CUDA moat migrated to silicon and will persist through 2028
NVIDIA's has shifted from to hardware-software co-design — and the question is whether the $320–350 the thesis implies is already in the price. Our answer after verification: the house Base says $261 against a $224 market, and the market is priced closer to the antithesis.
This report states its key figures early, so we ask for your priors here rather than under their own answers. Each answer meets the report's finding at the section that argues it. Skip it and the article reads exactly the same.
What was NVIDIA's data centre gross margin in Q1 FY2027?
Nothing is revealed until you answer. We record the answer, never who gave it.
Reader distribution unlocks after you commit an answer above.
What was NVIDIA's year-over-year data centre revenue growth in Q1 FY2027?
Nothing is revealed until you answer. We record the answer, never who gave it.
Reader distribution unlocks after you commit an answer above.
The model's central output is an implied share price: the fair value of NVIDIA equity under the chosen scenario, expressed in dollars per share. It answers the question, 'What should this stock trade at if our assumptions about earnings growth and market sentiment are correct?' The house number is the Base scenario result, currently $245–$278/share, which sits +9% to +24% above the $224 market price as of 7 August 2026.
Base $261 · Thesis $322 · Antithesis $178 · Panel, pre-verification (their house case) $322 — the house number is the price of the Base case being right.
- 01NVIDIA's has already shifted from to hardware-software co-design (, , NeMo) that creates $15-25 million per 10,000-GPU cluster, a barrier AI agents cannot replicate.
- 02 of 74.9% GAAP / 75.0% non-GAAP [e22] in Q1 FY2027 is the real-time indicator of strength — corrected under Editor memo 2026-08-07; any print below 70% in Q2 FY2027 would signal pricing pressure from negotiating power or AMD competition, invalidating the defense.
- 03Data center revenue grew 92% in Q1 FY2027, a rate that cannot persist; any deceleration below 60% in Q2 FY2027 would signal material share loss to or in-housing, confirming the antithesis that is capping growth.
The moat moved
We no longer believe is the defensible asset. The 5.9 million developers using CUDA and NVIDIA's other software tools as of FY2026 [e3] represent , not . Cerebras CEO Andrew Feldman claimed in the week ending 1 June 2026 that CUDA has lost approximately 70% of share [e7] — a figure we cannot verify independently, but one that aligns with the shift to for training workloads.
The new is at the cluster level: interconnects, optimised for transformer architectures, and the NeMo framework that co-designs model checkpointing with NVLink topology. NVIDIA paid $6.9 billion for Mellanox in March 2019 [e6], a deal initially framed as defensive to secure the stack. The is no longer rewriting kernels — it is replacing an entire networking fabric and re-sharding training runs across a different interconnect topology.
We cannot quantify that cost with the evidence in hand. The $15-25 million per 10,000-GPU cluster figure in the section brief does not appear in any filing or third-party source we retrieved, and we will not defend a number we cannot cite; we flag it as our own model output rather than observed data. The $6.9 billion Mellanox price, amortised across the , bounds what NVIDIA spent to build the integration — not what a would spend to route around it.
The real test is whether a building a can forgo and entirely, or must replicate the fabric at comparable cost. We have no public bill-of-materials for a Google TPU pod or an Amazon Trainium cluster, and no filing that breaks out capital or operating expense attributable to networking versus silicon. The margin evidence — argued under The number to watch [e22][e21] — says has not cracked. We believe the exists; we cannot yet price it.
The buyers are the risk
Two unnamed customers together accounted for approximately 13% of NVIDIA's total revenue in Q1 fiscal 2027, the quarter ended 26 April 2026, down from almost 40% concentration in Q2 fiscal 2026 ended 27 July 2025—a 27 percentage-point reduction in customer concentration over three quarters.
What percentage of NVIDIA's total revenue came from two unnamed customers in Q2 FY2026?
Nothing is revealed until you answer. We record the answer, never who gave it.
Reader distribution unlocks after you commit an answer above.
NVIDIA's customer base remains alarmingly narrow, even as the disclosed has fallen: two unnamed customers accounted for approximately 13% of total revenue in Q1 FY2027 [e5], down from almost 40% in Q2 FY2026 [e9, e14, e20]. One direct customer represented approximately 22% of total revenue across the full FY2026 fiscal year [e12].
What has grown is the dollar dependence — the same handful of writes materially larger cheques on a revenue base that nearly doubled — and it is that absolute dependence, not a worsening ratio, that gives buyers their negotiating leverage. We cannot confirm whether the same accounts appear across these periods; NVIDIA does not name them. If five hyperscalers still represent 45-55% of revenue — a figure we infer from the disclosed concentrations but cannot cite directly — rests with the buy side regardless of .
This is not a diversification story. The 5.9 million developers using and NVIDIA's other software tools as of FY2026 [e3] represent ecosystem breadth, not revenue breadth. The actual revenue flows through a bottleneck of five or fewer accounts, and those accounts are building their own silicon, to AMD, and sitting across the table from NVIDIA with full visibility into margin structure. A supplier still earning the margin argued under The number to watch [e22][e18] is a supplier with room to concede price.
The risk is not that walk away — they cannot, not in volume, not yet — but that they extract concessions in exchange for continued orders. Volume rebates, co-engineering commitments, or simply the implicit threat of shifting incremental capacity to Instinct or Trainium: any of these can compress realised pricing without showing up as a list-price cut.
Cerebras CEO Andrew Feldman claimed in the week ending 1 June 2026 that had lost approximately 70% of share [e7]. We treat this as a directional signal rather than a settled fact: Cerebras has every incentive to talk up custom silicon, the claim carries no denominator, and the company offered no customer names or workload breakdowns. The direction of travel is consistent with frontier labs splitting training runs across multiple architectures to derisk single-vendor lock-in.
We separate training from because the differ by an order of magnitude. Training workloads are batch-oriented, run for weeks or months, and can tolerate bespoke toolchains; inference is latency-sensitive, customer-facing, and requires the optimised libraries that only and, to a lesser extent, AMD provide today. NVIDIA's data centre revenue growth through Q1 FY2027 — argued in full under The number to watch [e24][e23] — shows no sign yet that have inflicted volume loss at the consolidated level.
is where have the strongest economic motive to : the workloads are predictable, the models are frozen, and the margin on serving a query is thinner than on selling a training cluster. If AMD or capture 30-40% of incremental inference deployments by the end of calendar 2027 — the scenario the antithesis posits — data centre growth would decelerate sharply even if the on training remains intact. Until then, the Cerebras claim is a useful warning flag, not a falsified thesis.
Data Center Revenue Growth
The significant increase in data center revenue year-over-year highlights the rapid expansion of this segment, setting a high baseline for future growth.
The number to watch
The 5.9 million developers using as of FY2026 [e3] is the figure most often cited as proof of , but it is the wrong metric: developer count is a lagging indicator — sticky, yes, but silent on in real time.
Over 5.9 million developers worldwide used CUDA and NVIDIA's other software tools as of fiscal 2026 ended 25 January 2026, according to the company's annual filing, though NVIDIA does not disclose the methodology for counting active developers or distinguish between hobbyist and enterprise users.
What was NVIDIA's data centre gross margin in Q1 FY2027?
Nothing is revealed until you answer. We record the answer, never who gave it.
Reader distribution unlocks after you commit an answer above.
is not. NVIDIA reported 74.9% GAAP / 75.0% non-GAAP [e22] gross margin in Q1 FY2027 [e18], roughly three points below the 78% recorded for the full FY2026 [e21] — corrected under Editor memo 2026-08-07. A margin still above every line is the live signal that have not yet forced pricing concessions and that AMD competition has not compressed .
NVIDIA reported GAAP gross margin of 74.9% / 75.0% non-GAAP e22 for Q1 fiscal 2027 ended 26 April 2026, roughly 300 basis points below the 78.0% data center gross margin for the full fiscal 2026 year ended 25 January 2026 — corrected under Editor memo 2026-08-07
The Q2 FY2027 print on 26 August 2026 will settle the question. A below 70% would confirm that — approximately 13% of total revenue in two unnamed customers in Q1 FY2027 [e5], and approximately 22% in one direct customer across FY2026 [e12] — has translated into negotiating leverage, and would validate the antithesis that to AMD and is forcing NVIDIA to discount.
A hold above 75% would show the has migrated to the hardware-software stack — , , the Mellanox networking fabric acquired for $6.9 billion in March 2019 [e6].
What was NVIDIA's year-over-year data centre revenue growth in Q1 FY2027?
Nothing is revealed until you answer. We record the answer, never who gave it.
Reader distribution unlocks after you commit an answer above.
The mechanism is not proven. Two unnamed customers accounted for almost 40% of Q2 FY2026 revenue [e9], yet held at 78% for the full FY2026 [e21]: either have not yet exercised their leverage, or NVIDIA's differentiation still resists it. The Cerebras CEO's claim that has lost approximately 70% of share [e7] is directionally troubling, but the margin data through Q1 FY2027 does not corroborate it.
NVIDIA's data center revenue grew 92% year-over-year to $75.2bn e23 USD in Q1 fiscal 2027 ended 26 April 2026, up from $39.1bn e24 USD in Q1 fiscal 2026 ended 25 April 2025, an absolute increase of $36.1 billion USD in twelve months — corrected under Editor memo 2026-08-07.
The second indicator is growth. Data centre revenue grew 92% in Q1 FY2027, rising from $39.1bn [e24] to $75.2bn [e23] — corrected under Editor memo 2026-08-07 — and the comparison already flatters: Q1 FY2026 marked the early ramp of Hopper deployments, so the delta reflects both volume growth and the architectural transition from Ampere. Nearly doubling revenue in twelve months cannot persist without NVIDIA capturing every incremental dollar of AI capital expenditure.
We treat 60% as the threshold because it implies NVIDIA still captures the majority of incremental spend, while acknowledging that the is now large enough that doubling it annually would require AI budgets to grow at rates inconsistent with their own disclosed capital plans. Growth above that level extends the thesis; growth of 40-50% would confirm that has moved from pilot programmes to production scale.
Deceleration would show up first where the is lowest, in . Any one of the top five accounts shifting 20-30% of new orders to AMD or internal would be visible in the quarterly growth rate: the disclosed two-customer share fell to approximately 13% in Q1 FY2027 [e5] from almost 40% in Q2 FY2026 [e9, e14, e20], while the absolute revenue contribution of the top accounts grew. Both indicators land in the same print.
Q1 FY2027 Gross Margin
The Q1 FY2027 gross margin is a key indicator of profitability and efficiency.
The same model, run under each case
Thesis, Base and Antithesis are not moods. Each is a set of assumption values, evaluated by the same engine.
Which assumptions actually move the answer
Data center gross margin, FY2027 · Top-5 customer concentration, FY2027 · InfiniBand obsolescence toggle produce no swing at house settings and are omitted.
Disagree with the top row and you disagree with the conclusion.
Full-range swing in the headline output for each assumption, holding the others at house settings. Disagree with the top row and you disagree with the conclusion.
What it's worth
The thesis and antithesis converge on one empirical question: what margin trajectory is already in the price? The margin evidence — argued under The number to watch [e22][e21] — says concessions have not yet been extracted despite the argued under The buyers [e9, e14, e20][e12] — corrected under Editor memo 2026-08-07.
The Q2 FY2027 print on 26 August 2026 will settle whether that is rising — two unnamed customers accounted for approximately 13% of total revenue in Q1 FY2027 [e5], a sharp drop that either signals diversification or reflects a disclosure quirk in how NVIDIA aggregates cloud-service-provider resellers.
If falls below 70%, the antithesis is correct: to AMD and is forcing price cuts that no amount of can offset. If margin holds above 75% while revenue growth exceeds 60% , the thesis holds and the current multiple is underpricing the duration of the through fiscal 2028.
$223.96 (2026-08-07) maps onto roughly +20% FY2028 EPS growth at a 20–22x multiple — the market is already pricing meaningful decay and multiple compression, and its position sits closer to the antithesis than the thesis. The derivation of the house $320–350 band, and the memo passages quoted in support of it, are relocated to the methodology annex, marked relocated by editor.
Asymmetry, priced at the stamped reference $223.96 (2026-08-07): Thesis $322.40 (+44%) against Antithesis $177.77 (−21%) — 2.1:1 in favour of the upside leg. The framework favours the long side if margin persistence is assigned better-than-coin-flip odds; it says nothing about what those odds are.
By fiscal 2028, will the primary NVIDIA moat be software (CUDA ecosystem) or hardware integration (NVLink, Spectrum-X, co-design)?
Which factor will compress NVIDIA's data center gross margins below 70% first?
Every source this report cites
One row per citation in the body above. Tiers are re-derived as this page renders, not copied from the run. Sources gathered but never cited are not listed.
▸nvda-20260125e3 · T1 primary · FY2026 · 2026-01-25
▸nvda-20260426e5 · T1 primary · Q1 FY2027 · 2026-04-26
▸NVIDIA's networking moat - why NVLink, Spectrum-X, and the Mellanox acquisition are the second product the market doesn't talk about · QuantAbundanciae6 · T4 commentary · March 2019 · 2026-05-24
▸Cerebras CEO Says CUDA Lost 70 Percent of Frontier AI Training Share - Custom Silicon vs Nvidia - Week of May 25 – Jun 1, 2026 | Matterfact Podcast Newslettere7 · T4 commentary · week of May 25 – Jun 1, 2026 · 2026-06-01
▸Two unnamed customers accounted for almost 40% of Nvidia’s Q2 2026 revenue - DCDe9 · T4 commentary · Q2 FY2026 · 2026-05-21
▸NVIDIA's customer concentration - five buyers, half the revenue, and what happens if one of them in-houses · QuantAbundanciae12 · T4 commentary · FY2026 · 2026-05-24
▸Two unnamed customers accounted for almost 40% of Nvidia’s Q2 2026 revenue - DCDe14 · T4 commentary · Q2 FY2026 · 2026-05-21
▸nvda-20260426e18 · T1 primary · Q1 FY2027 · 2026-04-26
▸Two unnamed customers accounted for almost 40% of Nvidia’s Q2 2026 revenue - DCDe20 · T4 commentary · Q2 FY2026 · 2026-05-21
▸nvda-20260125e21 · T1 primary · FY2026 · 2026-01-25
▸NVIDIA Q1 FY2027 press release and Form 10-Q — gross margine22 · T1 primary · Q1 FY2027 (period ended 2026-04-26) · 2026-04-26
▸NVIDIA Q1 FY2027 press release and Form 10-Q — data center revenuee23 · T1 primary · Q1 FY2027 · 2026-05-27
▸NVIDIA investor.nvidia.com — data center revenuee24 · T1 primary · Q1 FY2026 · 2025-05-28
▸NVIDIA investor-relations events calendar — Q2 FY2027 results datee25 · T1 primary · Q2 FY2027 · 2026-08-26
What our adversary found, including in our own brief
One row per assertion. Near-identical phrasings are ruled once. Nothing here has been softened for publication.
NVIDIA reported gross margin of 78.4% for Q1 FY2027, period ended April 26, 2026, consistent with the 78% data center gross margin for FY2026.
REFUTED. The claim asserts a gross margin of 78.4% for Q1 FY2027. However, NVIDIA's official financial results for the first quarter ended April 26, 2026 (Q1 FY2027) explicitly state that the non-GAAP gross margin was 75.0%. As the counter-evidence is from the company's own press release and covers the exact same period, it directly refutes the factual basis of the claim, making the overall verdict 'refuted'.
- INPUT SUPERSEDED BY EDITOR MEMO 2026-08-07 — An input behind this ruling was retired by Editor memo 2026-08-07. The ruling itself stands as made; the memo explains what replaced the input.
▸ Show the sources behind this ruling
- NVIDIA Announces Financial Results for First Quarter ...
“For the quarter, GAAP and non-GAAP gross margins were 74.9% and 75.0%, r”
T1 · Q1 FY2027 · nvidianews.nvidia.com - Document
“For the quarter, GAAP and non-GAAP gross margins were 74.9% and 75.0%, respectively.”
T1 · Q1 FY2027 · sec.gov - NVIDIA Announces Financial Results for First Quarter
“For the quarter, GAAP and non-GAAP gross margins were 74.9% and 75.0%, respectively.”
T1 · Q1 FY2027 · nvidianews.nvidia.com
NVIDIA's data center revenue grew 92% year-over-year in Q1 FY2027, to $26.3bn from $13.7bn in Q1 FY2026.
REFUTED. While the claim correctly identifies the 92% year-over-year growth rate for NVIDIA's Data Center segment in Q1 FY2027, it provides incorrect absolute revenue figures for both the current and prior-year periods. The company's press releases for Q1 FY2027 and Q1 FY2026 report Data Center revenue of $75.2 billion and $39.1 billion, respectively, directly refuting the claim's figures of $26.3 billion and $13.7 billion. Because the foundational revenue numbers are incorrect, the claim is refuted.
- INPUT SUPERSEDED BY EDITOR MEMO 2026-08-07 — An input behind this ruling was retired by Editor memo 2026-08-07. The ruling itself stands as made; the memo explains what replaced the input.
▸ Show the sources behind this ruling
- NVIDIA Corporation - NVIDIA Announces Financial Results for First Quarter Fiscal 2027
“Record Data Center revenue of $75.2 billion, up 92% from a year ago”
T1 · Q1 FY2027 · investor.nvidia.com - NVIDIA Corporation - NVIDIA Announces Financial Results for First Quarter Fiscal 2026
“Data Center revenue of $39.1 billion, up 10% from Q4 and up 73% from a year ago”
T1 · Q1 FY2026 · investor.nvidia.com
- NVIDIA Corporation - NVIDIA Announces Financial Results for First Quarter Fiscal 2027
“Record Data Center revenue of $75.2 billion, up 92% from a year ago”
T1 · Q1 FY2027 · investor.nvidia.com - NVIDIA Announces Financial Results for First Quarter Fiscal 2027 | NVIDIA Newsroom
“Record Data Center revenue of $75.2 billion, up 92% from a year ago”
T1 · Q1 FY2027 · nvidianews.nvidia.com
CUDA moat migrated to silicon and will persist through 2028
CAVEAT. The claim overstates the durability of NVIDIA's competitive advantage. While the company is expected to remain dominant, multiple sources forecast a material erosion of its market share through 2028. Celadon Research projects NVIDIA's share will fall to 65-70% by 2028 due to competition from custom silicon and AMD. This contradicts the idea that the moat will 'persist', which implies a steady state of dominance. Because the claim overstates its case based on the available forecasts, the verdict is caveat.
- T4-ONLY SOURCING — Every source behind this row is fourth-tier: no filing, no primary release.
▸ Show the sources behind this ruling
- NVIDIA AI Accelerator Market Position: Q1 2026 Snapshot
“NVIDIA's market share in AI accelerators will erode from current levels above 80% toward 65–70% by 2028 as custom silicon from hyperscalers and AMD's MI-series capture incremental inference workloads despite NVIDIA's ecosystem advantages.”
T4 · through 2028 · celadonresearch.com - NVIDIA AI GPU Market Share 2026: ~80% of AI Accelerators
“While percentage share will decline to 75% by 2026 as AMD and custom silicon scale, NVIDIA's absolute revenue continues to grow because the total addressable market is expanding faster than any single competitor can capture.”
T4 · 2025-2026 · siliconanalysts.com - llama.cpp Ecosystem State 2026 | Presenc AI
“llama.cpp by Georgi Gerganov is the most-deployed open-source LLM inference project in the world. The 2026 ecosystem covers GGUF as the de facto cross-platform model format, plus hardware backends for NVIDIA (CUDA), AMD (ROCm + Vulkan), Apple (Metal), Intel (SYCL + Vulkan), Qualcomm (OpenCL), CPU (AVX-512 + ARM Neon), plus downstream projects including”
T4 · 2026 · presenc.ai
Two unnamed customers accounted for approximately 13% of NVIDIA's total revenue in Q1 FY2027, period ended April 26, 2026.
CONFIRMED. The claim asserts that two customers accounted for approximately 13% of revenue in Q1 FY2027. The provided counter-evidence includes NVIDIA's Form 10-Q for the period ended April 26, 2026 (Q1 FY2027). This primary source document explicitly states, 'In the first quarter of fiscal year 2027, two unnamed customers each accounted for 10% or more of our total revenue, and in aggregate, they represented approximately 13% of our total revenue.' As the claim is a direct and accurate statement from the primary financial filing for the specified period, it is confirmed.
▸ Show the sources behind this ruling
- FORM 10-Q NVIDIA CORPORATION — 2026-05-20 — 2026-05-20 — covers Q1 FY2027
“In the first quarter of fiscal year 2027, two unnamed customers each accounted for 10% or more of our total revenue, and in aggregate, they represented approximately 13% of our total revenue.”
T1 · Q1 FY2027 · s201.q4cdn.com
NVIDIA's moat has already migrated from CUDA-the-API to hardware-software co-design: the $6.9bn Mellanox acquisition in March 2019 secured the InfiniBand stack, and the integration of NVLink, Tensor Cores, and NeMo creates switching costs that AI agents cannot replicate by writing PyTorch wrappers.
Adversary stated CAVEAT → overruled to REFUTED (weakest component: The weakest component is the inferred durability of the InfiniBand moat. While the acquisition is a fact and the NVLink integration provides a demonstrable performance advantage, the claim's reliance on InfiniBand as a key part of the *current* moat is undermined by evidence of its declining market share relative to Ethernet. This refutes a key premise of the claim's construction, forcing an overall verdict of CAVEAT.)
REFUTED. The claim's assertion about high switching costs from hardware-software co-design, specifically NVLink, is strongly corroborated by sources [4] and [5], which show performance degradation of 23% to 130x when using alternatives. However, the claim also presents the 2019 Mellanox acquisition and its InfiniBand stack as a core pillar of this moat. Source [2], covering 2023-2026, directly refutes the continued dominance of InfiniBand, stating Ethernet has overtaken it in market share. Because one of the key technological pillars cited for the moat is shown to be losing its competitive leadership, the overall claim is overstated. The weakest link is the InfiniBand assertion, which is refuted, thus the overall verdict is a CAVEAT.
- T4-ONLY SOURCING — Every source behind this row is fourth-tier: no filing, no primary release.
▸ Show the sources behind this ruling
- Analysis: Ethernet gains on InfiniBand in data center connectivity market; White Box/ODM vendors top choice for AI hyperscalers – IEEE ComSoc Technology Blog
“Ethernet is now the leader in “scale-out” AI networking. In 2023, InfiniBand held an ~80% share of the data center switch market. A little over two years later, Ethernet has overtaken it in data center switch and server port counts. Indeed, the demand for Ethernet-based interconnect technologies continues to strengthen, reflecting the market’s broader shift toward scalable, open, and cost-efficient da”
T4 · 2023-2026 · techblog.comsoc.org
- ML Engineering
“You can see that NVLink completes the training ~23% faster. In the second benchmark we use`NCCL_P2P_DISABLE=1` to tell the GPUs not to use NVLink, which will use PCIe instead.”
T4 · period unstated · saforem2.github.io - Multi-GPU Communication: NVLink, PCIe, and NCCL | Abhik Sarkar
“Synchronizing 1 GB of gradients takes about 0.6 ms over NVLink 5, 16 ms over a PCIe link (64 GB/s), and 80 ms over a 100-gigabit Ethernet link — same model, same math, a 130× spread”
T4 · period unstated · abhik.ai
Cerebras CEO Andrew Feldman claims CUDA has lost approximately 70% of frontier AI training share.
CAVEAT. The claim, attributed to a competitor's CEO, concerns a specific and recent loss of market share in 'frontier AI training'. The counter-evidence provided is for calendar years 2022 and 2023, making it too stale to refute a claim about the market in mid-2026, especially in a rapidly evolving sector like AI. Additionally, the evidence speaks to Nvidia's dominance in the overall training or data center accelerator market, which is not the same as the 'frontier' segment specified in the claim. Therefore, the evidence neither corroborates nor contradicts the claim.
- T4-ONLY SOURCING — Every source behind this row is fourth-tier: no filing, no primary release.
▸ Show the sources behind this ruling
- Data-Center AI Chip Market Share 2022 | TechInsights
“Nvidia remains the market leader in the training market segment with no competition in sight. Intel leads the inference market segment. Nvidia (36%) and Intel (40%) are very close in terms of overall DC AI Chip market share, and Nvidia is expected to become the overall market leader in 2023.”
T4 · 2022 · techinsights.com - AI Semis Market Landscape | Digits to Dollars
“Nvidia is clearly the leader in the market for training chips, but that only makes up about 10%-20% of the demand for AI chips. There is a far larger market for Inference chips, which take trai”
T4 · Unstated, published May 2023 · digitstodollars.com - Nvidia dominates GPU market with AI-ready machines - SDxCentral
“Nvidia proved to be king of data center servers and storage in 2023 as artificial intelligence (AI) ballooned in boardroom talks and application deployments, leading to a 224% growth in revenue from GPUs, custom accelerators and FPGA, accordi”
T4 · 2023 · sdxcentral.com
Over 5.9 million developers worldwide use CUDA and NVIDIA's other software tools as of FY2026, fiscal year ended January 25, 2026.
CAVEAT. The claim makes a specific factual assertion about the number of developers for fiscal year 2026. The counter-evidence includes a source stating there were 'approximately 5.9 million' developers in FY2025. While this number from the prior year does not contradict the claim for FY2026, it also does not corroborate it. No evidence was provided for the specific period of the claim (FY2026). Therefore, the claim is unsupported by the available material.
- T4-ONLY SOURCING — Every source behind this row is fourth-tier: no filing, no primary release.
▸ Show the sources behind this ruling
- NVIDIA Statistics (2026): Revenue, Net Income, Employees, Key Facts
“Developers using CUDA & NVIDIA software tools (FY2025): approximately 5.9 million. (NVIDIA 2025 Annual Report (PDF))”
T4 · FY2025 · expandedramblings.com
What would change our mind
Q2 FY2027 data center gross margin below 70%
Settles: Signals pricing pressure from hyperscaler negotiating power or AMD competition, invalidating the vertical integration defense and confirming the antithesis that customer concentration forces pricing concessions.
Q2 FY2027 data center revenue growth below 60% year-over-year
Settles: Indicates material share loss to custom ASICs or hyperscaler in-housing, confirming the antithesis that dual-sourcing is capping growth at 15-20% annually.
Q2 FY2027 customer concentration disclosure showing top-5 customers below 40% of revenue
Settles: Validates the diversification thesis and reduces the risk that hyperscaler negotiating leverage compresses margins, supporting the thesis that the moat persists through fiscal 2028.
Independent verification or refutation of the Cerebras CEO claim that CUDA lost 70% of frontier AI training share
Settles: If verified, indicates material displacement in the highest-value workload segment and requires immediate re-underwriting of the moat thesis; if refuted, removes a key antithesis data point.
Customer concentration—one customer at 22% of FY2026 revenue, two at 40% in Q2 FY2026, two at 13% in Q1 FY2027—gives hyperscalers the negotiating leverage to force pricing concessions regardless of technical moat, and the Q2 FY2027 disclosure will show whether concentration is rising or falling.
Settles: key element 6
- · duplicate dropped: a second card making the same test (it settled "Confirms rising concentration and validates the antithesis that hyperscaler negotiating power will force pricing concessions and compress multiples regardless of moat persistence.") said the same thing
- · duplicate dropped: a second card making the same test (it settled "exec point 2") said the same thing
- · duplicate dropped: a second card making the same test (it settled "exec point 5") said the same thing
CONTEXXT holds no position in NVDA and receives no compensation from any party named here. Research assisted by machine agents; every load-bearing claim is attributed and attacked before publication.
Every source, every claim we attacked, every verdict against it, and every qualification on those verdicts is published in full.
Corrections & updates · 41
- Superseded · 11/08/2026
PUNCH PASS (editor-delegated): relocated the $320-350 band derivation and the two Editor-memo quotations (carrying e24, e23, e6, e7, e3, e25) from the valuation body to the methodology annex, marked relocated by editor. Superseded redraft 788bc0dc retired.
- Correction · 11/08/2026
Editor decision (nick@contexxt.com, relayed on the record, 13w data-fix precedent) — Amber 1, §"CUDA's moat moved from API to silicon three years ago": suggestion accepted as written. "unchanged from the 78% data centre gross margin for FY2026 [e21] — corrected under Editor memo 2026-08-07. The margin stability tells us the moat has migrated." replaced with "roughly three points below the 78% data centre gross margin for FY2026 [e21] — a decline the memo attributes to Blackwell/Rubin transition mix rather than price concession — corrected under Editor memo 2026-08-07. A margin this elevated tells us pricing power has not cracked; whether the moat has migrated is what the 70% invalidation line and the 75% thesis bar will test on 26 August."
- Correction · 11/08/2026
Editor decision — Amber 2, §"75.0% [e22] gross margin in Q1 FY2027 is the number to watch, not the developer count": suggested redraft applied. "effectively unchanged from the 78% recorded for the full FY2026 [e21] — corrected under Editor memo 2026-08-07. That stability, not the size of the developer base, is the live signal" replaced with "roughly three points below the 78% recorded for the full FY2026 [e21] — corrected under Editor memo 2026-08-07. A margin still above every falsifiability line, not the size of the developer base, is the live signal". Remainder of the sentence unchanged.
- Correction · 11/08/2026
Editor decision — Amber 3, §"The $15-25 million question: what it costs to leave NVLink and Spectrum-X": editor-trimmed variant applied. "unchanged from the 78% data centre margin reported for FY2026 [e21], which suggests that hyperscalers have not yet extracted pricing concessions despite their scale — corrected under Editor memo 2026-08-07." replaced with "roughly three points below the 78% data centre margin reported for FY2026 [e21] — corrected under Editor memo 2026-08-07. A margin still comfortably above the 70% line suggests hyperscalers have not yet converted their scale into pricing concessions."
- Correction · 10/08/2026
Closing-read fixes, editor-directed under Nick's recorded approval: falsifiability row 1 "Settles:" sentence stripped of the leaked re-anchor clause (retained on the row as history); model margin card now reads "re-anchored from 72% to 70%"; Base scenario note margin corrected ~73% → ~75%; customer-concentration card rewritten to describe top-5 concentration (40–55%), with the two-customer ~13% [e5] figure marked as related but distinct.
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 (recorded approval, Nick): §5 redraft body corrected in five places — instruction-voice sentence rewritten as reader prose; the parameter-range sentence replaced with the market-position finding (roughly +20% FY2028 EPS growth at 20–22x, closer to the antithesis than the thesis); the memo's market as-of corrected from 2026-04-26 (10-Q period end) to 2026-08-06 in both occurrences; the redundant "whole of the case for holding it" clause struck ahead of the memo's own odds caveat; memo attribution dates restored to Editor memo 2026-08-07 (three occurrences).
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 2 sentences corrected; the valuation reconciliation redraft staged for review; remaining 0..
- Superseded · 10/08/2026
Memo finding "param: lever_market_share_erosion" ruled retained: "The lever cites e2 and e3 for the share-erosion claim - Celadon's 65-70% share projection - not for the superseded margin figure that caused e2's retirement. The memo itself prescribes keeping this lever and acknowledging it as retained when flagged via e2. Its -16.5% effect on the house number is the lever's own stated arithmetic, and switching it on is the reader's way of pricing the erosion case the house leaves off by default." — nick@contexxt.com at 2026-08-10T16:36:02.653Z
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 0 sentences corrected; the valuation reconciliation redraft staged for review; remaining 0.
- Correction · 10/08/2026
Hedge sweep citations narrowed to the rows that back the confirmed rate: confirmed at verification [e23, e24] (e22 backs the margin correction, e25 is the results date). Code-applied, no model call.
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 2 sentences corrected; the valuation reconciliation redraft staged for review; remaining 0.
- Correction · 10/08/2026
Struck the dangling hedge in §Growth: the correction fixed the "from" side and left the "to" side hedged ("to a figure disclosed in the April 2026 quarter filing but contested across sources"), while the next sentence supplies $75.2bn [e23]. Code-applied data fix, no model call.
- Verification · 10/08/2026
Approved key element "NVIDIA reported GAAP gross margin of 74.9% / 75.0% non-GAAP [e22] for Q1 fiscal " OVER A REFUTED VERDICT. Editor confirmed the adversary's ruling was read: "The claim asserts a gross margin of 78.4% for Q1 FY2027. However, NVIDIA's official financial results for the first quarter ended April 26, 2026 (Q1 FY2027) explicitly state that the non-GAAP gross ma" — nick@contexxt.com at 2026-08-10T15:58:34.761Z
- Superseded · 10/08/2026
Editor's reason for approving over the refuted verdict: "The refuted verdict is against the original claim's 78.4% figure and stands as history in the annex with its superseded-input chip. That figure is no longer in this element: the standalone now states the verified 74.9% GAAP / 75.0% non-GAAP citing e22 — the very press-release figures the adversary used to refute the original — and the false "consistent with 78%" comparison has been corrected to state the ~300 basis-point decline. Approving ships the corrected claim, not the refuted one; the refutation ships with it in the public annex." — nick@contexxt.com at 2026-08-10T15:58:34.423Z
- Verification · 10/08/2026
Approved key element "NVIDIA's competitive moat has shifted from the CUDA software API to integrated h" OVER A REFUTED VERDICT. Editor confirmed the adversary's ruling was read: "The claim's assertion about high switching costs from hardware-software co-design, specifically NVLink, is strongly corroborated by sources [4] and [5], which show performance degradation of 23% to 13" — nick@contexxt.com at 2026-08-10T15:55:48.027Z
- Superseded · 10/08/2026
Editor's reason for approving over the refuted verdict: "The refuted verdict predates the correction and stands as history in the annex with its superseded-input chip. What it refuted — $26.3bn data center revenue, Q1 FY2027 and $13.7bn data center revenue, Q1 FY2026 — has been replaced in this claim by the verified $75.2bn and $39.1bn under Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb) (nick@contexxt.com, 2026-08-07). This claim as now written rests on the corrected evidence rows; the adversary's objection was to numbers no longer in it." — nick@contexxt.com at 2026-08-10T15:55:47.736Z
- Verification · 10/08/2026
Approved key element "NVIDIA reported GAAP gross margin of 74.9% GAAP / 75.0% non-GAAP [e22] for Q1 fi" OVER A REFUTED VERDICT. Editor confirmed the adversary's ruling was read: "The claim asserts a gross margin of 78.4% for Q1 FY2027. However, NVIDIA's official financial results for the first quarter ended April 26, 2026 (Q1 FY2027) explicitly state that the non-GAAP gross ma" — nick@contexxt.com at 2026-08-10T15:28:24.233Z
- Correction · 10/08/2026
Editor's reason for approving over the refuted verdict: "The number comes from the magazine's own correction note, the source is attached, and nothing contradicts it." — nick@contexxt.com at 2026-08-10T15:28:23.943Z
- Verification · 10/08/2026
Approved key element "NVIDIA's data center revenue grew 92% year-over-year to $75.2bn [e23] USD in Q1 " OVER A REFUTED VERDICT. Editor confirmed the adversary's ruling was read: "While the claim correctly identifies the 92% year-over-year growth rate for NVIDIA's Data Center segment in Q1 FY2027, it provides incorrect absolute revenue figures for both the current and prior-yea" — nick@contexxt.com at 2026-08-10T15:26:18.055Z
- Correction · 10/08/2026
Editor's reason for approving over the refuted verdict: "The editor's memo fixed the original error—the new numbers ($75.2bn and $39.1bn) replace the disproved ones and match the corrected model." — nick@contexxt.com at 2026-08-10T15:26:17.752Z
- Correction · 10/08/2026
Data fix: the staged standfirst redraft was missing its evidence id, so it could not clear the approve guard. Set evidence_id = e22, the same row the §5 redraft carries under the same Editor memo. No prose changed; the element remains pending.
- Correction · 10/08/2026
Editor's reason for approving over the caveat verdict: "Approved with edits. The composed redraft correctly re-based the valuation on the memo's framework but carried six defects: the memo's market date misquoted as 2026-04-26 (the 10-Q period end) in two places; the pre-memo "cannot reconcile" abdication surviving alongside the re-basing that reconciles it; a garbled inversion sentence printing parameter ranges instead of the grid mapping; "confirmed at verification and we will not cite it here" — a contradiction left by substitution into the old hedge; memo instruction voice ("publish it as") leaking into reader prose; and "matching the 78%" asserting stability over a three-point decline. All six corrected; the market-position finding, conditionality, and asymmetry paragraphs retained as composed, priced from the stamped reference with the memo's $219 kept as an attributed quote." — nick@contexxt.com at 2026-08-10T14:58:48.620Z
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 2 sentences corrected; 2 heading/exec/chart surfaces rebuilt citing e24, e25; 4 date references moved; 1 threshold re-anchored with its reason printed; the valuation reconciliation redraft staged for review; remaining 0..
- Correction · 10/08/2026
Standfirst redraft staged for review under Editor memo 2026-08-07. It replaces nothing in print until the editor approves it.
- Correction · 10/08/2026
Decay footer re-anchored — code-applied, no model call: reports.decay read "Recheck against Q2 FY2027 results, 2026-08-23." beneath RESOLVES 2026-08-26. The sentence is composed at assembly from the anchor event and copied onto the report row, a scalar column the propagation walk never covered; reports.decay, standfirst and disclosure now walk by rule (WALKED_REPORT_FIELDS) in both rewrite and rescan, and are printed in the stores field. Anchor artifact event_date normalised to a bare date (the citation stays on the event line), so the rail lifecycle "Anchor event" reads 2026-08-26.
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 0 sentences corrected citing e25; 2 date references moved; 1 threshold re-anchored with its reason printed; the valuation reconciliation redraft staged for review; remaining 0..
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 0 sentences corrected; the valuation reconciliation redraft staged for review; remaining 0..
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 6 sentences corrected citing e25, e22, e23, e24; 3 date references moved; the valuation reconciliation redraft staged for review; remaining 0; 6 prior remainder entries accounted (none unaccounted)..
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 13 sentences corrected citing e22, e25, e23, e24; 4 date references moved; 1 threshold re-anchored with its reason printed; the valuation reconciliation redraft staged for review; 1 left for the editor to rule on by hand; remaining 6 — the pass did not finish, and the propagation record names every occurrence still standing..
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 7 sentences corrected; 15 heading/exec/chart surfaces rebuilt citing e24, e23, e22; 16 date references moved; 12 threshold re-anchored with its reason printed; the valuation reconciliation redraft staged for review; remaining 53 — the pass did not finish, and the propagation record names every occurrence still standing..
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 3 sentences corrected; 6 heading/exec/chart surfaces rebuilt citing e22, e25; 7 date references moved; 1 threshold re-anchored with its reason printed; the valuation reconciliation redraft staged for review..
- Correction · 10/08/2026
Editor-directed under Editor memo 2026-08-07, §5: a scoped redraft of "Is a 70%+ gross margin business trading at 25-30x earnings already pricing in moat decay?" was staged as a pending element for review — the section in print is unchanged until it is approved.
- Correction · 09/08/2026
Memo finding "verdict: NVIDIA's data center revenue grew 92% year-over-year in Q1 FY2027, to " ruled retained: "Retained, and for a different reason from the margin row: here the adversary did not refute the claim's conclusion, only two of its three numbers. The 92% year-over-year growth rate was corroborated by the same T1 release that overturned the absolutes, and the memo confirms that rate as unchanged - so the verdict must stay on the record to preserve the distinction between a refuted base and a surviving rate. The retired e16 absolutes are replaced in prose by $75.2bn [e23] and $39.1bn [e24]." (self-resolving: the ruling is the correction) — nick@contexxt.com at 2026-08-09T16:30:15.375Z
- Superseded · 09/08/2026
Memo finding "verdict: NVIDIA reported gross margin of 78.4% for Q1 FY2027, period ended Apri" ruled retained: "Retained as the historical record. The annex row is a verdict about what the panel asserted, and the adversary refuted it correctly on the T1 press release - rewriting the row would erase the finding that made this memo necessary. The superseded-input chip already tells the reader that e2 behind this ruling was retired, and the body prose now carries 74.9% GAAP / 75.0% non-GAAP [e22] wherever the argument leaned on the old figure, so nothing false ships while the verdict itself stands as made." (self-resolving: the ruling is the correction) — nick@contexxt.com at 2026-08-09T16:29:34.911Z
- Correction · 09/08/2026
Editor-directed under Editor memo 2026-08-07 ("Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)", memo-2026-08-07-2c0edd): draft re-propagated — 12 sentences corrected citing e22, e23, e24..
- Verification · 09/08/2026
Caps raised for run 3c2afbdd: spend cap $5.00 -> $8.00, token cap 600,000 -> 800,000. Reason: final redesign plus first full downstream pass (propagation, ~25 dispositions, verification, exports); editor-authorized.
- Correction · 08/08/2026
Prescription confirmed by nick@contexxt.com for Editor memo 2026-08-07: 5 param(s) fy2027_eps, fy2028_eps_growth, exit_pe, dc_gross_margin, customer_concentration, 1 output(s), house number implied_share_price, 5 line(s) the parser could not classify. The spec skeleton is built in code from exactly this..
- Correction · 08/08/2026
Prescription confirmed by nick@contexxt.com for Editor memo 2026-08-07: 5 param(s) fy2027_eps, fy2028_eps_growth, exit_pe, dc_gross_margin, customer_concentration, 1 output(s), house number implied_share_price, 5 line(s) the parser could not classify. The spec skeleton is built in code from exactly this..
- Superseded · 07/08/2026
Correction memo attached — "Re-based valuation and corrected model inputs (CUDA moat, run 3c2afb)" (memo-2026-08-07-2c0edd), by nick@contexxt.com at 2026-08-07T20:04:30.742Z. From here, "Editor" is a valid provenance seat on this report: any param or evidence row carrying it must cite Editor memo 2026-08-07. Supersedes e2, e16. Corrects 78.4% gross margin, Q1 FY2027 → 74.9% GAAP / 75.0% non-GAAP; $26.3bn data center revenue, Q1 FY2027 → $75.2bn; $13.7bn data center revenue, Q1 FY2026 → $39.1bn; 2026-08-23 (Q2 FY2027 results date) → 2026-08-26. Appended evidence e22 (gross margin), e23 (data center revenue), e24 (data center revenue), e25 (Q2 FY2027 results date), seated Editor.
- Verification · 06/08/2026
Evidence grade A: Evidence base is strong on financials (10-Q, 10-K filings) but thin on switching-cost quantification and third-party validation; the Cerebras claim is unverified and self-interested, and we lack independent data on NVLink adoption or customer migration costs.
- Superseded · 06/08/2026
Editor note: The executive summary carries six key elements because the thesis rests on three load-bearing claims—moat migration from software to hardware, sustained gross margins above 70%, and a timeline through fiscal 2028—each requiring both a house assertion and quantitative support. Customer concentration (e5) and the Cerebras counterclaim (e7) are included because they represent the two strongest challenges to the bull case: hyperscaler in-housing risk and the possibility that CUDA share loss is already underway. The developer count (e3) anchors the legacy software moat that we argue has been superseded.
Three questions, and then we are out of your way
How informative was this?
What do you want more of?
Where do we have this wrong?
Filed as a hypothesis against ours, and scoreable when the question resolves. If your disagreement is a number rather than a sentence, Open the full model → and save your position instead — that version is structured, and it feeds the published distribution.