Thesis: Q3 FY26 missed EPS (-$0.33 adj vs -$0.24) but printed $3.1B ARR under contract (MSFT $1.9B/yr + NVDA $0.7B/yr) and $2.2B cash; the in-line/no-touch branch fired — hold off at $63 RSI 60 until cohort earnings, scale only on a pullback to $50-55

Refresh of 2026-05-05-iren-deep-dive — re-evaluated against the actual Q3 FY26 print (reported 2026-05-07, three weeks ago). The original was a pre-earnings binary-catalyst call ($54.74, conviction low, "watch don't position"). This update closes that loop. Do not read the original's prices/numbers as current.

What the print actually said (Q3 FY26, quarter ended 2026-03-31)

The original framed this as a binary squeeze-or-implode. The actual print was mixed, not binary — a headline EPS miss wrapped around a materially de-risked contract book.

(a) AI HPC mix vs legacy BTC mining. Revenue is still 77% Bitcoin mining ($111.2M) / 23% AI Cloud ($33.6M) — the pivot is funded but not yet in the P&L at scale. The sequential revenue decline ($240M→$185M→$145M over three quarters) is the BTC-mining side compressing (hash price / difficulty), not the AI side shrinking. The thesis is still forward-loaded: AI revenue is contracted but not yet recognized.

(b) MSFT deal + the new NVIDIA disclosure — this is the real news. The original's #1 open question ("what's the MSFT take-or-pay structure?") is now answered, and a second anchor showed up:

• Microsoft: $9.7B total → ~$1.9B average annual revenue (consistent with a ~5yr term). This is the high-magnitude read the original flagged as the bull case.
• NVIDIA (new): a $3.4B, 5-year air-cooled Blackwell AI-Cloud contract → ~$0.7B/yr ARR, plus NVIDIA received a 5-year right to purchase 30M shares at $70/share (~$2.1B potential investment). NVIDIA taking equity warrants at $70 (above the current $63.54) is a strategic vote that did not exist at the original write-up.
• Total ARR under contract: $3.1B ($1.9B MSFT + $0.7B NVDA + $0.5B Prince George GPU), with a stated target of $3.7B ARR by end of CY2026.

(c) FY26 guide vs capex ramp. Capacity buildout: 480MW on track for 2026 (Horizon 1–4), 1,210MW in build for 2027 (Childress H5–6, Sweetwater 1), against a secured 5GW power pipeline + the 490MW Nostrum (Spain) acquisition. Capex is running ~$1.48B/quarter in investing outflow — the ramp is accelerating, funded by cash + GPU financing + the convertible notes.

(d) Cash vs burn — the original's loudest bear point is partly wrong. The original said "FCF -421.9% / -$780M burn" and "clean balance sheet (D/E 0.00)." The print revises both:

• Operating cash flow is POSITIVE (+$75.3M) and Adjusted EBITDA is +$59.5M at a 41% margin. The cash drain is entirely capex (-$1.48B investing), not operations. "Burning 4× revenue" conflates investing with operating burn — the buildout is the burn, and it is a choice, not a going-concern signal.
• The balance sheet is no longer debt-free. Total liabilities jumped from $1.39B (9/30) to $4.6B (3/31) on ~$3.69B of convertible notes. The original's "D/E 0.00 / equity-funded" thesis is stale — IREN now funds the ramp with converts (dilution risk via conversion + the NVDA warrants), not pure equity. Cash of $2.2B (3/31), ~$2.6B (4/30) covers near-term capex alongside GPU financing.

Current tape (2026-05-29)

• IREN: $63.54 · RSI 60.5 · +13.9% 7D · +39.6% 30D · +53.5% 3M · 17.3% off 52wk high · vsSMA20 +12.6% · strong-up, golden cross.
• The stock gapped through the original's $60 trigger and sits at $63.54, three weeks post-print, RSI a healthy 60.5 (not overbought). It did not implode on the EPS miss — the $3.1B ARR + NVDA equity vote outweighed the headline.

Cohort context (same summary)

• HUT: $124.83 · RSI 74.3 · +19.8% 7D · +64.7% 30D · +136.8% 3M · -1.4% off 52wk high — the cohort momentum leader, now overbought (RSI >74) and at fresh highs.
• CORZ: $26.85 · RSI 67.6 · +12.6% 7D · +34.3% 30D · +62.8% 3M · -4.8% off high — middle of the pack, still has room (RSI 67).
• IREN is the laggard on RSI (60.5) but the contract-book leader ($3.1B ARR vs cohort). The relative read from the original ("is HUT the better risk/reward?") resolves: HUT has run hardest (RSI 74, +137% 3M) and is the most extended; IREN is the de-risked-fundamentals name still digesting.

Decision tree — which branch fired

The print was an EPS miss but with a beat-grade contract disclosure — it does not cleanly map to "beat+raise" or "miss/guide-down." Applying the original's tree to the current tape:

• Not "beat + raise": EPS missed (-$0.33 vs -$0.24 consensus). The "chase to $60+ on volume" branch is technically already live (it gapped past $60), but it gapped on the contract news, not an EPS beat — chasing $63 at RSI 60 after a +39.6% 30D run is buying extension, not a fresh setup.
• Not "miss / guide-down → wait for $40 handle": there was no guide-down — the company raised the ARR target to $3.7B by CY2026-end. A $40 handle is not coming absent a cohort-wide unwind.
• → "in-line → no-touch until next cohort earnings" is the operative branch. The fundamentals de-risked (ARR + cash + NVDA equity) but the price already re-rated past the entry zone. No fresh entry at $63. Scale only on a pullback into $50–55 (vsSMA20 mean-reversion zone). Use next cohort earnings (HUT/CORZ) + the CY2026 ARR-target milestones as the next decision points.

Setup

• Current price: $63.54 (2026-05-29) · RSI 60.5 · +39.6% 30D
• Entry zone (refresh): $50–55 on a pullback toward SMA20 ($56.41) / SMA50 ($47.77). No chase at $63.
• Stop: $44 (carried — below SMA50, ~30d structural low).
• Target: $72 (carried — the NVDA warrant strike-adjacent level; $70 warrants cap near-term as resistance/dilution).
• Action: no-touch / watch. Fundamentals de-risked, price extended. Wait for $50–55 or a cohort-earnings reset.

Bull case (refreshed)

1. $3.1B ARR under contract, dual-anchored. MSFT $1.9B/yr + NVDA $0.7B/yr + Prince George $0.5B. The original's biggest open question (MSFT structure) is answered at the high end, and a second blue-chip anchor (NVDA) appeared. ARR target raised to $3.7B by CY2026-end.
2. NVIDIA took equity at $70. A 30M-share right at $70 (above spot $63.54) is the strongest strategic-validation signal in the cohort — NVDA does not write $2.1B option strikes on names it expects to fail.
3. Operating cash flow is positive (+$75.3M); Adjusted EBITDA +$59.5M at 41% margin. The "cash burn outruns the narrative" framing from the original is half-wrong — operations generate cash; only the buildout burns.
4. Power pipeline is the moat: 480MW (2026) → 1,210MW in build (2027) → 5GW secured + 490MW Nostrum (Spain). This is the Aschenbrenner watts-per-rack thesis with the most secured megawatts in the cohort.
5. Didn't implode on the EPS miss — held and gapped past $60, RSI a sustainable 60.5. The squeeze-or-implode binary resolved upward.

Bear case (refreshed)

1. Revenue is still 77% Bitcoin mining and declining sequentially ($240M→$185M→$145M). AI Cloud is only $33.6M/quarter — the $3.1B ARR is contracted, not recognized. The re-rate prices a ramp that has barely started hitting the P&L.
2. The balance sheet flipped to ~$3.69B convertible-note debt (total liabilities $1.39B→$4.6B in two quarters). The original's "clean / equity-funded" thesis is dead — converts + NVDA warrants = real dilution/conversion overhang.
3. EPS missed (-$0.33 vs -$0.24); net loss widened to -$247.8M (from -$155.4M). Operating income -$233.5M. The income statement is deeply negative and worsening as capex/D&A scales.
4. Capex -$1.48B/quarter and accelerating. Funded by cash ($2.2B) + GPU financing + converts. If GPU financing tightens or the AI revenue ramp slips, the cash runway compresses fast.
5. Price extended: +39.6% 30D, +53.5% 3M, already gapped past the $60 trigger. Buying $63 at RSI 60 after that run is chasing — the asymmetry is gone until a pullback.
6. Cohort is hot and crowded — HUT RSI 74.3 / +137% 3M is melt-up territory. A cohort unwind takes IREN with it regardless of its better contract book.

Catalysts

Open questions (carried / updated)

1. AI Cloud revenue trajectory — $33.6M this quarter vs $3.1B ARR contracted. How fast does recognized revenue close that gap? The Q4 FY26 print is the first real read.
2. Convertible-note terms — strike, maturity, conversion triggers on the ~$3.69B. Dilution math depends on these (NOT disclosed in the summary data; needs the 6-K / 10-Q detail).
3. GPU financing dependence — "near-term capex met through cash + operating CF + GPU financing + additional financing." How much is committed vs aspirational?
4. Insider behavior post-print — the original's loudest signal was -$66M insider selling pre-print. Has that continued or reversed after the contract disclosures? (Not re-pulled here — flag for follow-up.)

Verdict

No-touch / watch at $63. The "in-line → no-touch until next cohort earnings" branch fired.

The print materially de-risked the fundamentals versus the original write-up: the MSFT structure resolved at the high end ($1.9B/yr), a second blue-chip anchor appeared (NVDA $0.7B/yr + $70 equity warrants), operating cash flow is positive, and cash is $2.2–2.6B. But two original assumptions were wrong in opposite directions: the "clean / equity-funded balance sheet" is now ~$3.69B in converts (more bearish), while the "FCF -422% burn" overstated the operating drain (more bullish — the burn is capex, a choice).

The catch is price: IREN already gapped through the original's $60 chase trigger and sits +39.6% 30D at RSI 60.5. The fundamental de-risking is real but the easy asymmetry is spent. Do not chase $63. Scale into $50–55 on a pullback toward SMA20/50, stop $44, target $72 (capped by the $70 warrant level). Track HUT (RSI 74, most extended) and CORZ (RSI 68, mid-pack) as the cohort-flow tells; use their next earnings as the re-evaluation trigger.

Conviction: low — not because the thesis is weak, but because the entry is wrong here. The right setup is a pullback, not today's tape.

Sources

• Q3 FY26 financials (event truth): Massive data/stocks/IREN/massive/{income-statements,balance-sheets,cash-flow-statements}.json (filed 2026-05-08) — revenue $144.8M, net loss -$247.8M, operating CF +$75.3M, investing CF -$1.48B, total liabilities $4.6B, cash flow detail.
• Primary release: IREN "Business Update and Q3 FY26 Results," GlobeNewswire 2026-05-07 — BTC mining $111.2M / AI Cloud $33.6M split, Adjusted EBITDA $59.5M (41%), cash $2,213.3M (3/31) / ~$2.6B (4/30), convertible debt ~$3,687.8M, MSFT $1.9B/yr, NVDA $3.4B / $0.7B/yr + 30M shares @ $70, $3.1B ARR → $3.7B CY2026 target, 480MW/1,210MW/5GW capacity, Nostrum 490MW.
• SEC Form 8-K (FY2026): sec.gov/Archives/edgar/data/0001878848/000187884826000025/iren-20260507.htm
• EPS / consensus: -$0.33 adjusted vs -$0.24 consensus (miss).
• Price truth (current tape): ai-power (data_as_of 2026-05-29) — IREN $63.54 / RSI 60.5 / +39.6% 30D; HUT $124.83 / RSI 74.3; CORZ $26.85 / RSI 67.6.
• Original deep-dive: 2026-05-05-iren-deep-dive (format reference; prices stale).
• Perspective: 2026-05-03-ai-power-bottleneck (cohort frame).