Many people probably feel that NVIDIA's earnings are no longer surprising. When a company posts double-digit growth every quarter, is there still a point in scrutinizing the numbers? Yet the Q2 earnings released on August 26, 2026, weren't just a story about scale. GAAP net income exceeded Non-GAAP net income—a reversal of the usual relationship—and the main driver wasn't GPU sales but gains on equity securities. What's more, the $108 billion Q3 guidance the company presented still assumes zero data center compute revenue from China.
Inside the $96.2 billion quarter: the two pillars behind $89.0 billion in Data Center revenue
NVIDIA's fiscal 2027 second-quarter results (for the period ended July 26, 2026), announced after market close Pacific time on August 26, 2026, came in at $96.2 billion in revenue. That's up 106% year-over-year and 18% quarter-over-quarter—growth that remains remarkably strong. Converted to yen, this is roughly ¥15 trillion (based on Nikkei's August 27, 2026 exchange rate), an extraordinary figure for a single company's quarterly revenue by global standards.
The Data Center segment, NVIDIA's primary earner, posted revenue of $89.0 billion, up 117% year-over-year—outpacing the company's overall growth rate. Within that, Hyperscale—covering public cloud providers and the world's largest consumer internet companies—brought in $48.7 billion, while the newly created "ACIE" category (AI Clouds, Industrial, and Enterprise), which aggregates demand from AI-dedicated data centers and AI factories across diverse industries and countries, contributed $40.3 billion. Combined, these match the reported $89.0 billion figure. The two categories are roughly equal in size, revealing a business structure that simultaneously carries dependency on a handful of massive customers and diversification across a broad customer base. If that balance were to tip, the growth rate itself could be significantly shaken.
NVIDIA's motive for creating these two new categories is to make visible just how much investor risk stems from dependence on a small number of giant customers. If Hyperscale numbers were to slow sharply, the numbers make clear how much that alone could cut into overall company growth. By breaking out ACIE separately, the company also lets investors track, quarter by quarter, how much broader-based demand—from government agencies to industrial customers—is developing.
Jensen Huang, at the Q1 FY2027 earnings call in May 2026, said the easiest go-to-market is the hyperscaler, "because there are only five or six of them." The fewer the customers a business relies on, the greater the swing that a single customer's investment decision can have on overall performance. ACIE's $40.3 billion also represents the breadth of demand that helps offset this concentration risk.
Revenue outside Data Center is confined to the newly created "Edge Computing" category, at just $7.2 billion. That's up 13% quarter-over-quarter and 27% year-over-year, but it represents less than a tenth of total company revenue. NVIDIA's revenue structure has become even more concentrated around AI semiconductors for data centers.
The company is also returning capital to shareholders. Combined share buybacks and dividends totaled roughly $26 billion, with the next dividend set at $0.25 per share (payable October 1, 2026). The remaining share buyback authorization stands at approximately $99 billion, underscoring the company's capacity to fund both growth investment and shareholder returns from its earnings.
The unusual reversal: how equity securities gains pushed GAAP net income above Non-GAAP
GAAP net income came in at $59.68 billion (diluted EPS of $2.46), exceeding Non-GAAP net income of $53.9 billion (EPS of $2.22). GAAP operating income—the company's core earnings power—stood at $63.734 billion, meaning the bulk of net income still comes from semiconductor and systems sales. Yet the reversal in which GAAP net income exceeded Non-GAAP net income occurred because NVIDIA, starting in the first quarter of fiscal 2027, changed its methodology to stop excluding stock-based compensation expense from its Non-GAAP metrics. In the past, excluding stock-based compensation was the primary reason Non-GAAP figures came in higher; that effect is now gone.
What took its place was "equity securities gains, net," a $7.771 billion figure that appears only in GAAP net income and is excluded from Non-GAAP. NVIDIA's official earnings reconciliation table explicitly discloses this amount, noting it combines both realized and unrealized gains and losses. This gain is the primary reason GAAP net income surpassed Non-GAAP. However, NVIDIA has not disclosed a breakdown of which investments generated how much of this gain.
The underlying driver is the accounting treatment of NVIDIA's equity holdings. Marketable equity securities—those traded on public markets—must be marked to market each quarter and flow through the income statement, even as unrealized gains or losses. For non-marketable investments in private companies, the company also revalues holdings and reflects the change in income whenever there is an observable price change.
For NVIDIA, as of July 26, 2026, its publicly traded equity holdings carried $1.5 billion in net unrealized gains, while its non-marketable holdings recognized roughly $4.9 billion in valuation gains during the quarter (non-marketable holdings also saw some impairments and reclassifications to public holdings when certain companies made their public market debut during the period). Even without selling a single share, if the value of a portfolio company rises, that gain is booked as profit on the income statement—and if it falls, it's booked as a loss. As large tech companies like NVIDIA increase their equity stakes in operating companies, a growing share of variability unrelated to core business performance can seep into their reported earnings.
Among NVIDIA's disclosed equity investments, its stake in SpaceX stands out. In an August 14, 2026 Form 13F filing (a quarterly disclosure of institutional investor holdings), NVIDIA disclosed holding 122.8 million shares of SpaceX Class A stock. Valued at the SpaceX share price as of the calendar-quarter-end date of June 30, 2026 ($170.86 per share), that stake was worth approximately $21.0 billion. The shares originated from a $10 billion investment NVIDIA made in xAI in January 2026 (part of a $20 billion funding round). When SpaceX, valued at $1 trillion, absorbed xAI, valued at $250 billion, in a merger completed in February 2026—creating a combined entity valued at $1.25 trillion—NVIDIA's xAI stake converted into SpaceX shares.
That means a $10 billion investment ballooned to a valuation of $21.0 billion as of the end of June 2026, in less than half a year. But the paper gains didn't keep pace. By late August, SpaceX's share price had reportedly fallen to $140 per share, shrinking the value of NVIDIA's stake to roughly $17.2 billion. An asset that had more than doubled in a bit over six months lost nearly $4 billion in value within a matter of weeks.
How guidance of $108 billion can exceed actual results—while still assuming zero China data center revenue
The Q3 (August-October 2026) guidance NVIDIA provided calls for revenue of $108 billion (plus or minus 2%) and a gross margin of 74.0% (plus or minus 50 basis points, both GAAP and Non-GAAP). This forecast, which would exceed the $96.2 billion actual result, comes with a caveat, however. NVIDIA explicitly stated in its earnings materials: "NVIDIA is not assuming any Data Center compute revenue from China in its outlook."
Shipments of data center GPUs bound for China have continued to shrink amid a tangle of U.S. export restrictions and Chinese regulatory constraints. Sales of the H20 have essentially halted, and while limited shipments of certain products such as the H200 continue, China now accounts for less than 1% of Data Center revenue. A market that once represented substantial revenue has, quarter after quarter, remained excluded from the company's forward-looking assumptions.
And yet guidance still exceeds the $96.2 billion actual figure—because the business has already grown large enough that non-China demand alone, across both Hyperscale and ACIE, can sustain that growth.
China was once a market of considerable importance to NVIDIA, but current guidance has effectively given up on incorporating it at all. Conversely, this means that if regulations were to ease at some point in the future and shipments to China were to resume, that would represent pure upside to guidance. The current $108 billion figure is, in a sense, close to a floor representing NVIDIA's "underlying strength" without China.
On August 10, NVIDIA also announced signing a memorandum of understanding (MOU) with six major financial institutions—Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—to establish an independent lending platform for AI infrastructure. The framework aims to mobilize over $500 billion in third-party capital over time, though a definitive agreement has not yet been signed. Still, it's a move designed to bolster buyers' financing capacity—a way to secure demand-side investment capacity even without the enormous Chinese market. Supporting growth without China requires not just supply-side technological strength but also financial strategies that shore up buyers' access to capital.
The scale of this capital expenditure cycle has direct implications for Japanese semiconductor equipment stocks as well. Tokyo Electron and Advantest are highly sensitive to AI-related capital spending trends; when U.S. semiconductor stocks plunged sharply in July 2026, both companies' shares fell by roughly 9%. Whether growth can continue without China is a variable that affects not only NVIDIA itself but the earnings of domestic equipment makers as well.
Why the stock kept swinging even after beating estimates
Despite beating market expectations on both revenue and profit, NVIDIA's stock behaved erratically in trading following the announcement. Released after U.S. market close, the results were followed by an earnings call during which buying and selling crossed paths, producing a reaction too tangled to describe with a single volatility figure. The stock seemed to react less to the good results themselves than to whatever came up during the subsequent call.
Some market observers have pointed out that the stock has now fallen after earnings for four consecutive quarters despite beating expectations each time. If that pattern were to continue, it would mark five consecutive quarters—a sign, in itself, that the market has already priced in NVIDIA's high growth. Once beating expectations stops being a "surprise," the stock starts reacting less to whether the numbers were good or bad and more to what comes next.
The Hyperscale-dependent structure—reliant on just five or six customers—is itself a source of jitteriness in earnings reactions. If even one of those customers were to slow its investment plans, growth rates could quickly unravel. Behind the market's inability to simply cheer good results lies wariness about the concentration risk embedded in NVIDIA's very scale.
Rising AI memory prices are also complicating investor sentiment. Q2's actual gross margin was 75.0% on both a GAAP and Non-GAAP basis, but Q3 guidance calls for 74.0% (plus or minus 50 basis points)—a one-point decline. It's reasonable to view the stock's volatility as reflecting caution over a combination of continued revenue growth paired with a slight erosion in profitability.
Behind the disappearance of the Gaming category, AMD closes in
Starting with the fiscal 2027 first quarter (ended April 26, 2026, announced May 20, 2026), NVIDIA restructured how it discloses its market platform categories in earnings. The previous five-category breakdown—Data Center, Gaming, Professional Visualization, Automotive, and OEM & Other—was consolidated into two: Data Center (subdivided into Hyperscale and ACIE) and Edge Computing. That structure remains in place in this latest report. The company's formal accounting reporting segments remain "Compute & Networking" and "Graphics," but the granularity of revenue disclosure that investors actually see has changed with this market-category restructuring. NVIDIA explained the change as being intended to "better reflect current and future growth drivers."
Gaming was, for years, a standalone category NVIDIA disclosed separately in its quarterly earnings—arguably the face of the company. Growth rates for consumer GPUs and new product sales were long an important indicator for the stock market, but that standalone disclosure has now vanished with this restructuring. Investors no longer have a way to track the standalone momentum of the gaming business from official earnings reports.
Consumer GPU market trends are now buried within the $7.2 billion Edge Computing category. While NVIDIA has explained its reasoning for the restructuring, that explanation still doesn't reveal how the gaming business's standalone growth rate is trending.
The moves by the company's chasing competitor are also worth noting. AMD, in its calendar Q2 earnings announced August 4, 2026, reported that data center revenue doubled year-over-year. The company has already announced deployment agreements spanning multiple generations of AMD GPUs—up to 6 gigawatts with OpenAI (with the first phase being 1 gigawatt of MI450) and up to 2 gigawatts of the MI450 series with Anthropic—showing steady progress in securing major customers.
Still, the scale gap remains enormous. Even if AMD's data center revenue doubled year-over-year, its absolute figure remains far below NVIDIA's $89.0 billion in Data Center revenue. The contrast between an NVIDIA whose gaming business has become invisible and an AMD strengthening its presence in data centers underscores where the industry's main battleground truly lies.
What it would take for the $1 trillion Blackwell/Vera Rubin order pipeline to hold up
Backing NVIDIA's bullish guidance is its outlook for orders of next-generation GPUs. At the GTC keynote on March 16, 2026, Jensen Huang said the cumulative order visibility for Blackwell and Vera Rubin through 2027 stood at "at least $1 trillion." That's double the figure the company had previously cited—"$500 billion in very high-confidence demand and purchase orders spanning 2025 to 2026"—achieved in under a year. This figure, however, is not a confirmed backlog of signed contracts; it's the company's own estimate of demand visibility.
This $1 trillion order outlook is underpinned by a dual transformation. One is that NVIDIA has strengthened its identity as an "investor," being simultaneously the seller of semiconductors and a shareholder in buyers such as xAI and SpaceX. The other is that the business has transformed to the point where it can post guidance exceeding $96.2 billion even while assuming zero data center compute revenue from China—a structure now sustained entirely by demand outside China. Neither transformation is visible if one looks only at the momentum of GPU sales.
The conditions for the $1 trillion order outlook to avoid becoming a pipe dream are narrow: Hyperscale and ACIE investment must not slow down; the MOU with the six major financial institutions must advance to a definitive agreement, with the targeted $500 billion-plus in mobilized capital actually converting into real capital expenditure; and export restrictions targeting China must not worsen from current levels. As Jensen Huang put it: "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue. And demand is accelerating."
The next focal point will be NVIDIA's Q3 earnings report, expected around November 2026. If the company hits its $108 billion guidance, that will confirm the China-free growth model is becoming durable rather than a one-off. Alongside that, how far SpaceX's share price recovers, and how the momentum of AI investment ripples through the earnings of domestic Japanese equipment makers like Tokyo Electron and Advantest, will offer clues as to whether the $1 trillion order outlook is truly converting into real revenue. Demand for GPUs itself remains robust, but what this $96.2 billion quarter truly revealed is a company solidifying a structure that can hold up even without the enormous China market—while simultaneously expanding its bets as a shareholder beyond the boundaries of its own supply chain.