Built byAlpha Pod
Forensic Note · July 2026
Nebius Group · NBIS · Weekend forensic

The debt is extreme. The real question is whether the growth curve catches the capital curve.

Chanos actually understated it. Today Nebius spends closer to $10 of capital per $1 of revenue on an annualized basis — 18× on a trailing-twelve-month base. The $4× figure isn't the current state; it's roughly where the ratio lands if management delivers the 2026 revenue guide against today's debt stack. What isn't in question: Nebius already spends more of each revenue dollar on interest than any peer we can find. Whether that ends as a debt trap or a growth story depends entirely on the top-line curve landing where management says it will.

Original post · @RealJimChanos
"Many of the retail bulls on the neoclouds don't seem to understand how capital-intensive this business model is. For example, it is currently costing $NBIS over $4 in capital employed to generate $1 in revenue. The growth is coming at a very high cost (and commensurately low ROIC)."
Our reply
Correct. Their debt is insane and it eats into all their profits. But the open question is whether they grow into this capex or not.
Alpha Pod · what the filings actually show
Chapter 01 · Where Nebius sits

Every AI infrastructure company borrows to grow. Nebius spends 2–3× more of each revenue dollar on interest than any peer.

TTM interest expense divided by TTM revenue, ranked. Data centers, hyperscalers, and pure-play AI compute names on one axis. Nebius is the outlier by a wide margin.

Peer comparison: NBIS interest/revenue vs CoreWeave, DLR, ORCL, EQIX, IREN, APLD
Figure 1 · Peer comparison, trajectory, and break-even revenue Download PNG
The point

NBIS spends 13.8% of every revenue dollar on interest. The next-closest peer is CoreWeave at 9.3%. Applied Digital, Iris Energy, and Equinix all sit near 5%. This is not a "capex-heavy business" problem — every name here is capex-heavy. It's a Nebius problem, and the direction of travel is up.

Chapter 02 · How we got here

Six quarters ago, interest was rounding error. Today it is 16% of revenue.

Nebius resumed operations post-divestment with net cash. Interest income exceeded interest expense every quarter through Q1 2025. By Q4 2025 the cross was decisive. This is a very recent problem, and it's still accelerating.

Quarterly interest expense vs interest income Q4 2024 through Q1 2026
Figure 2 · Interest expense vs. interest income · Q4 2024 → Q1 2026 Download PNG
The point

Q4 2024: net financing cost was -58% of revenue (they earned money on cash). Q1 2026: net financing cost is +12.4% of revenue. That's a 70-point swing in six quarters — and interest income is still meaningful because $9.3B of raised cash sits on the balance sheet earning yield. Once that deploys into capex, the offset disappears.

Chapter 03 · What the coupon hides

The stated coupon says 1.9%. The economic cost is 5.09%. A $1.57B deferred obligation.

All six senior unsecured convertible notes accrete to 115–125% of face at maturity. That premium is a real cash obligation on the maturity date — it just doesn't show up as interest expense the way a coupon does. GAAP amortizes it through the P&L over the life of the note, which is why reported interest expense is climbing faster than the coupon math alone would predict.

Face vs accreted-at-maturity for all six NBIS convertible note series
Figure 3 · Cash face vs. accreted-at-maturity, per series Download PNG
The point

$8.5B in cash face today. $10.07B owed at maturity between 2029 and 2033. The gap — $1.57B — is deferred interest hidden inside a headline coupon that reads like a growth-stock number. Any analysis that uses the stated 1.9% coupon undercounts the real economic burden by a factor of 2.7×.

Chapter 04 · The Chanos ratio, tested

Chanos said $4 of capital per $1 of revenue. The current run-rate is 9.8× — worse than he claimed.

There's no single "right" way to test the claim. On a trailing-twelve-month revenue base it's 18×. On a Q1-annualized base it's 9.8×. On FY24 continuing operations it was 36×. On no reasonable measure is it currently at 4×. To hit the Chanos target, Nebius has to grow into it — the exact question our reply raised.

Capital stack and capital-employed per dollar of revenue
Figure 4 · Capital stack + capital-employed-per-revenue ratio, four bases Download PNG
The point

The Chanos $4 claim is directionally correct as a forward projection: if management hits the 2026 guide of $3.0–3.4B, the ratio compresses to 4.6–5.2× — right at his line. If they miss, it stays above 9×. This is the entire debate on Nebius reduced to one number.

Chapter 05 · The grow-into-it math

Under management's own guidance, interest-to-revenue collapses from ~14% today to ~5% by 2028.

Scenario A holds debt flat at $8.5B and models three revenue paths. Scenario B adds $2.5B in each of 2027 and 2028 to fund capex. In every path except a hard bear case with continued borrowing, interest normalizes.

Scenario grid: interest-to-revenue ratios under bull/base/bear revenue paths and static vs growing debt
Figure 5 · Interest / revenue under 18 scenarios · 2026–2028 Download PNG
The point

Base case 2028 (static debt, $9B revenue): 4.8%. Base case 2028 with another $5B borrowed: 7.9%. Bear case with growing debt: 10.9% — better than today. The math for growing into it exists on paper. It requires revenue landing where management says it will — a demand-side, sales-execution, GPU-utilization question, not a balance-sheet question.

Where Chanos is right

The current-state critique is defensible.

NBIS's capital-per-revenue ratio, interest-to-revenue ratio, and effective debt cost are all worse than every reasonable peer. The 1.9% coupon narrative undercounts real cost by 2.7×. On today's revenue base, Nebius is capital-inefficient by any yardstick. None of this is arguable.

Where the bull case survives

Every ratio compresses fast if the top-line lands.

Management's $3.0–3.4B 2026 guide, coupled with the $21.3B backlog reported at year-end, is what has to convert. If it does, the Chanos $4 rule is met by 2026 and interest normalizes below 5% by 2028. If revenue misses, the equity holders pay for the miss twice — once in dilution from the convert, once in growing accreted principal.

Alpha Pod is an educational publisher, not a registered investment adviser or broker-dealer. Everything here is for educational purposes only and is not investment advice. Charts are derived from filings but include forensic reconstructions and forward-looking scenarios; verify against primary sources before making any decision.