his options book — but you can own a ~$1.5B small-cap that is a pure bottleneck play (owned/secured power + AI data-center colocation) with the same convexity: if AI compute demand stays insatiable, even a marginal operator with secured megawatts gets pulled along, and on a tiny base the move is explosive. That is precisely what a long-shot is — low probability, high payoff, sized small.

What we're seeing in WYFI · the bull signals
A signed anchor + a real power pathway
A signed 10-year, ~$865M Nscale colocation deal (40 MW) is the anchor that turns story into contract. Flagship NC-1 (Madison, NC) has 54 MW utility power delivered with a Duke Energy pathway scaling to 99 MW within four years and ~200 MW long-term — secured power is the scarce asset in Leopold's whole thesis, and WYFI has a credible slice of it.
What we're seeing · the convexity
~$1.5B float against a ~1,500 MW pipeline under review
Real AI-native customers are already on the books (Cerebras, Modal Labs, Hyperbolic, Boosteroid), and the pipeline under review is ~1,500 MW gross. Against a ~$1.5B market cap, any one of those converting to a signed, energized lease is a needle-mover the large-caps simply can't match per dollar invested.
What we're seeing · the backstop & the catch
Parent Bit Digital funding — and the paused customer
Parent Bit Digital provides funding support (incl. a ~$100M loan) — a lifeline a standalone micro-cap wouldn't have. But keep the long-shot honest: the "Initial Customer" (70.7% of FY25 revenue) has paused, the balance sheet is the weakest in the cohort (~$76M cash vs ~$169M/qtr capex), and it's parent-dependent. This is why it's a lottery ticket, not a leader.

The bottom line: you hold NBIS and CRWV because they are the buildout — proven, validated by Leopold's biggest bets. You hold a small slice of WYFI because, in a thesis where trillions chase scarce megawatts, you want one convex, sized-to-lose call on the tail — the small name that 10x's if the buildout is everything Leopold says it is. Just respect the asymmetry: tiny position, total-loss tolerable, hard milestones (a re-started or replaced anchor customer, NC-1 energized, a second signed lease) as your scorecard.

In Plain English

The most successful AI-investing fund (run by Leopold Aschenbrenner) is built on one bet: AI's real shortage is power and data centers, not chips — so own the companies that have them. His biggest holding is Nebius, and he owns CoreWeave and IREN too, which means your two leaders are exactly what the smartest money owns. WYFI is the "tiny gamble" version of the same idea: a small company with secured power and a signed deal that could soar if the AI boom keeps roaring. You hold it small — like a lottery ticket — precisely because the upside is huge and the risk of zero is real.

§ The Best Competitors

The names from the field worth owning. The pivots that actually signed tenants and turned the power on.

Applied Digital$APLD · #4
▲ 6.0 / 10 · STRONG
The landlord-grade play: ~1.2 GW and $31B of 15-year take-or-pay leases anchored by CoreWeave plus an investment-grade hyperscaler, with Macquarie's $5B facility cutting dilution and phases energizing now. Back-end-loaded, and its GAAP operating loss is larger than headline (positive adjusted EBITDA), but the contracted backlog is genuine.
Riot Platforms$RIOT · #6
▲ 5.3 / 10 · DEFENSIVE
The balance-sheet pick for risk-averse exposure: by far the cleanest leverage (net debt/equity ~21%, only a $200M facility), 1.7 GW of approved ERCOT power, an AMD anchor lease, and Starboard pushing the AI re-rating. AI revenue is still small and single-tenant — speculative, but very well capitalized.
Core Scientific$CORZ · #5
▲ 5.4 / 10 · STRONG
Colocation revenue inflecting fast — 243 MW billable today, path to 590 MW by 2027. But ~76% of revenue is a single customer (CoreWeave), the CoreWeave buyout was terminated, and a new $3.3B 7.75% bond adds leverage. High operating beta to one counterparty.
The Fluidstack clusterWULF · CIFR · HUT
▬ 4.8–5.0 / 10 · SPECULATIVE
WULF (Core42 live + ~$13B Fluidstack/Google backlog), CIFR (AWS 300 MW/$5.5B, rent ~Aug 2026), HUT ($16.8B backlog incl. Anthropic, but ~93% still Bitcoin). All credible, all backlog-rich — but their AI revenue is mostly future, concentrated in Fluidstack/Google-backstopped deals, and years from full energization.
In Plain English

Among the ex-miners, Applied Digital has the biggest signed lease book, Riot has the safest finances, and Core Scientific's rent is ramping fastest (but it depends almost entirely on CoreWeave). TeraWulf, Cipher and Hut 8 all have big contracts on paper, but most of that money doesn't start flowing until 2026–2027.

§ Best Plays

If you're building the basket. Ranked by risk-adjusted quality, not raw upside.

1. NBIS — the only near-"core" holding here8.6
8.6 · LEADER
2. CRWV — leader, but size for the debt & volatility7.4
7.4 · LEADER
3. IREN — best pivot, already EBITDA-positive6.6
6.6 · STRONG
4. APLD — landlord-grade $31B backlog6.0
6.0 · STRONG
5. RIOT — the defensive, cleanest-balance-sheet option5.3
5.3 · DEFENSIVE
★ WYFI — the lottery ticket: tiny size, asymmetric upside3.6
3.6 · LOTTERY TICKET · SIZE SMALL

The shortlist: NBIS is the single best risk-adjusted play and the only name that approaches "core." CRWV earns a spot but should be sized as the high-beta leader it is. WYFI rides along as the lottery ticket — a tiny, sized-to-lose moonshot, not a core position. From the field, IREN (execution, not promises), APLD (contracted backlog), and RIOT (balance-sheet safety) provide the ballast — and IREN is the swap if you'd rather not run the WYFI risk at all.

In Plain English

If you're putting money to work: Nebius first, CoreWeave second (kept small for the debt), WhiteFiber only as a tiny lottery ticket, then IREN, Applied Digital and Riot from the competitor list for ballast. Keep the WhiteFiber bet small enough that a total loss wouldn't hurt — that's what makes it a lottery ticket rather than a mistake.

§ The Risks

What sinks the whole cohort. These are high-beta bets on AI demand staying insatiable.

Cohort-wide risks

  • AI-capex bubble: nearly every name is priced for the buildout continuing at full throttle; a hyperscaler spending pause repricing hits all of them (they already swing 7–15% intraday).
  • Capacity is mostly projected: contracted GW and backlog dollars are often years from energization — interconnect, permitting and GPU/transformer supply delays can slip timelines.
  • Customer concentration: CORZ (~76% CoreWeave), WYFI (70.7% paused), GLXY (sole CoreWeave tenant), and the Fluidstack/Google cluster all hinge on one counterparty.
  • Financing & crypto risk

  • Leverage & dilution: CRWV (~$25B debt) and IREN (rising converts) carry refinancing risk; capital-intensive builds mean ongoing equity dilution and dependence on open markets.
  • Crypto coupling: the pivots still carry Bitcoin exposure that drives violent GAAP swings (MARA ~$1.3B, RIOT ~$500M quarterly BTC marks) and erodes cash before AI revenue ramps.
  • Valuation: NBIS (~75x P/S) and the whole group price in flawless multi-year execution — little margin if a catalyst slips.
  • In Plain English

    The biggest danger is the same for all of them: if AI spending slows, this entire group drops hard. Most haven't actually built the capacity they've promised, many depend on a single big customer, and several still ride Bitcoin's price up and down. And the best ones are priced for perfection.

    § My Take

    John's read. Your instincts on the leaders are right — and WYFI is a fine lottery ticket if you size it like one.

    My Take — Johnny Li
    • You nailed the top two. NBIS and CRWV are the real leaders of this space, and it's not particularly close. Nebius especially — verified hypergrowth, actual margins, a fortress balance sheet, and Nvidia in the cap table. If I owned one name here, it's NBIS.
    • WYFI is your lottery ticket — own it as one. It's a legitimate ~$1.5B micro-cap, but its biggest customer just paused, so it's not a "leader" — it's a venture bet. That's fine if you frame it right: a tiny, sized-to-lose moonshot where the asymmetric upside (small float, the $865M Nscale deal, the NC-1 ramp) is the whole point. Just don't size it like NBIS — a lottery ticket is a small line, not a core holding.
    • Know the difference between the two tiers. NBIS/CRWV are tech businesses; the rest are power landlords. Don't price a landlord like a neocloud or vice versa. For the pivots, I only care about signed leases on energized megawatts — everything else is a pipeline slide.
    • The hedge to the lottery ticket is IREN. If you ever decide the WYFI risk isn't worth it, IREN is the clean swap: it already generates AI revenue, it's EBITDA-positive, it owns its power, and the Microsoft and Nvidia deals are signed. Some traders hold both — IREN for the steady pivot, WYFI for the moonshot.
    • How I'd build it: NBIS as the anchor, CRWV sized smaller for the leverage, WYFI as a small lottery-ticket slice, and IREN / APLD / RIOT from the field for diversified, signed-and-energized exposure. Avoid MARA and KEEL — no signed AI tenants, priced on hope. And remember the whole basket is high-beta on AI capex: position for 15% days, both directions.

    Want the live scorecard + the signed-vs-energized tracker for all 13 names?

    Join the Discord to find out! →
    discord.gg/nfrs · @Nefarioustrading
    Nefarious Trading
    Equity research and trading commentary — AI infrastructure, photonics, enterprise software, power semiconductors.
    AuthorJohnny Li
    Sources & Method
    Figures drawn from Q1 2026 (quarter ended Mar 31, 2026) SEC filings, company press releases and investor materials, cross-checked against StockAnalysis/CNBC/company IR; every material number (market cap, revenue, profitability, power capacity, marquee contracts) was independently fact-checked by a separate verification pass. Key data points: NBIS — ARR $1.92B (+674% YoY), Q1 rev $399M, ~45% core EBITDA margin, $9.3B cash, >3.5 GW contracted, Microsoft ~$17.4B + Meta ~$27B + Nvidia ~$2B equity. CRWV — Q1 rev $2.08B (+112%), $99B backlog, >1 GW live, ~$24.9B debt vs $2.2B cash. WYFI (WhiteFiber, Bit Digital spin-off, IPO Aug 2025) — ~$1.5B cap, FY25 rev ~$78–83M, "Initial Customer" 70.7% of FY25 rev paused; Nscale $865M colocation. IREN — TTM rev ~$757M, EBITDA-positive, signed $9.7B Microsoft + $3.4B Nvidia, Sweetwater 1.4 GW energized. APLD — $31B / 1.2 GW take-or-pay, CoreWeave anchor. CORZ — 243 MW billable, ~76% CoreWeave, $3.3B bond. RIOT — 1.7 GW power, AMD anchor, net debt/equity ~21%. WULF/CIFR/HUT — Fluidstack/Google + AWS backlogs, AI revenue ramps 2026–2027. GLXY — Helios 1.6 GW, CoreWeave 526 MW / $1B+/yr. MARA — rev −18% YoY, no signed AI tenant, capacity ~mid-2028. KEEL (formerly Bitfarms/BITF, rebranded ~Apr 1, 2026) — no signed AI leases, no DC revenue until 2027. Prices/caps ~June 22–24, 2026; highly volatile and may be stale by the time you read this. Scores are a Nefarious composite of three analyst lenses (growth & fundamentals, balance sheet & risk, AI moat & positioning), not a quantitative model output. Leopold Aschenbrenner / Situational Awareness LP framing per his 2024 "Situational Awareness: The Decade Ahead" essay and 2025–2026 reporting (Fortune, WSJ, Motley Fool, 13F-based portfolio trackers): fund grown to ~$20B AUM, ~+1,000% since 2024 inception / ~+270% in 2026; reported top holding Nebius (~39% of disclosed book), with CoreWeave and IREN forming a "neocloud trifecta," plus Core Scientific, Riot, Bloom Energy and crypto-miner exposure; thesis = long the AI-buildout bottlenecks (power, data centers, compute, memory, optics), short chipmakers. Portfolio/stake figures are from third-party 13F trackers and press reports and may be stale or imprecise.
    One trader's view — do your own research. The author's stated positions are not disclosed; nothing here is a recommendation to buy or sell any security. This cohort is overwhelmingly loss-making on a GAAP basis, capital-intensive, dependent on continuous debt/equity raises, and richly valued — these are high-beta bets on sustained AI-infrastructure spending and on flawless multi-year execution of largely unbuilt capacity. Contracted "GW" and backlog dollars are commitments/projections, not operating results, and several names carry severe single-customer concentration. Crypto-linked names (RIOT, MARA, HUT, GLXY, KEEL, IREN, WULF, CORZ) also carry Bitcoin price exposure that drives large mark-to-market swings. Ticker note: BITF now trades as KEEL (Keel Infrastructure); CIFR is now Cipher Digital; MARA was formerly Marathon Digital. Rankings, scores and the trade framing are the author's opinion. Figures are approximate, as-of late June 2026, and not live brokerage pricing. © 2026 Nefarious Trading.