Nvidia’s rare forecast for approximately 70% revenue growth in fiscal 2028—well above Wall Street’s prior 44% estimate—provided investors with an unusually specific long-range demand signal. Nvidia stock (NASDAQ: NVDA) initially fell more than 1% after Wednesday’s earnings release, reversed to gain nearly 5% in extended trading and rose 7.2% in Thursday premarket trading, Reuters reported. Nvidia and Amazon also announced that AWS plans to deploy 2 million additional Nvidia GPUs across its global infrastructure in 2027 and 2028.
The company also reported record fiscal second-quarter revenue of $96.22 billion and guided to $108 billion for the current quarter, beating Wall Street expectations on both measures. Yet the call also revealed that expensive memory is likely to push gross margin lower than previously expected, placing a direct profitability test alongside the growth outlook.
The central question for Nvidia investors is therefore not whether demand is growing. It is whether the company can convert that demand into durable margins and cash while managing supply shortages, customer financing and an expanding set of infrastructure commitments.
Nvidia Forecasts Approximately 70% Revenue Growth in Fiscal 2028
Nvidia’s preliminary fiscal 2028 outlook was unusually specific. The company expects revenue to grow approximately 70% in the year ending January 2028, compared with analysts’ average forecast of 44% before the results. Reuters described the projection as a rare year-ahead disclosure for a company that normally limits formal guidance to the next quarter.
Management said the forecast is constrained by available supply, not demand. Customer forecasts point to Nvidia’s business potentially doubling next year, but the company expects memory and other parts of the supply chain to remain bottlenecks through at least the end of fiscal 2028. That distinction supports the growth case, but it also means the 70% figure depends on Nvidia and its suppliers delivering an enormous capacity expansion on schedule.
AWS plans to deploy an additional 2 million Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs across its global infrastructure in 2027 and 2028. The expanded partnership also includes Vera CPUs, Nvidia’s Nemotron models on Amazon Bedrock and SageMaker, and Nvidia technology for Amazon’s warehouse robots. Nvidia did not disclose a contract value, and deployment over multiple years should not be treated as revenue already recognized.
The Vera Rubin platform has begun shipping and is expected to represent about 20% of Data Center revenue in fiscal Q3. That gives investors a near-term benchmark for the product transition: Rubin must scale rapidly while existing Blackwell systems continue to sell and constrained components are allocated across both platforms.
Nvidia Beats Revenue and Earnings Estimates
Nvidia’s fiscal Q2 results were well ahead of consensus:
- Revenue reached $96.221 billion, up 18% from the previous quarter and 106% from a year earlier.
- Adjusted earnings were $2.22 per diluted share, up 19% sequentially and 120% year over year.
- GAAP operating income rose 124% from a year earlier to $63.734 billion.
- GAAP net income more than doubled to $59.688 billion, or $2.46 per diluted share.
Wall Street analysts had expected $92.27 billion of revenue and adjusted earnings of $2.09 a share, according to the Associated Press. Nvidia therefore beat the revenue estimate by $3.951 billion, or about 4.3%, and the adjusted earnings estimate by $0.13 a share, or roughly 6.2%.
One qualification matters when reading the profit figures. Nvidia said GAAP other income included $7.8 billion of net gains from equity securities. Those gains helped reported net income but are market-dependent and should not be treated as recurring profit from selling chips, systems or software. The adjusted earnings figure, which excludes equity-security gains and certain other items, provides a cleaner comparison with Wall Street’s estimate.
Data Center Revenue Now Represents About 92.5% of Sales
Data Center revenue climbed 18% sequentially and 117% year over year to $89.023 billion. That represented approximately 92.5% of Nvidia’s total quarterly revenue, according to an ABBO News calculation.
Within Data Center, Hyperscale revenue reached $48.710 billion, up 102% from a year earlier. Revenue from AI Clouds, Industrial and Enterprise rose even faster—138%—to $40.313 billion. Outside Data Center, Edge Computing contributed the remaining $7.198 billion and grew 27% year over year.
The mix confirms Nvidia’s exceptional exposure to AI infrastructure, but it also shows the concentration risk. Nvidia’s Form 10-Q said one direct customer represented 16% of total quarterly revenue. The company also generates significant revenue from a limited number of indirect customers, some individually accounting for at least 10%. A slowdown in data-center spending, delayed site construction, or a shift toward customers’ in-house accelerators could therefore affect the overwhelming majority of current sales.
Nvidia is not simply a standalone GPU vendor. It sells processors, interconnects, networking, systems, software, and services across the broader AI chip stack. That integration can deepen its competitive advantage, but it also means investors must follow component supply, power availability, customer financing, and system deployment—not only GPU benchmarks.
The $108 Billion Outlook Beat Consensus, but Margin Is Set to Ease
For fiscal Q3, Nvidia expects revenue of $108 billion, plus or minus 2%. That creates a guidance range of $105.84 billion to $110.16 billion.
The midpoint is about 3.7% above the $104.19 billion LSEG consensus reported by Reuters and would represent another 12.2% sequential increase if achieved. Nvidia also forecast both GAAP and adjusted gross margin at 74.0%, plus or minus half a percentage point, compared with 75.0% in the reported quarter.
That one-percentage-point midpoint decline is financially meaningful at Nvidia’s scale. On $108 billion of revenue, a 1-point margin difference equals about $1.08 billion of quarterly gross profit. Absolute gross profit would still rise at the guidance midpoint because revenue is expected to grow, but the comparison shows why investors closely watch even small margin changes.
The pressure is expected to intensify before improving. On the call, Nvidia said sharply higher memory costs should push gross margin to a trough of roughly 71% to 72% in fiscal Q4, before it settles around 72% to 73% in fiscal 2028 as product price increases take effect. The company had previously expected a smaller cost increase. That reset is the clearest counterweight to the 70% revenue-growth forecast: sales can expand rapidly while each dollar of revenue generates less gross profit than investors had expected.
The company guided to approximately $9.2 billion of GAAP operating expenses and $9.0 billion on an adjusted basis. Continued spending on compute infrastructure and employees supports future product development, but it also raises the operating bar if revenue growth eventually slows.
China Remains Potential Upside, Not Part of the Base Case
Nvidia said shipments of Data Center Hopper products to China represented less than 1% of Data Center revenue during the quarter. Its Q3 outlook assumes no Data Center compute revenue from China.
The 10-Q adds three important qualifications. Nvidia recorded a $400 million H200-related charge for excess inventory and purchase obligations in the first half after demand diminished. Licensed H200s must be inspected in the United States and incur a 25% tariff when imported for that process; Nvidia said it has been unable to pass that tariff to customers. The company described itself as effectively shut out of China’s Data Center compute market under the current U.S. and Chinese rules.
That disclosure is an important update to earlier reports that China could approve limited Nvidia H200 purchases. Approval, licensing, and actual shipment are separate steps. The latest results show that China remained immaterial to Data Center revenue, while the current-quarter forecast treats any contribution as upside rather than part of Nvidia’s operating base case.
The exclusion also makes the $108 billion outlook more notable: Nvidia expects substantial sequential growth without relying on China. At the same time, it leaves a large market exposed to policy decisions, export controls, and domestic Chinese competitors.
Nvidia Disclosed $366 Billion of Future Commitments
The most consequential long-term disclosure was not on the income statement. In its CFO commentary filed with the SEC, Nvidia listed $366 billion of future commitments as of July 26:
- $279 billion for supply and capacity, primarily memory procurement;
- $29 billion for cloud-service agreements;
- $25 billion for data-center leases that had not started;
- $25 billion for equity investments; and
- $8 billion for capital expenditures.
Supplier commitments alone increased from $119 billion in the previous quarter to $279 billion. Nvidia is securing components years ahead to support demand and the Vera Rubin product cycle, but the increase also raises execution risk if customer deployment schedules, product mix, or AI spending change.
The timing is important. Nvidia expects $120 billion of the $366 billion in the remainder of fiscal 2027, $100 billion in fiscal 2028, and $98 billion in fiscal 2029. The remaining $48 billion extends from fiscal 2030 onward.
These commitments are not the same as funded debt, and the full amount is not due immediately. The 10-Q says certain supply agreements may be cancelable, rescheduled, or adjusted before Nvidia places firm orders, although changes can create additional costs. Equity-investment commitments are also subject to contingencies. Even with those qualifications, the figures show that Nvidia is reserving capital and capacity at a scale that requires demand to remain durable for several years.
A Separate $56 Billion Supports Third-Party Projects
Nvidia disclosed another $56 billion of commitments under arrangements intended to help selected customers secure land, power and data-center capacity. That total consists of $36 billion for AI-cloud agreements and $20 billion for data-center leases intended for third parties.
Under the AI-cloud arrangements, the cloud providers buy Nvidia infrastructure while Nvidia commits to purchase cloud services. The providers can stop supplying those services to Nvidia and sell the capacity to third parties at more advantageous rates, reducing Nvidia’s commitments as the capacity is used. If specified conditions are met, Nvidia will share in revenue generated from third-party customers. Nvidia also expects to reassign the separate third-party leases, which have terms of about 15 years and are expected to commence between fiscal 2028 and fiscal 2029. Their start dates depend on construction completion.
This structure can expand the addressable market by helping customers that are growing faster than their own balance sheets. It also links Nvidia more closely to customers’ financing, utilization, and credit performance. Investors should therefore distinguish ordinary product demand from demand that depends on Nvidia-backed infrastructure arrangements.
Nvidia said it entered into memoranda of understanding with six large investment firms to establish independent financing platforms designed to mobilize more than $500 billion of third-party capital for AI infrastructure over time. The 10-Q cautions that these preliminary arrangements may not lead to definitive agreements. At its option, Nvidia may provide limited residual-value support for portions of specific projects, subject to disciplined risk management and project-by-project evaluation. The fundraising target is not a $500 billion Nvidia cash commitment and should not be added wholesale to the company’s balance-sheet obligations.
On the call, Nvidia also said financing platforms with six large investment firms aim to raise more than $500 billion of third-party capital for AI infrastructure. That is an external fundraising target—not another $500 billion Nvidia cash commitment—and should not be added to the company’s balance-sheet obligations.
The connection between growth and financing is nevertheless material. Nvidia expects demand from AI labs for which it plans to use its balance sheet to contribute roughly one-quarter of its total business next year. That does not mean Nvidia will finance one-quarter of sales, but it shows why customer credit and project utilization now matter more to the investment case.
Maximum Gross Guarantee Exposure Reaches $108.5 Billion
Nvidia also reported maximum gross exposure of $108.5 billion under guarantees. Of that amount, $3.5 billion covers certain AI-cloud partners’ land, power, and shell lease obligations in the event of default.
The other $105 billion relates to guarantees signed in August for SB Energy’s PORTS-Pike Technology Campus in Ohio. Nvidia said the guarantees support approximately 4.25 gigawatts of IT load under 20-year leases to a customer that the filing identifies as an affiliate of OpenAI Group PBC. Subject to limited exceptions, the campus will exclusively host Nvidia AI infrastructure.
The $105 billion is a cap, not a current cash outflow. The guarantees become effective in phases as conditions are met, with the first expected in fiscal 2029, and the guaranteed amount is expected to decrease over each phase’s 20-year lease term. The 10-Q says the guarantees cover defined portions of lease and power payments—not the site’s full cost or all tenant obligations—and can terminate if OpenAI achieves a satisfactory credit rating. That distinction is essential: maximum guarantee exposure measures a contingent risk, not an amount Nvidia has already paid or recorded as ordinary debt.
Still, the arrangement shows how Nvidia’s role is expanding from supplying processors to helping enable the physical and financial infrastructure around them. That could reinforce demand, but it also creates risks that investors would not capture by tracking quarterly GPU shipments alone.
Cash Flow Lagged Profit as Receivables and Inventory Rose
Nvidia generated $24.077 billion of GAAP operating cash flow and $21.341 billion of company-defined non-GAAP free cash flow in Q2. Both improved from a year earlier, but free cash flow fell from $48.554 billion in the previous quarter despite the 18% sequential revenue increase.
The company attributed the sequential operating-cash-flow decline to working-capital adjustments and cash taxes. Accounts receivable ended the quarter at $63.1 billion, while days sales outstanding rose to 60 from 45 in Q1. Nvidia said the increase reflected extended payment terms on large, multi-quarter agreements with certain investment-grade customers.
Inventory rose about 22% sequentially to $31.6 billion from $25.8 billion as Nvidia prepared for the Vera Rubin ramp in Q3. That build can support another major product cycle, but investors should verify that inventory converts into shipments, receivables, and ultimately cash on schedule.
Nvidia ended the quarter with $56.6 billion of cash, cash equivalents, and marketable debt securities. It also issued $25 billion of senior unsecured notes, taking long-term debt to about $32.4 billion from roughly $7.5 billion at the end of fiscal 2026. The company returned approximately $26 billion to shareholders through buybacks and dividends and had about $99 billion left under its repurchase authorization.
The balance sheet remains substantial, but this quarter illustrates why revenue, accounting profit, and cash flow should be analyzed separately.
What Nvidia Investors Should Watch Next
The bullish case remains straightforward. Revenue more than doubled, both major Data Center customer groups grew rapidly, Q3 guidance topped consensus without assuming China sales, and management expects approximately 70% fiscal 2028 growth despite supply constraints. AWS’s 2 million-GPU plan and Rubin’s expected 20% share of current-quarter Data Center revenue give investors concrete checkpoints behind that outlook.
The risk case has become broader. Nvidia must manage a margin decline that could reach 71% to 72% in fiscal Q4, $366 billion of disclosed future commitments, a separate $56 billion of third-party project commitments, maximum gross guarantees of $108.5 billion, customer concentration, longer payment cycles, and a large inventory build. It must do so while hyperscale customers develop custom chips and policymakers continue to shape access to China.
Retail investors should track five figures in the next update:
- Revenue against the $108 billion midpoint and the $105.84 billion lower end of guidance.
- Rubin’s contribution against management’s target of about 20% of Data Center revenue.
- Gross margin against the 74.0% Q3 midpoint and the 71% to 72% Q4 trough forecast.
- Days sales outstanding and free cash flow, not only net income.
- Changes in supply commitments, guarantees, and China revenue actually shipped.
The Bottom Line
Nvidia delivered a powerful quarter, an above-consensus quarterly forecast, and a rare long-range revenue-growth projection. The shares reversed an initial after-hours decline during the earnings call, when management disclosed the approximately 70% fiscal 2028 revenue-growth outlook and discussed the expanded AWS partnership. Those disclosures coincided with the reversal, but the sequence does not prove they were the only drivers of the move or that the longer-term risks have disappeared.
The more useful conclusion is that Nvidia’s investment case is changing. AI demand remains the engine, but the company is increasingly securing supply, committing to infrastructure, and supporting customer buildouts years in advance. If that demand remains durable, those commitments can protect capacity and expand Nvidia’s market. If deployment, financing, or utilization falls short—or memory costs remain elevated—the same operating model can amplify risk.
After a $96 billion quarter, Nvidia no longer needs only to prove that it can sell more AI compute. It must also show that an increasingly capital-intensive ecosystem can convert those sales into margins and cash at the scale investors expect.
