How DeepStack settles a question · Part 11
Do orders become profit?
Do the orders turn into profit?
From Part 11: The Backlog Is Not the Quarter. 15 voices on 3 sides, attributed and dated; the DeepStack Call with its history; 5 dated tests that settle it; the results as they come in.
The mechanism
AI suppliers report record order books, and the near book will ship: prepaid, supply-constrained demand converts into revenue at the pace of capacity. Profit arrives later than revenue. Between the order and the margin sit ramps that absorb fixed costs ahead of volume, payment terms stretched to a year at Nvidia for its best-rated customers, revenue deferred until a deployment is complete, a far book outgrowing the near one three to one at Microsoft, 113 percent against 37 percent, and assets whose life the buyers' own lenders now cover with guarantees, $28 billion at Meta. So the Call reads the near book as revenue, the far book as a promise and the margin as the number that arrives last, and expects a margin that lags the ramp to be reported as a demand miss.
Approved by Marco Cauduro on October 10, 2026.
The DeepStack callSplit
The orders will ship. The profit comes later, and a margin that lags the ramp will be reported as a demand miss.
Our Call takes the side “Shipping is not earning”.
Call history
- v2Revenue yes, margin not yetConviction Moderate (3/5)Side: Shipping is not earningSide named by the owner.
- v1Revenue yes, margin not yetConviction Moderate (3/5)First call, with the story.
13 voices
Funded, not forecast
The hyperscaler thesis is that AI makes businesses they already own more valuable, that owning the customer relationship makes them less dependent on predicting which model wins, and that broad agent usage creates recurring compute demand, while…
What matters for hyperscaler investment is the relationship between earnings growth and the capital required to produce it rather than when capex peaks; Meta in 2023 illustrates the cash-flow math, with operating cash flow going from $50.5B to…
The binding constraint on NVIDIA is supply rather than demand: unconstrained revenue growth was described as potentially above 100% YoY while guidance was near 70% growth because of supply limits, and the CFO framed the FY2028 70% growth outlook as…
Fabrinet's next step depends on turning new manufacturing roles in NPO, CPO, and scale-across into real volume; it will be necessary to confirm which steps Fabrinet handles in NPO, CPO, and ELS production programs, whether multi-rail programs show…
Muse is not important because of its underlying model; the real story is that Meta is assembling every component an AI needs to act on a user's behalf in the real economy: dedicated compute, a memory of the user, access to logins, a way to pay, and…
In the August 14, 2026 edition, the author argues: Demand for the theme is already funded rather than forecast: TSMC guides to 52-56 billion dollars of 2026 capex, chip equipment sales just set a record, and the silicon photonics market is modelled…
Unlike Amazon, Google and Microsoft, which had strong free cash flow, strong balance sheets and little or no debt pre-AI era letting them own their datacenters and equipment, Oracle entered the AI era already carrying dozens of billions in debt…
The majority of Oracle's 26 billion QoQ RPO growth in Q1 came via prepayment or bring-your-own-hardware arrangements, meaning the new contracts won't require incremental capital from Oracle and, according to Oracle, support its guidance that…
Group Up added capacity too slowly, which will limit growth for the next two years, and with the production ramp slipping to the end of 2028 plus the largest company-wide salary increase in its history amid a Taiwanese engineer shortage, there may…
Among the three memory names Daniel Romero holds, only Samsung is rated a Buy, because of foundry, advanced packaging, new memory architectures, unique vertical integration for base dies versus SK hynix, and because it is the cheapest of the three…
Tae Kim believes that if OpenAI were public, its stock would have gapped up 30% or more on hitting $70 billion in ARR, a blowout versus expectations on the level of Nvidia in May 2023, and that the figure proves his August report that OpenAI's ARR…
Despite Muse's initial traction, its long-term success is far from assured; Meta can cross-promote it across Facebook, Instagram, and WhatsApp, but competitors with similar distribution advantages are coming, with OpenAI likely being the biggest…
Consensus expects hyperscalers' operating cash flow to more than triple by 2030, outpacing the sharp increase in AI-driven capex, based on FactSet and Apollo Chief Economist sources for Google, Meta, Amazon, Microsoft and Oracle.
12 voices
A backlog is a promise
Amazon's backlog grew from $200 billion at a weighted average life of 3.8 years (September 30, 2025) to $496 billion at a weighted average life of 6.4 years (June 30, 2026), just nine months later, which Michael Burry identifies as the same type of…
Amazon's free cash flow has gone negative again, roughly negative $8 billion, while combined uncommenced leases and purchase commitments of $267 billion are significant and growing, up 81% over the last nine months, and long-term debt raised jumped…
SB's backlog isn't 100% guaranteed revenue, because most of its contracts have completion time clauses letting the customer exit the agreement or get a discount if a project is delayed by more than a year.
SB Energy, purportedly a data center company, has zero operational data centers, and while it has a $439 billion backlog, only 9% of contracts have actually broken ground.
There is a chain where NVIDIA sells chips to hyperscalers, hyperscalers' capital spending depends on the future health of OpenAI and Anthropic, and both are currently losing substantial sums, with OpenAI's revenue growth having slowed relative to…
George Noble's bear case for AI rests on the question of actual return on investment; he argues that Nvidia announcing massive funding packages just to keep customers like OpenAI able to keep buying chips shows the money underneath the boom is not…
Roughly 15% of revenue is tied directly to EU and state tenders where timing is unknown and awards are not always publicly announced, and a significant portion of private-sector spending is supported by EU or government funding incentives, so the…
Ilyda is currently assessed as a genuinely good business at a boring price, with issues around revenue visibility post-2027; Kevin says his process leans more on future predictability, so he needs a better estimate of the future than just the CEO's…
There is substance behind the growth confidence: management expects to finish FY2026 with at least $10B in backlog, says it has secured supply to support at least 30% revenue growth in FY2027, and believes it has secured supply to extend that…
Gaetano is much more bullish on Ciena over the past two weeks given upgraded guidance and the ECOC demos, and thinks Ciena is one company being under-appreciated for its late 2027/early 2028 time frame; CFO Marc Graff said the roughly $14B fiscal…
JunkBondInvestor leans toward a Panther Creek lease being signed, but expects it in 2027 rather than on management's 2026 timeline, because assembling a delivery schedule, construction budget and risk allocation that both a tenant will sign and a…
The line between AI borrowers who can keep paying up and borrowers who are squeezed follows cash flow and financing options more than sector: a borrower with a strategic counterparty (like Meta or Alphabet), a public stock, and lenders competing…
10 voices
Shipping is not earning
Edelbridge Alpha's model assumes $11.1 billion of FY28 revenue, materially above consensus expectations of roughly $9.6 billion, and notes reaching it would require OCS to scale quickly, high-power laser shipments to ramp with CPO and NPO…
At current levels Edelbridge Alpha does not find the risk/reward attractive, because reaching their target requires strong execution on manufacturing ramps, sustained pricing power, and continued demand growth, and earnings must translate into cash…
Management expects 40% to 50% of backlog to convert to revenue in Q3 alone, roughly $57M to $71M, and Multibagger Ideas views this conversion as making it achievable for the second half to deliver more than $12.7M of net income to common needed to…
Multibagger Ideas states the position earns its next tranche if Q3 results in December show backlog converting as guided and gross margin moving back toward 30%, or if Jordan becomes a definitive award; another receivable write-off or gross margin…
Management said substantially all of the backlog is expected to be completed within the next twelve months and expects approximately 40-50% of the $142.3 million backlog to convert to revenue in the third quarter, implying $57-71 million of Q3…
Schwar Capital Research will be watching whether Q3 revenue conversion clears the 40-50% backlog guidance ($57-71 million) and beats the $61.2 million Q3 fiscal 2025 record, whether book-to-bill stays above 1x, whether gross margin moves toward the…
Late capacity, margin squeeze until 2028.
Fabrinet: new roles must become profit.
NVIDIA has extended payment terms for investment-grade customers from 90 days to a full year, lengthening days sales outstanding by 14 days.
Nvidia's top five customers accounted for seventy percent of accounts receivable for the quarter, and some of those customers including OpenAI are not in especially healthy financial condition.
8 voices
Voices that cross sides
Tower already has $1.3B of contractual silicon photonics revenue commitments for 2027, backed by roughly $290M of customer prepayments, and management has said its actual 2027 SiPho expectations are substantially higher than that contractual…
GFS is adding SiGe capacity to an optical business where demand already exceeds supply, and GFS has said its SiGe capacity is oversubscribed through 2027.
Off-balance-sheet AI financing is back because costs are exploding: Nvidia raised prices 15%, data centers are expensive and slow to build, and the big cloud companies that used to be capital-light cash machines will spend something like $700…
An inflection point will arrive at which expectations will no longer be rewarded so heavily and proof of concept will be required in the results.
Record backlog, 1% growth at CeoTronics; PPIH's Q3 test.
Nvidia supply-bound; Fabrinet's roles still have to earn.
HBM capacity exploding at TOWA, added too slowly at Group Up.
Edelbridge Alpha rates Vertiv's execution risk 3.5/5, noting that while hyperscaler capex demand looks strong, the bigger question is whether Vertiv can scale manufacturing fast enough to keep up without sacrificing quality or margins, calling this…
Where they cross
Who answers whom.
The specific points where one voice meets another: the same evidence read two ways, or the same mechanism with a different sign.
Gaetano qualifies Kevin, on a condition
Gaetano qualifies Kevin, on the evidence
Tae Kim contradicts Michael Burry, on the mechanism
Mia Silverio contradicts Tae Kim, on the evidence
Tae Kim qualifies Ren, on a condition
Tae Kim qualifies Ren, on the mechanism
Tae Kim qualifies Steve Eisman, on a condition
Steve Eisman contradicts Tae Kim, on the mechanism
Tae Kim qualifies Steve Eisman, on the evidence
Tae Kim qualifies Steve Eisman, on the mechanism
Ren qualifies Gaetano, on the evidence
Michael Burry qualifies Gaetano, on the evidence
Steve Eisman contradicts Aurelion Research, on premise
Steve Eisman qualifies Edelbridge Alpha, on the evidence
Ren contradicts Steve Eisman, on the mechanism
Michael Burry contradicts Gaetano, on the mechanism
Gaetano contradicts Tae Kim, on the evidence
Torsten Slok contradicts Michael Burry, on premise
Gaetano qualifies Tae Kim, on the evidence
Michael Burry contradicts Gaetano, on the evidence
Ren contradicts Tae Kim, on the mechanism
Steve Eisman contradicts Gaetano, on the mechanism
Steve Eisman qualifies Mia Silverio, on the evidence
JunkBondInvestor qualifies Torsten Slok, on a condition
Alternative explanations
Three other readings.
What would have to be true for the Call to be wrong, as the other side argues it.
- The margin is not the number to watch: what matters is earnings growth against the capital required to produce it, and Meta in 2023 shows the cash-flow math working once the ramp passes (the case Gaetano makes).
- Supply, not demand, binds the quarter: Nvidia's unconstrained growth could run above 100 percent against guidance near 70 percent, so a shortfall is a supply story (the case Aurelion Research makes).
- The far book is less than a promise: Amazon's backlog went from $200 billion at a 3.8-year average life to $496 billion at 6.4 years in nine months, a stretching, not a filling (the case Michael Burry makes).
What settles it
The dated tests.
The tests the story set, each with its specification where the owner has written one and the piece of the argument it verifies; on the Docket, resolved on the Results page.
Microsoft reports: the share of backlog due within twelve months, 30% last quarter, and whether the near book's growth closes on the far book's 113%; at Alphabet, whether the TPU contracts inside cloud backlog get a size; at Meta, which reports October 21, whether the $10.8 billion of restricted cash stays restricted.
Part 11: The Backlog Is Not the Quarter Pendingp11-t01Pre-specifiedChecks the near-book claim
- Question
- Did Microsoft's quarterly report for the quarter ended September 30, 2026 put the share of its remaining performance obligations expected to be recognized as revenue within the next twelve months at 30% or less?
- Source
- Microsoft Form 10-Q for the quarter ended September 30, 2026, revenue note on remaining performance obligations (SEC EDGAR, CIK 0000789019)
- Rule
- Read the sentence in the 10-Q's revenue note that gives the percentage of revenue allocated to remaining performance obligations Microsoft expects to recognize over the next twelve months. The answer is yes if the share is 30% or less (the book keeps drifting outward); no if it is above 30% (the near book is catching up). The growth of the near and far portions against last year's 37% and 113% is DeepStack's reading into the Result, not the answer.
- If the source is late, revised or silent
- If the 10-Q is not public on EDGAR by November 15, 2026, or states no twelve-month percentage, the test is inconclusive, and the Result says what was read and what was missing.
The Call, before the dateExpects yes. Gives the yes answer 65%.
0 · no100 · yes
Nvidia reports: days sales outstanding after the move to a year of payment terms, whether five customers still carry 70% of receivables, and whether guidance near 70% moves toward the unconstrained figure as supply arrives.
Part 11: The Backlog Is Not the Quarter Pendingp11-t02Pre-specifiedChecks the payment-terms claim
- Question
- In NVIDIA's Form 10-Q for its fiscal quarter ended October 25, 2026, is accounts receivable, net at quarter end, divided by the quarter's revenue and multiplied by 91, at or above 51.1 days?
- Source
- SEC EDGAR, NVIDIA Form 10-Q for the quarter ended October 25, 2026 (balance sheet and statement of income), CIK 0001045810
- Rule
- DeepStack computes days sales outstanding from two figures the 10-Q prints: accounts receivable, net on the condensed consolidated balance sheet at October 25, 2026, and revenue on the condensed consolidated statement of income for the three months then ended, times the 91 days of the quarter, rounded to one decimal. The answer is yes if the figure is 51.1 days or more; no if it is below 51.1. The line is the average of the four quarters before the quarter ended July 26, 2026, computed the same way: 54.1 (July 27, 2025), 53.3 (October 26, 2025), 51.4 (January 25, 2026) and 45.4 (April 26, 2026); the quarter ended July 26, 2026 gave 59.6. The payment terms the filing describes are read alongside and do not decide the test.
- If the source is late, revised or silent
- If the 10-Q is not public on EDGAR by December 15, 2026, the test is inconclusive, and the Result says what was read and what was missing; the next reading is December 31, 2026. If the period the 10-Q covers is not 91 days, the rule uses the number of days its period dates give. Figures are read as first filed; a later amendment does not reopen the test.
The Call, before the dateExpects yes. Gives the yes answer 75%.
0 · no100 · yes
Oracle reports: whether new remaining performance obligations keep arriving prepaid or with the customer's own hardware, the condition for a capex peak in fiscal 2027 and 2028.
Part 11: The Backlog Is Not the Quarter Pendingp11-t03Open questionChecks the far-book claim
Specification pending
Perma-Pipe's third quarter: 40% to 50% of the $142.3 million backlog converting, $57 million to $71 million, with gross margin moving toward 30% and no second receivable write-off.
Part 11: The Backlog Is Not the Quarter Pendingp11-t04 · monthOpen questionChecks the ramp-margin claim
Specification pending
- The optics ramp
Ciena's $10 billion backlog into shipments, Lumentum's path to $11.1 billion of fiscal 2028 revenue and Coherent's statement that it sells everything it can make, each measured against the margin the company targets.
Part 11: The Backlog Is Not the Quarter Pendingp11-t05Open questionChecks the optics-ramp claim
Specification pending
What changed
The record of Part 11.
The Part’s publication, every change to the Call, every result that came in and the next dated test, oldest first.
- PublishedPart 11 published; the Call: Revenue yes, margin not yet.
- Call updatedThe Call: Revenue yes, margin not yet. The orders will ship. The profit comes later, and a margin that lags the ramp will be reported as a demand miss. Our Call takes the side “Shipping is not earning”. by Marco Cauduro
- Next testNext: Oct 28, 2026, Microsoft reports: the share of backlog due within twelve months, 30% last quarter, and whether the near book's growth closes on the far book's 113%; at Alphabet, whether the TPU contracts inside cloud backlog get a size; at Meta, which reports October 21, whether the $10.8 billion of restricted cash stays restricted. Pre-specified
Sources (9)
- Apollo - hyperscaler credit rests on one consensus assumption
- Aurelion Research - in conversation with Nvidia
- Schwar Capital Research - Perma-Pipe's record backlog
- Prof G Markets - SB Energy's $439 billion backlog
- JunkBondInvestor - Keel Infrastructure's lease
- JunkBondInvestor - Credit Weekly, who the hike hits
- 100 Bagger Hunting - Ilyda's visibility past 2027
- Daniel Romero - portfolio update, Group Up's ramp
- Aurelion Research - Ciena and Fabrinet, the road ahead
Market data are as of the dates cited. Voices are paraphrased from public writing and attributed by name.
Analysis and opinion, not investment advice. Voices are paraphrased from public writing and attributed by name; DeepStack shows no score, ranking or accuracy for any person. Every question, one page each.