R40

Tracking what matters: Rule of 40, free cash flow, margins. Zero BS. 📈 (

[R40 @r40_io]

3.5T in revenue is a lot 🤣.. it is calculated by taking jan revenue 2033 x 12 ... or sum of the quarters revenue... just asking because probably the difference is like the size of a mid size country in EU

[Muskonomy @muskonomy]

NEWS: Elon Musk says SpaceX could hit $3.5 trillion in revenue by around 2033.

Morgan Stanley's model has SpaceX reaching $3.5 trillion in revenue by 2040. Elon Musk thinks it happens far sooner.

He replied: "My best guess for ~$3.5T revenue is roughly around 2033 fwiw."

That puts Elon Musk about seven years ahead of the bank. Others pointed out the 2040 figure is already cautious, close to half of what SpaceX itself is aiming for.

[R40 @r40_io]

The interesting from $AFRM earnings:

They guided a slowdown and then printed another 30%+ GMV quarter. On 7 May they told the Street FQ4 GMV would be $13.15–13.45B, or +26–29% — the first sub-30% quarter after ten in a row. They did $14.1B, +36%. That is the 11th consecutive 30%+ print, and every guided line (revenue, RLTC, both margins) came in above the high end.

The $4.62 EPS is a tax item they had telegraphed. GAAP net income is 1.6B because they released the U.S. deferred−tax valuation allowance (1.45B non-cash benefit). Pretax is 169M, about **0.48** a share. Street was at $0.33 GAAP and had not sized the release. Levchin, verbatim: he “had to check with my accountant” and called the $1.6B “currently ludicrous.”

First $50B GMV year, and he is already talking about $100B. FY2026 GMV was $50.2B. FY2027 is guided to more than $64B, with GAAP operating margin “more than 14.5%” and adjusted “more than 30.5%.”

Card is now the D2C story. 5.2M active cardholders (+125%), $2.8B Card GMV (+124%), 19% attach. Direct-to-consumer GMV +49% to $4.7B, “driven entirely by Affirm Card.” Top-five partner share fell to 42% from 46%.

Funding got cheaper while the book grew. Average cost of funds 5.8%, −103 bp year-over-year. Funding capacity $30B. Net cash 1.5B, +400M in a year. Fourth consecutive ‘X’ ABS with a duration-weighted yield under 6%.

Credit is not the scare in this letter. DQ30+ ex-Peloton and Pay-in-X is 2.5% (+19 bp YoY, −26 bp sequentially). Recent monthly-installment cohorts are tracking ~3.5% lifetime losses; Pay-in-4 vintages still under 1% of GMV.

[R40 @r40_io]

Interesting from Marvell:

Revenue $2.7393B beat the guide mid-point by $39.0M and non-GAAP EPS beat at $0.94 vs $0.93 — but GAAP diluted EPS printed $0.33 against a guided $0.37 ±$0.05, below the range. GAAP opex also overshot: $995.9M against ~$960M guided. Stock comp of $326.2M (from $207.6M last quarter, $153.6M a year ago) is the whole gap.

Data centre is now the company. $2,171.5M, 79% of revenue, +46% YoY and +18% QoQ. Communications and other fell 3% sequentially to $567.8M. A year ago the split was 74/26.

Growth is accelerating three quarters running — +27.6%, +36.5%, and +51.8% implied by the $3.150B Q3 guide. Murphy raised the revenue outlook for both fiscal 2027 and fiscal 2028 without attaching a number to either; the 6 October Investor Day is where that lands.

But margin is guided down into the ramp. Non-GAAP gross margin 58.9% reported → 57.5–58.5% guided, and GAAP opex guided up to ~$1.015B. Custom silicon carries a lower margin than merchant connectivity, and the release explicitly flags "significant acceleration in our Custom business beginning in the second half of fiscal 2027."

Last quarter's $0.04 GAAP EPS was an artefact. The Celestial AI earn out revaluation was $250.7M in Q1 and $52.0M now; GAAP net income went $34.5M → $308.0M on revenue up 13.3%. The GAAP/non-GAAP wedge is 2.8x this quarter, not the 20x that made headlines in May.

Capex up 167% YoY for a fabless designer — $126.7M vs $47.5M, 4.63% of revenue vs 2.37%. FCF grew only 15.6% against 36.5% revenue growth and fell sequentially ($478.8M vs $483.1M).

[R40 @r40_io]

$MRVL (Marvell) expected earnings after close:

https://t.co/iqe3NwpMaB

[R40 @r40_io]

This is interesting $IREN: Pricing is going up with scarcity: recent 3-year contracts at >$20M per MW (IT) with ~2-year payback, and active discussions at ~$25M/MW

25M/MW is a lot! anyone that owns GW should be happy

[R40 @r40_io]

AI Cloud revenue $70.5M passed Bitcoin mining $66.7M — 51% of revenue, from 23% one quarter earlier. AI Cloud roughly doubled QoQ (+110%) while mining fell 40%. Total revenue still went backwards, $144.8M → $137.2M, and −26.7% YoY.

The $1.8bn operating cash flow is not earnings. $1,722.2M of the $1,811.1M is deferred revenue — customer prepayments. Balance-sheet deferred revenue is now $1.84bn. On the site's own FCF convention the quarter is −$167.1M, and the full year is −$2.23bn. Anyone quoting "IREN generated $1.8bn of cash" is quoting a prepayment.

Mining is being written off. $450.4M impairment in the quarter, $638.8M for the year, described as decommissioning mining hardware as sites convert. That's the bulk of the $684.0M net loss and the $(620.4)M operating loss.

Adjusted EBITDA collapsed with gross profit flat. Gross profit $103.9M vs $104.9M; adjusted EBITDA $19.2M vs $59.5M, margin 41% → 14%. The entire delta is SG&A, $81.8M → $128.3M. They're spending ahead of the ramp, not losing unit economics.

The funding structure is the real story. $3.6bn investment-grade GPU financing at 6.0% for the Microsoft contract (funds 96% of associated GPU capex with prepayments), plus $2.8bn of non-IG facilities including $2.4bn led by Blue Owl and PIMCO at 9.0%. Customer prepayments run 45–55% of GPU capex. And yet they still issued $2.1bn of equity in the quarter on top of $3.0bn of new converts — the "capital-efficient" financing has not replaced dilution.

Pricing is going up with scarcity: recent 3-year contracts at >$20M per MW (IT) with ~2-year payback, and active discussions at ~$25M/MW. $4bn contracted ARR for 2026 capacity, $1bn operating as of 26 August; 2026 "largely sold out"; pipeline >5 GW. A new unnamed frontier AI lab joins Cohere, Perplexity, Figure, Prometheus, Fal and Higgsfield.

[R40 @r40_io]

$IREN expected earnings after close (alternative eps −$0.55 ; revenue $135.60M):

https://t.co/vDn6E7uAoE

[R40 @r40_io]

$AFRM EARNINGS:

EPS: $4.62
REV: $1.17B

massive EPS jump is due taxes to be confirm on call

[R40 @r40_io]

$AFRM (Affirm) expected earnings ($0.24–$0.50 range) after close:

https://t.co/j08EZsKz7q

[R40 @r40_io]

$AFRM Affirm and Shopify launch Shop Pay Installments in Australia

[R40 @r40_io]

$IREN Q4 EARNINGS FY2026:
Revenue $137.2M vs Est. ~$140M
EPS GAAP ~-$1.91 vs Est. -$0.55

[R40 @r40_io]

$MRVL Q2 EARNINGS FY2027:
Revenue $2.74B vs Est. $2.71B
EPS $0.94 vs Est. $0.94

[R40 @r40_io]

What to watch on $MRVL
(Marvell) earnings:

The Q3 revenue guide against roughly $3B. It is the cleanest test of management's promise that growth accelerates each quarter.

Data-center revenue and mix, against $1.833B and 76% in Q1. The consolidated beat matters less than which part of AI infrastructure produced it.

Gross margin, against a Q2 guide of 58.25–59.25% non-GAAP and 52.1–53.1% GAAP. Marvell is absorbing Celestial AI and XConn while scaling custom silicon; the gap between the two bases remains large.

[R40 @r40_io]

What to watch on Affirm earnings:

The fiscal 2027 guide, not the quarter. This is the year-end print, and it is the only number on Thursday that speaks to anything the model does not already assume.

The income tax line. If a valuation allowance release lands, it is the largest item in the release and it is non-cash. Read EPS before and after it, and treat any headline that does not separate them as unusable.

Operating margin against the guided 9.5–11.5%. This is where an operating beat has to show up. Above 11.5% is a genuine surprise; the Street's high estimate needs about 15%.

GMV against $13.15–13.45B, and the take rate against 8.52% trailing. Affirm puts GMV in the first paragraph of the letter. Divide revenue by it. The guide implies the first sub-30% growth quarter in eleven.

Delinquencies and funding costs. In March, management said delinquencies rose but pre-attributed roughly half the increase to a smaller denominator from tax-refund seasonality, and reported the lowest average cost of funds in three and a half years.

[R40 @r40_io]

What to watch on IREN earnings:

AI Cloud revenue for June. The comparison is the $33.6M reported in March; this is the income-statement proof of the transition.

Commissioned GPUs and utilization. Separate installed hardware from capacity that is merely ordered or planned.

The schedule for Horizon 2–4. Horizon 1 was delivered to Microsoft in August; each deployment is 50MW of critical IT load.

Capital still required to reach 480MW. Put the remaining bill against the $3.65B GPU financing closed in June and customer prepayments.

[R40 @r40_io]

@Ashton_1nvests If they move faster with this, it changes the value of
$UBER
in the forward model https://t.co/gSyXuTdst9

[R40 @r40_io]

@MatthewDR probably a developer use the wrong constant and the switch between cameras and lidars didn't work

[R40 @r40_io]

@niccruzpatane 3 years in testing + 6 years expanding = 200m waymo miles

vs

8 months model y testing + 3 months custom model =
380,000 cybercab miles

[R40 @r40_io]

@XFreeze this is great

[R40 @r40_io]

@StockSavvyShay copy of grok bots but with a few billions people as potential users