The S&P 500 edged lower after the close even though NVIDIA topped quarterly expectations and July core PCE inflation came in as forecast. The index had finished the cash session nearly unchanged at 7,675.70 before NVIDIA reported. NVIDIA posted revenue of $96.2 billion, above roughly $92 billion in consensus estimates, up 106% from a year earlier. Earnings per share came in at $2.22 versus the $2.09 estimate, more than doubling from $1.05 a year ago. On the macro front, core PCE rose 0.2% month over month and 3.3% year over year. Headline PCE increased 0.2% on the month and 3.7% on the year, with both headline readings running 0.1 percentage points above forecasts. NVIDIA shares ended regular trading at $209.66, down 1.59% from the prior close, according to TradingView. The stock later climbed to $218.72 in after-hours trading, a gain of 4.32%. Gross margin was 75%, with guidance pointing to a slight decline to 74% next quarter. Hyperscaler revenue more than doubled to $48.7 billion. Revenue from the AI cloud, industrial, and enterprise segment surged 138% to $40.3 billion. CEO Jensen Huang said demand was accelerating, calling AI an inflection point as computing begins generating "real revenue." NVIDIA carries the largest weighting in the S&P 500. The index rose 0.32% on Tuesday to 7,677.28, marking a third straight winning session. Futures were lower ahead of NVIDIA's Wednesday report. Markets have been balancing rate and valuation signals. The 30-year Treasury yield hit levels not seen in nearly two decades last week, and the subsequent pullback in yields helped tech shares regain some ground. Investors have also weighed valuation risks tied to the S&P 500's heavy concentration in artificial-intelligence-related names. Despite several recent earnings beats, NVIDIA shares have often struggled to hold gains as traders demand increasingly ambitious guidance to justify the company's valuation. The stock is up only modestly so far in 2026. The next major event on the calendar is Kevin Warsh's speech at Jackson Hole on Friday.

Galaxy is re-entering the U.S. retail crypto-lending market with a new collateralized borrowing product, according to CoinDesk. The firm said it launched the Crypto Portfolio Line of Credit on August 25 via its GalaxyOne platform, allowing eligible U.S. customers to access a revolving credit line and borrow cash against BTC, ETH and SOL without selling their crypto. Unlike earlier models that often required separate loans by asset, Galaxy's offering combines multiple cryptocurrencies into a single credit line. The company disclosed a variable annual interest rate of 8.99% and an initial loan-to-value ratio of 50%, implying that $100,000 of eligible collateral could support roughly $50,000 in borrowing capacity. Galaxy said approved funds are typically available instantly and can be used within the platform or withdrawn in USD or USDC. The company also emphasized that pledged collateral will not be rehypothecated, lent out or otherwise reused. It added that the platform continuously monitors collateral values and will notify users if prices fall before determining any next steps. For customers with already-staked SOL, Galaxy said they can keep SOL staked and continue earning staking rewards without unstaking. The service is offered through GalaxyOne Lending LLC and is currently available in 40 U.S. states. It is not available in California, Delaware, Idaho, Indiana, Minnesota, Mississippi, Missouri, Nevada or South Dakota. Galaxy's move comes as the market gauges whether retail crypto lending can regain traction after the sector's 2022 shakeout, when Celsius, BlockFi and Voyager collapsed, freezing customer funds and triggering forced liquidations that sharply curtailed retail activity. Galaxy said its new product is designed around a regulated platform, a clearer credit structure and collateral that is not lent out, aiming to balance rates, flexibility and asset protection using institutional-grade infrastructure.

UPDATE: U.S. stock futures moved higher after Nvidia surged more than 4% on results that topped expectations.

US companies delivered one of their strongest quarters in decades, with profits from current production surging in the second quarter of 2026. According to Bureau of Economic Analysis figures released August 26, profits from current production rose $400.9 billion in Q2. That compares with a $74.4 billion increase in Q1. After-tax profits as a share of gross value added climbed to 19.4% from 18.2%, the highest reading since the series began in the 1940s. On a seasonally adjusted annual rate basis, total corporate profits—adjusted for inventory valuation and capital consumption—reached $4,827 billion in Q2, about 9% above $4,426 billion in Q1. The profit surge arrived alongside signs of an economy still holding up. Real GDP increased at a 1.5% annualized rate in the second estimate, while nonresidential fixed investment grew at an 8.5% annualized pace. Richard Moody, chief economist at Regions Financial, said stronger profits are boosting business spending power: "Profit growth is freeing up cash that is helping support business capital spending, with capex growth extending beyond AI related investment." Equity-market results echoed the government data. For the S&P 500, blended earnings growth was roughly 50% year over year, and net profit margins rose to multiyear highs across the index. For investors, the combination of modest economic growth and a near-10% quarterly jump in profits points to a key driver: margin expansion, not volume growth.

Nvidia delivered another blowout quarter, reporting fiscal 2027 second-quarter revenue of $96.2 billion. The result surged 106% from a year earlier and rose 18% from the prior quarter, topping Wall Street's roughly $92.2 billion estimate by about $4 billion. The stock had slipped around 1.6% in regular trading on Aug. 26 amid typical pre-earnings volatility for mega-cap tech. After the report, shares reversed higher in after-hours trading. Data center revenue, which now represents about 92% of total sales, reached $89 billion, up 117% year over year. On the bottom line, GAAP net income was $59.7 billion, or $2.46 per diluted share. Non-GAAP EPS came in at $2.22, above the consensus near $2.10, extending Nvidia's streak to 15 consecutive quarters of results ahead of expectations. Guidance also beat forecasts. Nvidia projected fiscal Q3 revenue of $108 billion, plus or minus 2%, exceeding the approximately $104.2 billion consensus. On the earnings call, CEO Jensen Huang said "AI has reached its inflection point." The company also forecast 70% revenue growth for fiscal 2028. CFO Colette Kress cautioned that supply constraints could limit how quickly Nvidia can meet demand, citing production capacity as a potential bottleneck. Nvidia's market capitalization has now moved above $5 trillion. Even so, the stock is up only about 12% to 14% year to date, highlighting investor debate around valuation and revenue concentration. With data center sales accounting for 92% of the total, the business remains closely tied to the pace of continued AI infrastructure buildout.

Nvidia shares surged in after-hours trading after the chipmaker reported record revenue of $96.2 billion.

Amazon and NVIDIA said Amazon is set to deploy roughly 2 million high-end NVIDIA chips across 2027 and 2028 to train and run AI models, according to ChainCatcher. Amazon also disclosed in March that it plans to install 1 million NVIDIA chips in its AWS data centers. The agreement covers GPUs including Blackwell Ultra, Rubin, and Rubin Ultra.

Morgan Stanley says the U.S. AI data-center boom is running into a hard constraint: electricity supply. In a mid-August 2026 report, the bank estimates U.S. data centers will require about 68 gigawatts of new power capacity from 2026 to 2028. Roughly 30 gigawatts can be met through projects already under construction and existing contracted grid capacity. The remaining 38 gigawatts lacks a clear source. The bottleneck is the grid interconnection process. In parts of the U.S., interconnection requests can take five to seven years to approve and complete, while hyperscalers building AI infrastructure are seeking capacity immediately or within two to three years. With timelines misaligned, operators are increasingly turning to "behind-the-meter" generation, producing power onsite instead of relying on the broader grid. Morgan Stanley estimates onsite natural-gas turbines could contribute 15 to 20 gigawatts through this route. Fuel cells and nuclear co-location are also being explored, with some proposals placing data centers next to existing or newly built nuclear plants. The report notes these alternatives still face hurdles, including local opposition and labor shortages. Morgan Stanley identifies GE Vernova, Eaton, and Vertiv as key beneficiaries of the widening infrastructure gap. GE Vernova's gas-turbine backlog has risen from 100 gigawatts to 116 gigawatts, and the company is targeting 125 gigawatts by year-end. The data-center market is cited as a major driver, with orders in the segment reportedly doubling. Eaton, which supplies electrical components and power-management systems, reported an 85% surge in data-center orders in Q2 2026. Vertiv lifted its full-year revenue outlook to $14 billion after posting 24% year-over-year revenue growth in Q2 2026, with quarterly revenue of $3.27 billion. On the demand side, Microsoft and Amazon are projected to spend about $785 billion in capital expenditures in 2026, with the figure nearing $1 trillion in 2027. Global data-center construction spending is expected to approach $2.9 trillion through the end of 2028, while hyperscaler lease commitments are projected to exceed $1.2 trillion over the same period.

Nvidia delivered a blockbuster second quarter of fiscal 2027, posting revenue of $96.2 billion. The filing also highlighted two notable items: a $400 million charge tied to excess H200 GPU inventory and a high level of customer concentration, with one unnamed customer contributing 16% of quarterly revenue. H200 inventory charge The H200, a flagship Nvidia GPU designed for large-scale AI training and inference, features 141 GB of HBM3e memory and high-bandwidth performance aimed at data center deployments. For the quarter ended July 26, 2026, Nvidia recorded a $400 million provision against excess H200 inventory. Nvidia did not specify a single driver. Potential factors include shifts in demand, mismatches between production timing and orders, and geopolitical and regulatory frictions that have complicated sales of advanced chips in certain markets. The charge is far smaller than the $4.5 billion hit Nvidia took in fiscal Q1 2026, which was tied to older H20 inventory and directly linked to U.S. export restrictions on chip sales to China. Limited H200 shipments reached select customers in China, including ByteDance and Tencent, ahead of August 19, 2026. Those deliveries were reported at roughly 10,000 units each, well below Nvidia's previously accumulated backlog of more than two million units. The shortfall between what had been planned and what could ultimately be delivered likely underpins the $400 million write-down. Customer concentration: 16% from a single buyer Nvidia disclosed that one customer accounted for 16% of fiscal Q2 revenue. On $96.2 billion, that implies about $15.4 billion from a single buyer. The company did not identify the customer, consistent with common SEC disclosure practice once an individual customer exceeds the 10% reporting threshold. Market attention typically centers on major hyperscalers—including Microsoft, Meta, Amazon, and Google—which have been investing heavily in GPU clusters for AI infrastructure. This concentration is most evident in Nvidia's Compute & Networking segment, which includes data center GPU revenue. Export controls remain a strategic factor The prior quarter's $4.5 billion H20 charge underscored Nvidia's exposure to regulatory risk. The H200 inventory issue suggests constraints are not limited to older product lines, particularly given that reported H200 deliveries to ByteDance and Tencent were minimal versus earlier expectations. What to watch At $400 million, the H200 charge is manageable relative to Nvidia's scale—less than 0.5% of $96.2 billion in quarterly revenue. Investors will likely focus on two questions: whether excess H200 inventory is cleared through future sales or requires additional write-downs, and whether the 16% customer concentration increases or begins to diversify as more enterprises and sovereign AI initiatives enter the GPU market.