[NVDA] NVIDIA: Q3 FY2027 Earnings Preview on Rubin, Memory Costs and Cash Flow

NVIDIA designs GPUs, CPUs, networking chips and rack-scale AI systems, and its data center business now supplies more than 90% of revenue[1]. In this NVIDIA Q3 FY2027 earnings preview, the company is scheduled to report on 2026-11-18 for Q3 FY2027 (three months ending October 25, 2026)[2]. In the latest disclosed period, the second quarter of fiscal 2027 (ended July 26, 2026), revenue was US$96.22 billion, up 106% year over year and 18% sequentially, with data center revenue of US$89.02 billion[1]; GAAP gross margin was 75.0% and operating income was US$63.73 billion[3], while non-GAAP diluted earnings per share were US$2.22[4]. NVIDIA guided third-quarter revenue to US$108.0 billion, plus or minus 2%, assuming no data center compute revenue from China, with GAAP and non-GAAP gross margins of 74.0%, plus or minus 50 basis points[5]; on the call, management added that margins would bottom at 71%-72% in the fourth quarter[6]. The Drillr earnings calendar, updated on October 21, 2026, shows consensus third-quarter revenue of US$108.60 billion and EPS of US$2.491; the same source had expected US$92.27 billion and US$2.09 for the second quarter, and actual results came in above both[2].

Three things matter most in this report. First, Vera Rubin began production shipments in the third quarter[7] while management says growth is currently supply-constrained[8], so whether revenue approaches the US$110.16 billion top of the guidance range, and whether ACIE keeps outgrowing Hyperscale, will show whether the architecture transition opened a delivery gap. Second, rising memory prices have already moved the margin path from the roughly mid-70% full-year level management described in the first quarter[9] to a fourth-quarter trough of 71%-72%[6], and supply commitments rose from US$119 billion to US$279 billion, primarily for memory[10]; the third-quarter margin and the fourth-quarter guide will show whether the cost shock is larger than management judged in August. Third, days sales outstanding rose from 45 to 60 in the second quarter and operating cash flow fell from US$50.3 billion in the first quarter to US$24.1 billion[11]; whether cash catches back up with profit in the third quarter will show how much growth depends on payment terms and guarantees NVIDIA extends to customers.

Company Background and Business Structure

NVIDIA has moved from a graphics-card company to what it now calls "a data center scale AI infrastructure company"[12]. Incorporated in California in 1993 and headquartered in Santa Clara, it began with the GPU and later extended the GPU from gaming graphics into general-purpose accelerated computing[12]. Its platform is more than chips: it includes the CUDA development platform, the CUDA-X acceleration libraries, AI models and industry application frameworks, so the software customers build on this stack ties them to NVIDIA hardware[13]. In fiscal 2026 (ended January 25, 2026), revenue was US$215.9 billion, up 65% from the prior year[14].

Almost all of NVIDIA's revenue now comes from the data center, and the new disclosure framework splits customers into large cloud providers and a broader group of AI buyers. By reportable segment, Compute & Networking generated US$88.30 billion in the second quarter, or 91.8% of total revenue, and Graphics generated US$7.92 billion[1]. Starting in the first quarter of fiscal 2027, the company began reporting revenue by two market platforms and recast nine prior quarters[15]: second-quarter data center revenue was US$89.02 billion, of which Hyperscale (public clouds and the largest consumer internet companies) contributed US$48.71 billion and ACIE (AI Clouds, Industrial and Enterprise, covering AI-native clouds, sovereign AI and AI factories built by industrial and enterprise customers) contributed US$40.31 billion; Edge Computing (PCs, gaming, workstations, robotics and automotive) contributed US$7.20 billion[1]. In the second quarter the company also moved one customer from ACIE to Hyperscale because its business model changed, and recast that customer's prior-period revenue[16].

NVIDIA does not manufacture its own chips, and both its supply chain and its customer base are highly concentrated. TSMC and Samsung produce its wafers, packaging uses CoWoS technology, NVIDIA buys memory directly from SK Hynix, Micron and Samsung, and contract manufacturers such as Hon Hai, Wistron and Fabrinet handle assembly and testing[17]. Products reach end users through ODMs, OEMs, system integrators and distributors; in the second quarter one direct customer accounted for 16% of total revenue, and in the first half three direct customers accounted for 16%, 15% and 13%[18]. The company also estimates that one AI research and deployment company contributed a meaningful amount of revenue by buying cloud services from NVIDIA's customers[19]. By customer headquarters, second-quarter revenue was US$60.07 billion from the United States, about 62%, US$26.99 billion from Taiwan and US$7.88 billion from China including Hong Kong[20].

Beyond its operating assets, NVIDIA carries a growing book of outside investments. Non-marketable securities stood at US$51.16 billion at the end of the second quarter, up from US$22.25 billion at the start of the year[21]; among its equity-method investments in infrastructure financiers, the ones deemed variable interest entities carried a maximum loss exposure of US$4.7 billion, and because NVIDIA is not the primary beneficiary it does not consolidate them[22]. In September 2026 the company agreed to acquire the open-model platform Hugging Face for about US$11.9 billion, with closing expected in the first half of 2027[23].

Financial History and Current Position

NVIDIA's annual revenue more than tripled over three fiscal years, and nearly all of the growth came from the data center. Revenue rose from US$60.92 billion in fiscal 2024 to US$130.50 billion in fiscal 2025 and US$215.94 billion in fiscal 2026; fiscal 2026 data center revenue was US$193.74 billion, made up of US$162.36 billion of compute and US$31.38 billion of networking[24]. Net income that year was US$120.07 billion, up 65%, and diluted earnings per share were US$4.90[14].

The fiscal 2026 margin decline and the year's cash uses trace two threads: a product transition and shareholder returns. Gross margin fell from 75.0% in fiscal 2025 to 71.1% as the business moved from Hopper HGX systems to Blackwell full-scale data center solutions and as the company took a US$4.5 billion charge for H20 excess inventory and purchase obligations in the first quarter[25]. Operating cash flow was US$102.72 billion[26]; capital expenditure was US$6.04 billion, and NVIDIA also paid US$13.0 billion to Groq and spent US$17.50 billion on non-marketable equity[27], while repurchasing 282 million shares for US$40.4 billion[28].

In fiscal 2027, revenue and profit kept accelerating and gross margin returned to about 75%. Revenue was US$81.62 billion in the first quarter and US$96.22 billion in the second; second-quarter GAAP gross margin was 75.0%, against 74.9% in the first quarter and 72.4% a year earlier; operating expenses were US$8.41 billion, up 10% sequentially and 55% year over year; operating income was US$63.73 billion and GAAP net income was US$59.69 billion[3]. GAAP net income includes gains and losses on equity securities, which contributed US$7.77 billion in the second quarter, so operating income is the better gauge of operating performance[29]; GAAP and non-GAAP diluted EPS were US$2.46 and US$2.22[4].

Cash flow fell well behind profit in the second quarter, while the balance sheet expanded receivables, inventory and debt at the same time. Second-quarter operating cash flow was US$24.08 billion and free cash flow was US$21.34 billion, compared with US$50.34 billion and US$48.55 billion in the first quarter, bringing first-half free cash flow to US$69.90 billion[30]. Cash, cash equivalents and marketable debt securities totaled US$56.6 billion at quarter end, and the company issued US$25.0 billion of senior unsecured notes during the quarter[11]; accounts receivable were US$63.06 billion and inventory US$31.58 billion, up from US$38.47 billion and US$21.40 billion at the start of the year, and long-term debt rose from US$7.47 billion to US$32.37 billion[21]. NVIDIA returned about US$26.0 billion to shareholders through buybacks and cash dividends in the second quarter, had about US$99.0 billion of repurchase authorization left at quarter end, and pays a quarterly dividend of US$0.25 per share[4].

Operating Model

NVIDIA's revenue is set by how many systems it can deliver, not by how many it can book. In equation form, revenue is roughly data center system deliveries times the value content per system, plus Edge Computing revenue; management says revenue growth is currently supply-constrained and that unconstrained demand would be significantly higher[8]. Data center made up 92.5% of second-quarter revenue and Edge Computing 7.5%[1], so the availability of wafers, CoWoS packaging and memory passes through to data center revenue within the same quarter, while supply commitments have to be locked in several quarters ahead.

Gross margin depends on product mix, memory prices and inventory provisions, and operating profit then subtracts operating expenses that are mostly research and development. Gross profit equals revenue minus wafers, packaging, memory and components, tariffs, and inventory and warranty provisions; margins are usually lower at the launch of a new architecture, and higher memory prices enter cost quarter by quarter as committed supply is delivered. NVIDIA guided third-quarter GAAP operating expenses to about US$9.2 billion[5] and said on the call that full-year operating expense growth would be in the low 50s percent[6]. Because GAAP net income includes fair-value changes on equity stakes in AI companies, operating income or non-GAAP net income is the right lens for operating performance.

Cash flow depends on how much profit is collected on time and where the company deploys cash. Operating cash flow is roughly net income, minus equity investment gains, minus increases in receivables and inventory, plus increases in payables and customer advances; free cash flow then subtracts capital expenditure and principal payments on equipment[30]. In the first quarter, management said it would return about 50% of fiscal 2027 free cash flow to shareholders[31], while it has invested nearly US$50 billion in frontier AI labs[32] and provides off-balance-sheet guarantees on some AI cloud partners' data center leases[33]. NVIDIA states in its 10-Q that financing arrangements with certain investment-grade customers, including extended payment terms under large multi-quarter agreements, will continue to affect the timing of its operating cash flows[34].

Industry and Competitive Position

NVIDIA competes against three kinds of rivals: independent chipmakers, large cloud providers that design their own chips, and suppliers of custom silicon and networking gear. Its annual report names AMD, Huawei and Intel, as well as Alibaba, Alphabet, Amazon, Baidu and Microsoft, which design AI hardware and software in-house[35]. That means several of NVIDIA's largest customers are also potential substitutes.

NVIDIA's advantage lies in its integrated hardware-and-software stack and its annual cadence of rack-scale products. One CUDA software base spans training, inference and other accelerated workloads[13], and networking has grown alongside the rack-scale systems: fiscal 2026 networking revenue was US$31.38 billion, up about 142% from US$12.99 billion the year before[24]. The company says the Vera Rubin platform is already running at partners including CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius[36], and management says Rubin delivers 30 times the throughput per megawatt and 35 times lower token cost than Grace Blackwell Ultra; these are the company's own claims[8].

NVIDIA's weaknesses sit at two ends, customer in-house chips and upstream supply, and the available material does not allow a precise share comparison. Management acknowledges that key customers such as OpenAI and Anthropic are developing custom chips, while arguing that those customers will still rely on NVIDIA's full-stack platform and networking[37]; wafers, advanced packaging and memory all depend on a handful of suppliers[17]. NVIDIA does not split revenue between Blackwell and Rubin, and no comparable competitor data on the same basis is available, so this section can describe the competitive structure but cannot quantify share shifts.

Core Debates

NVIDIA guided third-quarter revenue to US$108 billion, about 12% above the second quarter, just as Vera Rubin began production shipments and while the company says growth is supply-constrained. Can shipments hold their pace through the architecture transition, and will AI clouds, sovereign and enterprise customers (ACIE) keep leading the growth?

This question matters because data center supplies more than 90% of revenue, and the third quarter is the first quarter of the switch from Blackwell Ultra to Vera Rubin. Second-quarter revenue of US$96.22 billion came in above the US$91 billion, plus or minus 2%, range the company gave in the first quarter[9], and the market has grown used to that pattern; a transition quarter, however, is when delivery gaps are most likely. ACIE represents the spread of demand from a few cloud providers to AI clouds, sovereign projects and enterprises, and whether it keeps leading determines whether growth can stop depending on the capital spending of a handful of large customers.

The second-quarter figures show that the main engine of growth has shifted from Hyperscale to ACIE. Data center revenue was US$89.02 billion, up 18% sequentially and 117% year over year, against US$75.25 billion in the first quarter and US$41.10 billion a year earlier; Hyperscale was US$48.71 billion, up 13% sequentially and 102% year over year; ACIE was US$40.31 billion, up 25% sequentially and 138% year over year, or 45.3% of data center revenue, compared with 42.8% in the recast first quarter and 41.2% a year earlier[1]. NVIDIA attributes ACIE growth to AI natives, enterprises and sovereign customers, as well as hyperscalers using AI clouds[38]; inventory rose from US$25.8 billion at the end of the first quarter to US$31.6 billion, which the company says is in preparation for introducing Vera Rubin in the third quarter[11], and Hopper shipments to China were less than 1% of data center revenue[39].

The financial chain runs from supply to revenue: wafer, CoWoS and memory supply sets how many Blackwell Ultra and Vera Rubin systems ship, those deliveries split into Hyperscale and ACIE revenue, and the total becomes data center and company revenue, largely within the same quarter. An alternative reading is that part of ACIE's second-quarter growth came from NVIDIA's own financing, guarantees and investments in AI labs rather than from independent end demand; if ACIE growth slows sharply in the third quarter while new guarantees or commitments keep expanding, customer expansion is leaning more on NVIDIA's balance sheet. The reporting basis is also moving: the second-quarter reclassification recast first-quarter ACIE to US$32.20 billion[16], the new framework has only two quarters of formal disclosure, and the company does not split revenue between Blackwell and Rubin, so none of this can yet be separated in the available data.

In the third quarter, watch whether revenue exceeds the US$110.16 billion top of guidance, whether data center grows at least as fast as total revenue, whether ACIE keeps outgrowing Hyperscale and holds its 45.3% share, whether the company confirms that Rubin has contributed revenue, and whether fourth-quarter guidance still excludes China data center compute[5]. Revenue below the US$108.0 billion midpoint would mean supply or the Rubin transition met resistance outside the guidance; an ACIE share falling back below 42.8% would mean customer expansion is slowing and growth is concentrating again in a few cloud providers.

NVIDIA's gross margin was 75.0% in the second quarter, but rising memory prices led it to guide 74% for the third quarter, and on the call management said margins would bottom at 71%-72% in the fourth quarter; supply commitments more than doubled from US$119 billion to US$279 billion, mostly for memory. Can the third-quarter margin stay within guidance, and will the fourth-quarter guide come in below what was said in August?

This question matters because NVIDIA's high margins rest on pricing power over the whole system, and memory is one of the few rack components priced by outside suppliers and rising fast. On second-quarter revenue of US$96.22 billion, each percentage point of gross margin equals about US$0.96 billion of quarterly gross profit[3]. By locking in supply with very large commitments, NVIDIA secured shipments but also took memory price risk onto itself; whether Rubin pricing covers those costs will not be fully visible until fiscal 2028, but third- and fourth-quarter margins will show the direction first.

The evidence so far shows a steady second-quarter margin, with the pressure appearing mainly in guidance and commitments. Second-quarter GAAP gross margin was roughly flat with the first quarter, and the company said Blackwell still makes up the vast majority of revenue[29]; inventory provisions of US$985 million and releases of US$177 million left a net drag of 0.8 percentage points on gross margin in the quarter and 1.0 point in the first half[40], against 2.6 points for fiscal 2026 including the H20 charge[25]. Supply and capacity commitments rose from US$119 billion at the end of the first quarter to US$279 billion, of which US$92 billion falls in the rest of fiscal 2027, US$87 billion in fiscal 2028 and US$88 billion in fiscal 2029[10]. In the first quarter, management still described full-year gross margin as in the mid-70% range[9]; on the second-quarter call it said memory prices had risen faster than anticipated, that margins would bottom at 71%-72% in the fourth quarter, and that they would settle at 72%-73% in fiscal 2028 as price increases take effect[6].

The financial chain works like this: higher HBM and DRAM prices raise the unit cost of committed supply and inventory, that cost enters cost of revenue when products ship, and gross margin falls; if inventory and purchase commitments tied to the Rubin transition need write-downs, inventory provisions push margin down further. Because the commitments span three fiscal years, memory prices feed into cost with a lag of several quarters. An alternative reading is that the margin decline comes mainly from one-off Rubin launch factors such as low early yields and ramp costs rather than memory; if third-quarter inventory provisions and inventory both rise while the company's language on memory softens, the problem is more about the product transition. NVIDIA does not disclose memory unit prices, and the fourth-quarter range comes only from a call summary rather than the written CFO commentary, so the size of the cost shock cannot yet be checked directly.

In the third quarter, watch whether GAAP gross margin stays at or above 73.5%, whether fourth-quarter margin guidance is at or above 71%, whether total supply commitments keep rising, whether the net inventory provision drag exceeds 1.0 point, and whether the company discloses pricing actions on Rubin systems[5]. A third-quarter margin below 73.5% or a fourth-quarter guide below 71% would mean the memory cost shock exceeded what management judged in August; a provision drag above 1.0 point would mean the architecture transition left obsolete inventory.

In the second quarter NVIDIA extended payment terms to some large customers: days sales outstanding rose from 45 to 60 and operating cash flow fell from US$50.3 billion to US$24.1 billion, while the company now guarantees up to US$108.5 billion of data-center lease obligations for AI clouds and OpenAI. Will third-quarter revenue convert back into cash on time, or is growth increasingly relying on credit NVIDIA extends to its customers?

This question matters because NVIDIA's profit has historically turned into cash almost immediately, funding large buybacks and investment in its ecosystem. The extended terms, guarantees and AI-lab investments that appeared in the second quarter show that some customers are growing faster than their balance sheets. If these are one-off terms on a batch of large contracts, cash flow should recover in the third and fourth quarters; if they become normal, NVIDIA is using its own credit to finance demand, which would weaken revenue quality and its capacity to return cash to shareholders.

The second-quarter evidence shows slower collection and more credit support arriving together. Accounts receivable were US$63.1 billion with 60 days sales outstanding, which the company attributed to extended payment terms on large multi-quarter agreements with certain investment-grade customers[11]; five direct customers accounted for 22%, 14%, 13%, 11% and 10% of the receivable balance[41]. Operating cash flow of US$24.08 billion equaled only 45% of non-GAAP net income of US$53.95 billion[30], even as the company returned about US$26.0 billion to shareholders[4]. In August, NVIDIA provided residual value guarantees for SB Energy's roughly 4.25-gigawatt campus in Ohio leased to OpenAI, with its aggregate payment obligation capped at US$105 billion[42]; earlier lease guarantees for AI cloud partners carried a maximum gross exposure of US$3.5 billion, AI cloud agreement commitments totaled US$36 billion, and data center leases not yet commenced and intended for third parties totaled US$20 billion[33], alongside US$25 billion of equity investment commitments[10]. Management said NVIDIA has invested nearly US$50 billion in frontier AI labs[32] and is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to set up independent compute financing platforms intended to mobilize more than US$500 billion of third-party capital[36].

The financial chain runs from credit to cash: extended terms on large multi-quarter agreements, guarantees for AI clouds and leases, and investments in AI labs support the purchasing power of balance-sheet-constrained customers, so revenue is recognized before cash is collected, receivables rise, operating cash flow trails profit, and free cash flow and room for shareholder returns shrink. An alternative reading is that the extended terms apply only to investment-grade customers with low default risk, that the higher DSO reflects a one-off contract structure, and that customer advances rising from US$160 million at the start of the year to US$2.8 billion partly offset the receivable build[43]. NVIDIA itself acknowledges that commercial arrangements expose it to counterparty risk[44], but it does not disclose the receivable balance owed by customers on extended terms, so it is not yet possible to tell which reading is closer to reality.

In the third quarter, watch whether DSO stays at or below 60 days, whether operating cash flow returns to at least 86% of non-GAAP net income, whether guarantee exposure and AI cloud commitments keep growing, whether buybacks and dividends keep exceeding quarterly free cash flow, and whether any single direct customer's share of receivables climbs back above 20%[34]. DSO staying above 60 days would mean extended terms are spreading; further growth in guarantees and commitments would mean demand leans more on NVIDIA's credit support.

Risks and Falsifiers

Export controls have all but removed China from data center revenue, and policy changes could also reach demand in other regions. The U.S. government allows only small H200 shipments to specific China-based customers, the chips must pass inspection in the United States and bear a 25% tariff that NVIDIA cannot pass on, and Chinese authorities have also restricted purchases; NVIDIA took a charge of about US$0.4 billion on H200 inventory and purchase obligations in the first half, and second-quarter China shipments were less than 1% of data center revenue[39]. Third-quarter guidance assumes no China data center compute revenue[5], the H20 charge in the first quarter of fiscal 2026 reached US$4.5 billion[25], and management says current Hopper shipments to China dilute margins[37]. If the third-quarter report adds no new export-control charges and the guidance assumption is unchanged, this risk does little to the current understanding.

Customer-designed chips would hit Hyperscale and frontier-lab purchasing growth directly. Key customers such as OpenAI and Anthropic are developing custom chips[37], and large cloud providers are themselves competitors[35]; two direct customers together accounted for 36% of fiscal 2026 revenue[45], and second-quarter Hyperscale revenue was US$48.71 billion[1]. If third-quarter Hyperscale revenue keeps growing sequentially and the company does not mention orders shifting away from large customers, this risk did not materialize in the quarter.

Execution risk in the Vera Rubin transition falls directly on data center revenue. Rubin integrates many complex custom components, and NVIDIA says in its 10-Q that it will ship Blackwell and Rubin systems in parallel, that it currently faces certain supply constraints, and that production scale and system complexity have caused and could again cause production delays[7]; management says Rubin will be the fastest product ramp in company history[8]. The third-quarter guidance range is about US$4.3 billion wide[5], and second-quarter data center revenue was US$89.02 billion. If third-quarter revenue is at least US$110.16 billion and the company confirms Rubin contributed revenue, the transition risk is falsified.

Memory price risk falls on gross margin and is magnified by the size of NVIDIA's commitments. Management says memory prices have risen faster than anticipated[6], and US$279 billion of supply commitments lock in high-priced purchases; if demand or product timing changes, only a limited part of those commitments can be cancelled, which could lead to further provisions[10]. On second-quarter revenue, each point of gross margin is worth about US$0.96 billion of quarterly gross profit, and inventory provisions already cut second-quarter margin by 0.8 points[40]. If third-quarter gross margin is at least 74.0% and the fourth-quarter guide is at least 72%, the memory shock is milder than management judged in August.

Balance-sheet-constrained AI clouds and model companies are a shared source of risk to NVIDIA's cash flow and credit exposure. These customers are growing faster than their ability to raise financing, and NVIDIA supports their purchases through extended terms, guarantees and equity investments[11]; if their funding dries up, NVIDIA could face delayed collections, guarantee payments or investment losses[44]. The related exposures include US$63.06 billion of receivables and US$51.16 billion of non-marketable securities[21], guarantees with a maximum gross exposure of US$108.5 billion[42][33], and US$25 billion of equity investment commitments[10]. If third-quarter DSO is at or below 60 days, operating cash flow is at least 86% of non-GAAP net income, and no new guarantees or commitments are added, this risk has not grown.

What to Watch Next

• Rubin transition and customer mix: watch whether third-quarter revenue exceeds the US$110.16 billion top of guidance (second quarter: US$96.22 billion; guide: US$108.0 billion, plus or minus 2%). Revenue below the US$108.0 billion midpoint would signal supply or transition friction outside the guidance.
• Rubin transition and customer mix: watch whether ACIE keeps growing faster than Hyperscale from US$40.31 billion, or 45.3% of data center revenue (42.8% in the recast first quarter). A share below 42.8% would signal growth concentrating again in a few cloud providers.
• Rubin transition and customer mix: watch whether NVIDIA confirms Rubin revenue and mentions any delivery gap, and whether guidance still excludes China data center compute. Confirmed Rubin revenue with revenue above the top of the range would falsify the transition risk.
• Memory cost and gross margin: watch whether third-quarter GAAP gross margin stays at or above 73.5% (second quarter: 75.0%; guide: 74.0%, plus or minus 50 basis points). A lower print would mean the memory shock exceeded the August view.
• Memory cost and gross margin: watch whether the fourth-quarter margin guide is at or above 71% (call: trough of 71%-72%) and whether supply commitments keep rising from US$279 billion. A guide below 71%, or a provision drag above 1.0 point, would weaken the current understanding.
• Extended terms and cash conversion: watch whether DSO stays at or below 60 days (first quarter: 45). DSO above 60 would mean extended terms are spreading.
• Extended terms and cash conversion: watch whether operating cash flow returns to at least 86% of non-GAAP net income (second quarter: 45%, or US$24.08 billion). A recovery would support the one-off contract-structure reading.
• Extended terms and cash conversion: watch whether guarantees (maximum gross exposure US$108.5 billion) and AI cloud commitments (US$36 billion) keep growing. Further growth would mean demand leans more on NVIDIA's credit support.

Conclusion

NVIDIA's business is driven by how many data center systems it can deliver, and its finances are currently in a phase of very strong profit and weakening cash conversion. Second-quarter revenue was US$96.22 billion, up 106% year over year, with a 75.0% GAAP gross margin and US$63.73 billion of operating income[3], but operating cash flow was only US$24.08 billion, or 45% of non-GAAP net income[30]. The central unresolved relationship has three layers: whether Rubin-transition deliveries can support the US$108.0 billion guide, how far memory costs will push gross margin down, and how much of the growth is bought with payment terms, guarantees and investments that NVIDIA itself provides.

As of October 22, 2026, the independent commentary published after the second-quarter results did not include a single interpretation that was signed, could be checked against its original text, and offered a substantive judgment on the three debates above; Drillr's broker research search returned only reports from February to April 2026. No outside views are therefore summarized here, and company guidance or the calendar's aggregated figures are not treated as independent interpretation. The third-quarter expectations described here come mainly from the guidance and call remarks NVIDIA itself gave in August[5][6].

The combination that would materially strengthen the current understanding is third-quarter revenue above US$110.16 billion with confirmed Rubin revenue, an ACIE share holding at 45.3% or higher, gross margin of at least 73.5% with a fourth-quarter guide of at least 71%, and at the same time a falling DSO and operating cash flow back above 86% of non-GAAP net income. Conversely, revenue below the US$108.0 billion midpoint, an ACIE share falling back below 42.8%, a fourth-quarter margin guide below 71%, or DSO staying above 60 days while guarantees and commitments keep growing would show that at least one of supply transition, memory cost or customer credit is less favorable than management judged in August.

Sources

[1] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · revenue by market platform · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[2] Drillr earnings calendar (updated 2026-10-21) · NVDA 2026-11-18 Q3 FY27 estimates · 2026-10-21 · Drillr earnings calendar · https://gateway.drillr.ai/mcp/private

[3] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · Q2 FY27 GAAP summary · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[4] NVDA 8-K filed 2026-08-26 (Q2 FY27 press release) · headline results and capital return · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27pr.htm

[5] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · Q3 FY27 outlook · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[6] NVDA earnings call 2026-08-26 (Q2 FY27) · guidance: margins and opex · 2026-08-26 · earnings-call · https://gateway.drillr.ai/mcp/private

[7] NVDA 10-Q filed 2026-08-26 · Vera Rubin production shipments and supply constraints · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[8] NVDA earnings call 2026-08-26 (Q2 FY27) · supply constraints and demand · 2026-08-26 · earnings-call · https://gateway.drillr.ai/mcp/private

[9] NVDA earnings call 2026-05-20 (Q1 FY27) · guidance and capital return · 2026-05-20 · earnings-call · https://gateway.drillr.ai/mcp/private

[10] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · supply and capacity commitments · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[11] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · receivables, inventory and operating cash flow · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[12] NVDA 10-K filed 2026-02-25 · business overview · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[13] NVDA 10-K filed 2026-02-25 · full-stack platform and CUDA · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[14] NVDA 10-K filed 2026-02-25 · FY2026 revenue growth · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[15] NVDA earnings call 2026-05-20 (Q1 FY27) · new reporting framework · 2026-05-20 · earnings-call · https://gateway.drillr.ai/mcp/private

[16] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · ACIE to Hyperscale reclassification · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[17] NVDA 10-K filed 2026-02-25 · fabless manufacturing and memory suppliers · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[18] NVDA 10-Q filed 2026-08-26 · Q2 FY27 customer concentration · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[19] NVDA 10-Q filed 2026-08-26 · indirect customers · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[20] NVDA 10-Q filed 2026-08-26 · revenue by customer headquarters · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[21] NVDA 10-Q filed 2026-08-26 · balance sheet · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[22] NVDA 10-Q filed 2026-08-26 · infrastructure financier investments · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[23] NVDA 8-K filed 2026-09-03 (Hugging Face acquisition agreement) · 2026-09-03 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000078/nvda-20260902.htm

[24] NVDA 10-K filed 2026-02-25 · FY2026 revenue by end market · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[25] NVDA 10-K filed 2026-02-25 · FY2026 gross margin and inventory provisions · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[26] NVDA 10-K filed 2026-02-25 · FY2026 cash flow and liquidity · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[27] NVDA 10-K filed 2026-02-25 · FY2026 capital expenditures · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[28] NVDA 10-K filed 2026-02-25 · FY2026 share repurchases · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[29] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · gross margin and operating expenses · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[30] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · free cash flow reconciliation · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[31] NVDA earnings call 2026-05-20 (Q1 FY27) · capital allocation update · 2026-05-20 · earnings-call · https://gateway.drillr.ai/mcp/private

[32] NVDA earnings call 2026-08-26 (Q2 FY27) · frontier AI lab investments and third-party financing · 2026-08-26 · earnings-call · https://gateway.drillr.ai/mcp/private

[33] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · AI cloud agreements and guarantees · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[34] NVDA 10-Q filed 2026-08-26 · financing arrangements and cash flow timing · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[35] NVDA 10-K filed 2026-02-25 · competition · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[36] NVDA 8-K filed 2026-08-26 (Q2 FY27 press release) · Vera Rubin production and compute financing platforms · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27pr.htm

[37] NVDA earnings call 2026-08-26 (Q2 FY27) · risks: memory, China, custom silicon, financing · 2026-08-26 · earnings-call · https://gateway.drillr.ai/mcp/private

[38] NVDA 8-K filed 2026-08-26 (Q2 FY27 CFO commentary) · Data Center and Edge revenue drivers · 2026-08-26 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm

[39] NVDA 10-Q filed 2026-08-26 · H200 China licensing · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[40] NVDA 10-Q filed 2026-08-26 · Q2 FY27 inventory provisions and gross margin · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[41] NVDA 10-Q filed 2026-08-26 · accounts receivable concentration · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[42] NVDA 8-K filed 2026-08-17 (SB Energy PORTS residual value guaranties) · 2026-08-17 · 8-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000069/nvda-20260817.htm

[43] NVDA 10-Q filed 2026-08-26 · deferred revenue and customer advances · 2026-08-26 · 10-Q · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm

[44] NVDA 10-K filed 2026-02-25 · customer financing and counterparty risk · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm

[45] NVDA 10-K filed 2026-02-25 · FY2026 customer concentration · 2026-02-25 · 10-K · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm