Earnings Scoreboard - Downstream and priced to prove it

“We're in, what I would characterise, as a 100-year flood on the memory pricing with exponential increases in memory prices."
Tim Cook, CEO of Apple, on the Q2 earnings call, 30 July 2026

Who’s scoring highest and why

From 28 July to 3 August, 176 S&P 500 companies, including 9 Dow Jones components (Amazon, Apple, Boeing, Chevron, Microsoft, Procter & Gamble, Coca-Cola, Sherwin-Williams and Visa) reported earnings. According to FactSet, the blended Q2 earnings growth rate for the S&P 500 now stands at 47.4% as of 31 July, up from 37.9% a week earlier and the 18.8% pace anticipated at quarter-end. As with Alphabet the week before, the headline overstates the operating reality. Amazon's reported EPS of $5.75 included a net non-operating gain of approximately $53 billion, primarily reflecting the markup of its stake in Anthropic, and the company's beat alone accounted for 76% of the past week's net dollar-level increase in index earnings. Excluding both Alphabet and Amazon, the blended growth rate falls to 28.8%. This would still mark the second consecutive quarter above 20% and the seventh consecutive quarter of double-digit earnings growth. On either measure, the quarter remains exceptional.

The top line tells a cleaner story. Blended revenue growth stands at 13.2%, the strongest since Q2 2022, and 77.2% of companies have reported revenue above estimates. At the sector level, Energy leads with earnings growth of 135.3%, a direct expression of the Strait of Hormuz disruption's transfer of income to producers and processors, followed by Communication Services at 109.8% and Consumer Discretionary at 90.7%. Healthcare remains the only sector reporting a y/o/y earnings decline, a fact that frames Tuesday's and Wednesday's pharmaceutical prints. The index's forward twelve-month P/E of 19.7x leaves little valuation slack should the discount rate continue to climb.

The FOMC held the federal funds target at 3.50% to 3.75% on 29 July for a fifth consecutive meeting, but the 9 – 3 vote carried three dissents in favour of a 25 bps increase. The probability of a September hike implied by CME FedWatch has risen to 62.7% as of 3 August, from 55.8% on 28 July and 46.1% on 2 July. The 30-year Treasury yield closed the FOMC session at 5.193% and the 10-year at 4.657%. Every duration-sensitive print this week, from Disney's streaming ambitions to Constellation's and Vistra's decade-long power contracts, will be discounted against that curve.

The themes defining this week: Energy's forward curve, AI's multi-layer cross-examination and Healthcare's bifurcation

Refining margins face their first direct test of durability since the Strait of Hormuz disruption redrew the energy complex. Valero's 297.2% q/o/q EPS increase last week was driven almost entirely by a geopolitically-induced supply shock. This week's Marathon Petroleum print, with consensus EPS estimates having been revised upward by 94.1% since the beginning of the quarter, will determine whether that windfall is broadening across the domestic refining complex or concentrated in companies with particular regional or crude-sourcing advantages. The critical question for investors is one of duration. Refining margins tend to mean-revert once global refined-product inventories rebuild. The pace of that rebuild depends on how quickly the underlying Middle East supply disruption resolves. However, this geopolitical dividend may last six months to a year, extending beyond the initial supply shock because global refined-product inventories take far longer to rebuild than crude-oil inventories alone. Investors should weigh this dynamic against Occidental Petroleum's and Chevron's more upstream-focussed exposure.

AMD’s print can shape the expectations within the Compute and Silicon layer beyond Nvidia. Microsoft's and Amazon's CapEx disclosures validated that hyperscaler AI infrastructure CapEx cycle remains robust, but neither company disclosed the precise silicon-vendor mix underlying that spending. AMD's Data Centre segment, expected to reach $6.511 billion in Q2 revenue, roughly 2x the year-ago figure, is the most direct proxy investors have for whether AMD's Instinct MI350 accelerators and the upcoming MI450 platform are capturing a meaningfully growing share of that hyperscaler capital-spending pool, particularly given the expanded Meta partnership that is expected to begin contributing gigawatt-scale deployment revenue in H2 2026. 

If AMD's print and forward guidance confirm accelerating Data Centre momentum alongside continued greater-than-70% y/o/y server-CPU growth, it would suggest the AI-compute space is genuinely bifurcating between multiple viable suppliers rather than consolidating around a single name. Arista’s print this week, the same day as AMD’s, provides a complementary signal because its networking fabric revenue is largely independent of the accelerator vendor ultimately favoured by hyperscalers. As a result, its results offer a useful cross-check on the broader trajectory of AI infrastructure spending, even if AMD’s data centre numbers disappoint.

The electrification and grid-modernisation trade tests how broadly AI-driven power demand is spreading. GE Vernova's, Vertiv's and Dominion Energy's results over the past two weeks established that the physical infrastructure supporting AI data-centre deployment, encompassing turbines, transformers, thermal management and regulated transmission capacity, remains capacity-constrained across nearly every node of the supply chain. This week's Constellation Energy and Vistra prints extend that theme into the generation layer itself, testing whether nuclear and merchant power producers are capturing pricing power commensurate with the demand growth that grid-equipment makers have described.

The distinction between Constellation's largely contracted, nuclear-heavy generation portfolio and Vistra's more merchant, Texas-ERCOT-exposed fleet offers investors a useful natural experiment. If both companies deliver strong results despite meaningfully different exposure to regulated versus competitive power markets, it suggests the AI-driven electricity-demand thesis is broad enough to lift power generators regardless of market structure. This would be a considerably more bullish signal for the sector than if only the more contracted, regulated names outperform. 

Albemarle's results this week, while nominally a lithium and battery-materials story rather than a grid-infrastructure one, offer an adjacent read on the broader electrification theme. This is because sustained high lithium pricing reflects continued strong demand for battery storage, an increasingly important complement to intermittent renewable generation as grid operators seek to balance AI-driven baseload demand growth against decarbonisation commitments.

Consumer and healthcare earnings challenge demand resilience seen earlier this earnings season. McDonald's deliberately difficult y/o/y comparison, Molson Coors' expected second consecutive quarter of double-digit percentage EPS decline and Fox's negatively-revised earnings estimate collectively suggest that not every corner of the consumer-facing space is experiencing the earnings strength characteristic of the broader S&P 500 thus far this reporting season. Read alongside Uber's continued double-digit percentage bookings growth and Eli Lilly's nearly 40% revenue growth guidance, the picture that emerges is one of genuine bifurcation. Consumer categories tied to convenience, mobility and differentiated healthcare innovation continue to show robust demand. However, more commoditised categories such as quick-service restaurants and traditional brewers, are exposed to secular volume decline among younger demographics, and therefore face a considerably more challenging setup.

This week’s preview: refining margin’s duration, AI's physical layers at play, GLP-1’s growth engine and consumer resilience

Major Earnings on the Docket

• Marathon Petroleum opens the week on Tuesday before the market opens, with the earnings call scheduled at 11:00 am EDT. Consensus calls for EPS of $14.27, against $3.96 a year earlier, an increase of 260.4% y/o/y, on revenue of $40.866 billion. The Street’s unusually wide estimate range of $13.38 to $15.55 reflects uncertainty over how much of Q2’s crack-spread expansion Marathon captured versus how much was offset by crude-cost inflation and turnaround activity. Furthermore, the consensus EPS figure has been revised upward by 94.1% from $7.35 at the end of March, a signal of just how much the Strait of Hormuz-related disruption to global refined-product flows has reshaped the earnings outlook for domestic refiners.

The Refining & Marketing segment is the entire story. The Street has pencilled in a refining margin of $32.84 per barrel, against $17.73 a year earlier, on net throughput of approximately 2.995 million barrels per day (bpd) and utilisation guided near 94.0%. The comparative evidence set a demanding bar last week. Valero delivered Refining operating income of $4.444 billion, 3.7x the prior year, on throughput of 2.950 million bpd and a Refining gross margin per barrel of $23.62, up 91.3% y/o/y and ahead of the $22.96 estimate. For Q2, the Street has modelled Marathon’s Operating income to increase 291.9% y/o/y to $4.910 billion, reflecting a margin expansion from 6.8% the prior year to 17.4%. Gross margin per barrel is anticipated to be $32.42. Marathon's Gulf Coast and Mid-Continent footprint is, if anything, better positioned for the export pull that redirected product demand toward US capacity during the disruption.

Two further variables warrant attention. First, capital returns, where the buyback programme of roughly $5 billion remains central to the per-share story and where the pace of repurchases will indicate management's own view of mid-cycle earnings power. On its Q1 earnings call, the company reaffirmed its capital-return priorities, noting that share repurchases would continue to be supported by strong cash flow and MPLX distributions through the year. Following the additional $5 billion repurchase authorisation announced on 5 May, the company would have had $8.6 billion available under its share repurchase programmes as of 31 March. Second, and most consequentially, forward commentary. Management's assessment of product inventories, which remain depleted globally, and of the durability of capture rates into Q3. Marathon provides the most direct listed read of the extraordinary refining-margin environment’s potential second act into Q3.

• McDonald’s also reports Tuesday, before the bell. The earnings call is scheduled at 8:30 am EDT. McDonald’s offers a read on the health of the value-conscious consumer at a moment when fast-food peers have flagged decelerating traffic. Consensus calls for EPS of $3.32, up 4.0% y/o/y, on revenue of $7.129 billion, up 4.2% y/o/y. 

The central dynamic is a deliberately tough year-over-year comp. Q2 2025 benefited from a strong Minecraft-branded promotional campaign in April that drove global comparable sales up 3.80% and US comparable sales up 2.50% in the base period. Management has already flagged that April 2026 ran slightly negative in both the US and International Operated Markets segments as that comparison rolled off. The US segment, which posted 3.90% same store sales (SSS) growth in Q1, is expected to decelerate sequentially. The Street's US comparable-sales expectation of 0.82% carries visible downside risk, with some estimates closer to 0.50%. The composition is as important as the headline figure, as investors will assess traffic against average cheque. A pricing-led comp alongside declining traffic would reinforce evidence that low-income consumers, a critical cohort for McDonald’s, are continuing to pull back. The McValue platform and the national beverage and McCafé initiatives are the company's structural answer, designed to rebuild visit frequency at the cost of near-term cheque growth.

The International Developmental Licensed Markets segment, which includes China and other franchise-heavy geographies, is also guided toward slower sequential growth from Q1’s SSS growth of 3.40% to estimates of 1.84% in Q2, even as management continues to emphasise market-share gains against value-oriented competitors. 

Margins are the second-order question. Operating margin, which stood at 46.0% in Q1 against a full-year guidance range in the mid-to-high 40% range, will be watched closely for evidence that value-menu promotions and beverage-platform investments are not eroding profitability even as they support traffic. For Q2, Operating margin is anticipated to be 47.6%, reflecting a slight decrease of 30 bps from a year ago. Q2 EBITDA margin is modelled at 55.5%, also decreasing by 30 bps from a year ago. McDonald’s print will serve as a signal of the degree to which US consumers are shifting more aggressively toward value tiers, a theme reflected in commentary from other quick service and retail names throughout this season.

• Pfizer reports Tuesday at 6:45 am EDT, with the call scheduled at 10:00 am EDT. Consensus calls for EPS of $0.68, down 12.6% y/o/y from $0.78, on revenue of $14.399 billion, down 1.7% from the year-ago $14.653 billion, against reaffirmed full-year guidance midpoint revenue of $61 billion and EPS of $2.80 to $3.00.

The revenue composition carries more information than the headline. The COVID franchise, guided to approximately $5 billion for the full year, continues its structural fade. The quarter's growth must come from the ex-COVID portfolio: the Seagenderived oncology assets, led by Padcev's expanding bladder-cancer footprint, the Vyndaqel franchise in cardiomyopathy, and Eliquis, whose approaching loss of exclusivity remains the largest medium-term revenue cliff in large-cap pharma. 

Investors will look for evidence that oncology's growth rate can outrun the erosion elsewhere, and for an update on the manufacturing-optimisation and cost-realignment programme, now targeting roughly $7.2 billion in cumulative savings. This has become the primary driver of the earnings bridge in the absence of top-line acceleration. Gross margin, which came in at 76.4% in Q1 against a mid-70s full-year target, is anticipated to be 76.2% in Q2.

Two strategic threads frame the call. The first is obesity, where Pfizer's route back into the incretin race — after discontinuing its own oral candidates — runs through business development. Any commentary on deal appetite will be read against a Lilly print twenty-four hours later that will demonstrate what the category is worth. The second is pricing policy, where the sector's exposure to IRA negotiation rounds and most-favoured nation proposals keeps a persistent discount on the group's multiple.

• Qnity Electronics reports Tuesday before the open, with the call at 8:00 am EDT, in only its second quarterly report as a standalone company following its 1 November 2025 spinoff from DuPont de Nemours and its addition to the S&P 500. The company's Q1 beat at the top and bottom lines was driven by organic growth linked to AI adoption, advanced-node transitions and advanced packaging and thermal-management demand, prompting management to raise full-year 2026 sales guidance to $5.225 to $5.375 billion from a prior $4.970 to $5.170 billion range. Consensus calls for Q2 revenue of $1.365 billion, against Q1's $1.315 billion, and EPS of $1.07, within full-year guidance of EPS of $3.80 to $4.14.

Qnity is the season's purest listed expression of semiconductor process materials. Its two segments map directly onto the AI buildout, as a constituent of the Process Materials and Fabrication Equipment layer of our Alpha Vibes framework. The Semiconductor Technologies segment supplies CMP pads, slurries, photoresists and advanced cleaning materials to leading edge logic and memory fabs, providing a consumables-based view of fab utilisation because materials volumes closely track wafer starts in real time. For Q2 Semiconductor Technologies’ sales are anticipated to be $718 million. Interconnect Solutions, which grew 25.0% y/o/y in Q1 to $593 million, supplies the advanced substrates and films that AI accelerators and high-density packaging consume in increasing intensity per device. For Q2, the segment’s revenue is projected to reach $640 million.

Management has indicated that data-centre-exposed applications represent roughly 15-20% of the portfolio and constitute its fastest-growing demand vertical. The print arrives with unusually strong corroborating evidence. Lam Research guided Q3 mid-point range sales to $8.100 billion, 20.5% sequential q/o/q growth, raised its 2026 wafer-fab-equipment outlook to the low-$150 billion range and reported NAND systems revenue that more than doubled sequentially. If wafer starts are following equipment installation at anything like that cadence, Qnity's consumables volumes should be inflecting. 

Following the Q1 guidance increase, investors will watch for further margin expansion, with EBITDA margin expected at 30.5% in Q2 after reaching 31.1% in Q1, as volumes scale. They will also look for incremental commentary on recently announced semiconductor manufacturer collaborations, which management cited as evidence of strengthening customer relationships across the AI value chain.

• Arista Networks reports Tuesday after the close, with the call at 4:30 pm EDT. Consensus calls converge above the company’s guidance with Q2 EPS estimates of $0.89, up 21.4% y/o/y from $0.73, on revenue of $2.830 billion, up 28.4% y/o/y. The company’s Q2 guidance contemplates revenue of $2.800 billion and EPS of $0.88, with operating margin guided to 46% to 47% and gross margin to 62% to 63%.

Arista occupies the networking layer of the AI buildout. Its print provides this week’s clearest indication of the continued strength of AI-driven networking capital spending, as suggested by last week’s hyperscaler CapEx guidance from Microsoft and Amazon.

Arista's back-end AI networking franchise, Ethernet switching deployed inside accelerator clusters, is a direct beneficiary among networking vendors. The recently launched 7060Xseries 1.6-terabit platforms extend the roadmap into the next optics generation. The variables that will decide the reaction are familiar and were outlined at Bank of America’s Global Technology Conference on 3 June by Jayshree Ullal, CEO & Todd Nightingale, COO. First, the AI-networking revenue trajectory against the company's stated multi-year targets, and the breadth of that revenue beyond the two largest cloud customers, whose concentration remains the bear case's anchor. Second, gross margin, where supply-constrained optics and component-cost inflation, the same memory-and-components pressure Apple flagged in its guidance, test the company's pricing power. For Q2, gross margin is anticipated to be 62.40%. Third, competitive positioning against both white-box designs built on merchant silicon and the vertically integrated networking stack of the dominant accelerator vendor, a tension that resurfaces with every hyperscaler architecture decision.

Recent templates provide a cautionary backdrop. Vertiv fell 12% on a revenue miss it attributed to project timing, and Corning fell 14% despite a beat, because at these multiples the guide is the print. Arista's deferred-revenue and purchase-commitment disclosures, which lead reported revenue, may matter more than the quarter itself.

• SpaceX also reports Tuesday after the close, with the webcast scheduled at 4:30 pm EDT, in its first quarterly report as a public company since June's IPO. Consensus calls for a loss of $0.23 per share on revenue of $6.826 billion, a sequential acceleration from Q1's $4.694 billion. 

The company reports across three segments, and each tells a different story. Connectivity, anchored by Starlink, is the only consistently profitable business and the print's foundation. The Street has pencilled in segment revenue of $3.836 billion, at an EBITDA margin of 61.9%, building on a subscriber base of 10.3 million across 164 countries as of 31 March, up 105.0% from 5.0 million a year earlier. The decomposition matters more than the headline. Subscriber additions against average revenue per user, the enterprise and government mix, churn and the pace at which the V3 satellites deployed by Starship's 25 July test flight expand network capacity will determine whether Connectivity reads as a compounding utility.

The second storyline is the one the market is genuinely litigating: whether Starlink's cash flows can fund the rest of the AI ambition. The AI segment, assembled through the February merger with xAI and spanning X, Grok, Cursor and third-party compute leasing, generated $818 million of Q1 revenue against a $1.943 billion operating loss and negative EBITDA of $609 million, while absorbing $10.107 billion of quarterly CapEx and an outflow of FCF of $9.060 billion. This was a burn rate that places SpaceX squarely inside the season's defining tension, reporting into the same cost-of-capital examination that separated Microsoft from Meta last week. For Q2, FCF is expected to be negative $1.980 billion and CapEx to reach $13.224 billion, from $10.107 billion in Q1.

Deployed compute capacity stood at approximately 1 gigawatt at the end of Q1 and is expected to roughly double by year-end. The full-year revenue expectations embedded in current forecasts imply H2 revenue of $27.822 billion. This is a ramp that rests almost entirely on AI compute contracts and Starlink, and that management must either underwrite with disclosed commitments or walk down. 

The Space segment's commercial arc beyond government launch contracts hangs on Starship's timeline, where the thirteenth test flight's successful V3 deployment marks operational progress that does not yet translate into revenue. 

Beyond the headline financials, the print carries unusual structural significance because it activates the company's staggered insider lock-up schedule. This means that the first tranche of as much as 20% of restricted shares, up to 911.5 million shares, worth roughly $116 billion at current prices, the largest lock-up expiration on record, become eligible for sale starting 6 August, two trading days after the print. An additional 10% of locked-up shares could be released if SpaceX shares close more than 30% above the $135 IPO price on at least five of the ten trading sessions preceding the report, a threshold the stock is unlikely to clear given it currently trades more than 30% below the level required. This ensures that whatever management says Tuesday evening will be tested by supply before the week is out.

• AMD closes Tuesday after the market at 4:15 pm EDT, with the call at 5:00 pm EDT. Management's own Q2 guidance calls for revenue of $11.200 billion, plus or minus $300 million, representing growth of 45.7% y/o/y and 9.2% q/o/q, with gross margin guided to 56.0%. Consensus has converged just above that midpoint, with the Street modelling revenue of $11.310 billion and EPS of $1.62, representing growth of more than 236.7% y/o/y. 

The segment mix is the print. Data Centre revenue is the most important line item, with consensus expecting the segment to reach $6.511 billion in sales, 2x the year-ago figure, driven by the continued ramp of Instinct MI350-series accelerators targeting generative AI training and inference workloads, alongside server CPU revenue that management has guided to grow more than 70% y/o/y. Because the MI350 generation, rather than the more advanced MI450 platform, is expected to account for the bulk of recognised Q2 revenue, investors will focus on forward commentary regarding the MI450 and Helios rack-scale architecture ramp under the expanded Meta partnership, where shipments supporting an initial gigawatt-scale deployment are scheduled to begin in H2, alongside any incremental detail on deployment timelines with OpenAI and other hyperscale customers. Given AMD’s position as the most direct alternative to Nvidia in AI accelerators, this print, arriving the week after Microsoft’s and Amazon’s CapEx disclosures, will be viewed as a test of the extent to which the hyperscaler spending commitments announced last week are translating into diversified AI-silicon demand. The inevitable comparison will be to Nvidia's cadence. The credibility of AMD's claim to a durable second-source position in training and a genuine contender position in inference rests on sequential acceleration here. Additionally, Intel’s Q2 print, which pointed to continued strength in data centre CPU demand across enterprise and cloud customers, support the case for the segment’s revenue above consensus.

Beyond Data Centre, Client revenue is forecast to reach $3.005 billion, up 20.2%; y/o/y and a decline of 27.3% y/o/y in Gaming revenue to $815 million, as the console cycle ages.

Three surrounding dynamics shape the reaction function. First, recent hyperscaler CapEx reaffirmations define AMD’s demand pool, making allocation share the key issue. Second, export controls on China-bound accelerators continue to limit a market that once contributed meaningfully to Data Centre GPU revenue, so any commentary on licensed-product traction could move estimates. Third, gross margin trajectory and outlook will be closely watched for signs of scale economics, as early accelerator silicon has historically diluted mix during lower-yielding ramps.

• Walt Disney reports Q2 earnings on Wednesday at 6:40 am EDT, with the webcast scheduled at 8:30 am EDT. Consensus calls for EPS of $1.86, up 15.3% y/o/y from $1.61, on revenue of $25.392 billion, up 7.4% from the year-ago $23.650 billion, against a guidance in which management has pointed to operating income of $5.3 billion for the quarter.

There are three segments with three narratives. In Entertainment, the streaming portfolio crossed into double-digit operating margin for the first time last quarter at 10.6%, and the durability of that profitability, built on price increases, the ad tier's growth and password-sharing enforcement, is the print's central claim. The Direct-to-Consumer (DTC) streaming segment, encompassing Disney+, Hulu and ESPN's direct-to-consumer offering, remains the primary swing factor for investor sentiment, given the platform's transition toward sustained profitability after years of investment-driven losses. Subscriber trends alongside average-revenue-per-user (ARPU) trajectory will be closely parsed for evidence that price increases implemented over the past year have not materially dented engagement. For Q2, Disney+ monthly ARPU is projected to be $8.31 worldwide, divided into Domestic ARPU at $8.33 and International ARPU at $8.28 ARPU Disney International. The Entertainment’s segment revenue is projected to increase 9.1% y/o/y to $11.675 billion. The segment’s Operating revenue is projected to increase 54.7% y/o/y to $1.581 billion, reflecting an operating margin of 13.5%, an expansion of 399 bps from the year ago figure of 9.5%. 

In Sports, the quarter is expected to absorb a 14.0% y/o/y decline in operating income to $892 million, despite a 5.3% y/o/y increase in segment revenue, reflecting higher programming costs. The more consequential disclosure will be the early performance of ESPN’s DTC service, particularly subscriber conversion from the linear base, engagement trends and the betting and fantasy integrations intended to support the platform thesis.

In Experiences, domestic parks have sustained per-capita spending growth of roughly 5.0%, even as attendance declined 1.0% last quarter. The cruise expansion continues to incur pre-opening costs ahead of new capacity additions. For Q2, the segment’s operating margin is expected to improve to 28.4%, from 27.7% a year earlier, on revenue growth of 7.4% y/o/y to $9.758 billion. The consumer question embedded in the segment is the same one McDonald’s raises a day earlier, but viewed from the opposite end of the income distribution scale. Can the premium experiential consumer continue to absorb price increases?

Capital returns remain central to the equity story, with the buyback targeted at or above $8 billion for the fiscal year and $5.5 billion completed through H1. As Hugh Johnston, Senior Executive Vice President and CFO, outlined at the MoffettNathanson Media, Internet and Communications Conference on 14 May, management reiterated its commitment to double-digit DTC operating margins in 2026. He indicated that content spend is expected to reach $24 billion in 2026, with greater emphasis on international DTC growth, while parks and cruise investments continue under a $60 billion global CapEx plan over ten years, supported by strong early ROIC.

Disney’s print, read alongside Fox’s on Thursday, will bookend the season’s evidence on the extent to which streaming economics have matured into a model the market can capitalise.

• Eli Lilly reports Wednesday, 5 minutes later, at 6:45 am EDT, with the earnings call scheduled at 10:00 am EDT. It faces a genuinely bifurcated setup between headline consensus expectations and the market's likely focus on a single product line. Consensus calls for revenue of $20.685 billion, implying growth of nearly 33.0% y/o/y, and EPS of $6.01, reflecting a decline of 4.8% y/o/y, accounting for a $2.8 billion acquired research-and-development charge that will weigh on this quarter’s figure. Lilly's full-year guidance contemplates revenue of $82 to $85 billion and EPS of $35.50 to $37.00, reflecting y/o/y of 49.7% at the midpoint.

Mounjaro Q2 sales are expected to rise 71.7% y/o/y to $8.925 billion, while Zepbound Q2 sales are projected to reach $4.672 billion, reflecting growth of 38.2% y/o/y. Together, the incretin franchise is expected to exceed $13.5 billion for the quarter across Mounjaro and Zepbound, with investors likely to assess the print across three principal dimensions. The first is volume: prescription trends have remained strong. With supply now largely resolved after two years as the binding constraint, the debate shifts to the durability of demand and its sensitivity to price. The second is price: management has guided to low-to-mid-teens US price erosion for the year as the channel mix broadens, rebates deepen and the LillyDirect direct-to-consumer channel scales. This quarter’s realised per-unit trajectory will either validate or challenge that guidance. The third is geography: international launches remain early and underpenetrated, leaving meaningful room for incremental contribution over the medium term.

The second strategic driver is orforglipron, sold under the name Foundayo. As a once-daily oral GLP-1 obesity pill and the only approved product in its class with no food or water restrictions, Foundayo remains central to Lilly’s next leg of franchise expansion. Consensus now models $102.8 million in Q2 sales, down from $200 million at the beginning of the quarter, reflecting the view that the initial commercial ramp has been slower than investors first anticipated. Full-year 2026 US Foundayo sales are estimated at $967.9 million. The launch economics, including manufacturing cost, pricing architecture and the Medicare access bridge that began on 1 July, will be critical in defining the franchise’s next phase and its competitive defence against Novo Nordisk and emerging oral alternatives. Pipeline readouts in adjacent obesity indications, from sleep apnoea to cardiovascular outcomes and next-generation Retatrutide, further extend the addressable-market thesis.

The market will focus on new-to-brand prescription trends; payer and pharmacy-benefit-manager channel mix, including the LillyDirect direct-to-consumer platform; and updates on the broader Orforglipron Phase III trial program spanning diabetes, weight maintenance and obstructive sleep apnoea. Any of these factors could alter the multi-year growth trajectory already embedded in Lilly’s premium valuation.

• Uber Technologies reports Wednesday at 6:55 am EDT, with the webcast scheduled at 8:00 am EDT. Consensus calls for EPS of $0.80, up 27.7% y/o/y from $0.63, on revenue of $14.235 billion, up 12.5% from the year-ago $12.651 billion. EBITDA is expected at $2.782 billion against management's guidance of $2.750 billion at midpoint and gross bookings guided to $56.25 to $57.75 billion, implying constant-currency growth of 21.9% y/o/y at the midpoint.

The platform metrics will define the print. The Street expects Mobility gross bookings of $29.027 billion and Delivery gross bookings of $26.910 billion, supported by monthly active platform consumers of 207.9 million, up from 180.0 million a year earlier. That 15.5% y/o/y user-growth rate remains the foundation of the bull case, as higher frequency and cross-platform attachment, reinforced by Uber One membership of approximately 50 million, compound on top of a larger user base. Total take rate for Q2 is expected at 24.9%, up from 24.6% in Q1, but below 27.1% a year earlier. The margin architecture remains central to the FCF narrative: the EBITDA margin is expected to expand faster than bookings, with incremental margins flowing through and Q2 EBITDA margin pencilled in at 19.3%, implying expansion of 53 bps q/o/q and 257 bps y/o/y. Commentary on insurance-cost inflation, historically the P&L's least controllable line, will be especially important for the Mobility take-rate outlook.

The strategic overlay is autonomy. Uber has positioned itself as the demand-aggregation layer for autonomous vehicles, supported by partnerships that span Waymo's expanding city footprint and a broader set of international AV operators. Each quarter's disclosure on AV trip volumes and partner economics is therefore scrutinised for evidence that the platform is capturing value, rather than ceding it, as autonomy scales. The bear case holds that AV owners ultimately disintermediate the network. The bull case holds that fleet-utilisation economics require the demand density Uber controls. Management commentary on partnership economics, together with any update on international regulatory developments, will likely shape the print's reaction. Uber's results will be read alongside Lyft's commentary on Thursday as a high-frequency signal of discretionary spending on transportation services at a time when retailers and restaurant chains are flagging more cautious consumer trends.

• Albemarle reports Wednesday after the close, with the call scheduled for Thursday at 8:00 am EDT. The company enters the print against a markedly improved lithium-pricing backdrop, a reversal that has reshaped the earnings outlook for the world's largest lithium producer. Consensus calls for EPS of $3.20, compared with $0.11 a year earlier, with estimates ranging from $2.78 to $4.02 across 15 forecasts. Q2 revenue is expected to reach $1.607 billion, up 20.8% y/o/y.

The key context is the sharp recovery in lithium spot prices, which have risen from multi-year lows near $10,050 per metric ton a year ago to $21,200 per metric ton as of 3 August. Because Albemarle prices most contracts roughly one quarter in advance, the sequential price improvement realised in Q1 should continue to flow through Q2 results. The company's scenario-based 2026 guidance underscores the sensitivity of earnings to realised lithium prices: at an average price of $10 per kilogram, management guided to net sales of $4.1 to $4.3 billion and EBITDA of $0.9 to $1.0 billion. However, at $20 per kilogram, those figures increase to $5.7 to $6.0 billion in net sales and $2.4 to $2.6 billion in EBITDA.

The print breaks down into price, volume and cost. Realised lithium pricing is the central swing factor. Albemarle's contract structures cause realised prices to lag spot, making the quarter's average realisation the clearest measure of how much of the recovery has reached the P&L. Volumes introduce a separate tension. The company idled hydroxide conversion trains at Kemerton during the downturn and has emphasised capital discipline over growth. Commentary on restart economics and ore-grade trends at the Greenbushes joint venture will define the volume trajectory into 2027. On cost, management has flagged $70 to $90 million of Middle East-related supply-chain and logistics headwinds for the year, creating a direct transmission channel from the Hormuz disruption into specialty-chemicals economics.

Investors will also focus on Energy Storage segment volumes, which grew 14.0% y/o/y in Q1 alongside a 51.0% pricing increase, as well as any commentary on the Foreign Entities of Concern provisions under the expanded Inflation Reduction Act framework. Management has suggested that those provisions could support a durable pricing premium for Western-sourced lithium relative to Chinese production. The strategic question for the call is the quality of the recovery: a demand-led and durable upturn, driven by energy-storage and EV demand pulling the market into deficit, or a supply-led and fragile rebound dependent on curtailments that higher prices may ultimately reverse. Albemarle remains the clearest listed read on lithium's cycle position, with read-through for the broader battery-materials and EV supply chain.

• SanDisk reports Wednesday at 4:05 pm EDT, with the call scheduled for 4:30 pm EDT. Management guided Q2 revenue to $7.750 to $8.250 billion and EPS to $30 to $33, while consensus stands at $34.80 following a Q1 in which revenue rose 251.0% y/o/y to $5.950 billion and adjusted gross margin reached 78.4%. For Q2, sales are expected to reach $8.440 billion, implying a further acceleration to 344.0% y/o/y growth. Comparisons predating the February 2025 separation from Western Digital, together with subsequent share-count changes, continue to distort screen-level history and should therefore be assessed on the company's own reporting basis.

The NAND cycle has entered an unprecedented phase, with SanDisk positioned at its centre. AI data-centre demand for enterprise SSDs, high-capacity flash used for training-data staging, inference caching and the storage tiers beneath GPU clusters, has met years of industry capital discipline, creating the pricing backdrop reflected in the company's elevated gross margins. Corroboration this earnings season has come from both sides of the storage ecosystem. Seagate reported nearline exabyte demand effectively sold out into 2027, with hard-drive gross margin of 52.7%, while Lam Research disclosed that NAND equipment revenue more than doubled sequentially. The industry is now adding capacity, a development that confirms the cycle while also defining its eventual normalisation. 

Memory-cost inflation has become a cross-sector P&L line, cited by Apple in pricing decisions and by Amazon in CapEx guidance. SanDisk is one of the most direct listed expressions capturing it. At the segment level, sequential growth is expected to accelerate, with Data Centre sales projected to rise to $2.742 billion in Q2 from $1.467 billion in Q1, and Edge sales expected to increase to $4.459 billion from $3.663 billion.

The call will centre on three variables. The first is the trajectory of long-term supply agreements, which management has indicated exceed $42 billion in aggregate commitments. These convert current spot-price strength into multi-year revenue visibility. The second is execution on the BiCS8 and BiCS10 node ramps, which will determine cost-per-bit progress and therefore the durability of margin expansion. The third is the fiscal 2027 guide, arriving only days before the company's 13 August investor day and likely to anchor expectations for the next stage of the cycle. SanDisk’s results this week provide one of the clearest reads on the Memory and Storage layer of the AI buildout.

• Constellation Energy reports Thursday before the open, with the call scheduled at 10:00 am EDT. Consensus calls for revenue of $7.789 billion, up 27.7% y/o/y, but down from $11.122 billion in Q1. It calls for EPS of $2.34, against $1.91 a year earlier, with the dispersion itself notable, as estimates have drifted lower over the past quarter even as the power-demand narrative has strengthened, suggesting the Street is still calibrating the earnings power of Constellation Energy, post-Calpine.

Three factors will frame the print. The first is the contracting pipeline, where Constellation's twenty-year agreement to supply Meta from the Clinton nuclear plant, the Crane Clean Energy Centre restart at Three Mile Island for Microsoft and the broader queue of hyperscaler negotiations represent the clearest monetisation of data-centre power scarcity. Dominion underscored that scarcity last week by quantifying more than 50 gigawatts of demand in Virginia alone. The disclosures most likely to move the multiple are the cadence of new power-purchase agreements, their pricing relative to the PJM curve and management's stance on co-location compared with grid-supplied structures. The second is the capacity-market tailwind. PJM's latest auction cleared at $325 per megawatt-day, securing roughly 18,875 megawatts for Constellation and translating into approximately $2.24 billion of gross annual capacity revenue for the 2028 to 2029 delivery year. This creates locked-in earnings visibility that did not exist two years ago, although regulatory proposals to cap future auctions limit the scope for straight-line extrapolation. The third is Calpine integration. The acquisition, which closed in Q4 2025, roughly doubled the generation fleet and added gas flexibility that complements the nuclear base. Synergy capture and combined-entity FCF disclosure will shape the standalone-versus-pro-forma earnings debate.

The company's $3.9 billion CapEx plan and $5 billion share-buyback authorisation, both announced earlier this year, reflect management's confidence in sustained demand for carbon-free electricity from data-centre operators seeking to secure long-term, reliable power supply as AI compute deployment accelerates. Constellation reports as the purest listed expression of the thesis that electricity is the AI buildout's binding constraint, the same thesis Vistra tests a day later.

Last week’s themes and the bigger picture

The market prices the buildout's bill, energy banks its windfall and the physical economy keeps compounding

Meta's Wednesday print was a study in contrasts: a clean top-line beat overshadowed by an earnings miss and a CapEx bill that's starting to spook even patient investors. Revenue reached $60.801 billion, up 28.0% y/o/y and topping the $60.224 billion consensus. Advertising from the Family of Apps contributed $59.363 billion, powered by a 13.2% increase in ad impressions and a 12.0% rise in average price per ad. EPS landed at $6.18, badly missing the $7.19 consensus and down 13.5% y/o/y. Net income fell 13.6% y/o/y to $15.848 billion, dragged by $2.4 billion in legal charges and $1.18 billion in severance costs tied to May layoffs, alongside a jump in the effective tax rate to 16% from 11%.

CapEx increased 82.1% y/o/y to $30.116 billion for the quarter, up from $16.538 billion a year ago, pushing FCF down to just $784 million from $8.549 billion in Q2 2025. Operating margin compressed to 31.3% from 43.4% a year ago. Management raised the low end of full-year 2026 CapEx guidance to $130 to $145 billion.

Shares fell in after-hours trading, marking a ninth straight losing session heading into the print. Bloomberg news framed it alongside Microsoft's stronger Azure print the same night as a ‘study in contrasts’ for AI monetisation.

The narrative forming on the Street is one of patience wearing thin. On the earnings call, analysts pressed CEO Mark Zuckerberg on which product lines would show ‘quantifiable, material ROIC’ by 2026–27, as well as Meta's plan to simultaneously lease out and purchase compute capacity. Meta is increasingly viewed as still searching for that ‘undeniable’ payoff, even as Zuckerberg insists AI is already accelerating the core business.

Microsoft’s results delivered a clear, broad-based beat that drew praise from the Street rather than scepticism. Microsoft revenue reached $90.007 billion, ahead of the $87.620 billion consensus and the company’s guidance of $87.800 billion. Q2 EPS of $4.74 came in higher than the $4.24 Street estimate. Intelligent Cloud segment’s revenue $39.306 billion, up 31.6% y/o/y. The key metric investors were watching, Azure growth, came in at 39.4%, in line with the 39% to 40% guidance range.

Shares jumped in extended trading, a sharp contrast to the muted-to-negative reactions that have greeted Alphabet and Meta’s AI-capex-heavy prints this earnings season.

CapEx discipline became the core narrative. Microsoft delivered the Azure beat without materially lifting its CapEx outlook, keeping guidance broadly in line rather than raising it sharply as markets had feared after Alphabet’s recent increase.

Microsoft also reported that Copilot, its AI offering embedded across Office applications, reached 30 million paid users, up from 20 million in the prior quarter, though the company continues to face challenges converting its large user base into paying AI customers.

Microsoft cut its calendar 2026 CapEx forecast to roughly $175 billion from about $190 billion after extending the estimated useful life of its data centres and office buildings to 25 years from 15. The accounting change will shift more future data centre leases into operating leases, which are excluded from CapEx, while leaving the company’s underlying investment plans unchanged.

The Street’s emerging view is that Microsoft is meeting the standard investors have set for hyperscalers this earnings season. It is converting AI-led cloud demand into revenue quickly enough to justify infrastructure spending, rather than relying on faith in a future payoff. With 2027 CapEx guidance also in focus, the quarter was read as evidence that Microsoft’s spending is generating revenue within the same fiscal year, easing the credit-quality and ROIC concerns weighing on other AI-infrastructure names.

Amazon's Thursday print completed the AI-capex trilogy after Meta and Microsoft on a distinctly bullish note. Net sales reached $200.606 billion, up 19.6% y/o/y and crossing the $200 billion mark for a single quarter for the first time, comfortably ahead of the $197.035 billion consensus. EPS came in at $5.75 versus a $1.82 Street estimate, though that figure was inflated by a $53.4 billion non-operating pre-tax gain tied to Amazon's stake in Anthropic. Operating income rose 43.2% to $27.461 billion and net income surged to $62.647 billion from $18.164 billion a year earlier.

AWS stole the show. Cloud revenue reached $42.232 billion, up 36.8% y/o/y, its fastest growth in 18 quarters and well ahead of the 31.3% growth analysts had modelled. AWS operating income reached $16.621 billion, ahead of the $13.624 billion consensus, with AWS operating margin expanding to 39.4%. CEO Andy Jassy noted that ‘AWS is booming... our AI and Chips businesses each eclipsed run rates of more than $25 billion,’ pointing to Trainium adoption from Anthropic and OpenAI as evidence the custom-silicon bet is paying off. Advertising also contributed to the positive sentiment, increasing 26.2% to $19.809 billion.

CapEx totalled $54.208 billion for the quarter, up 68.4% y/o/y and above the $49.353 billion Street estimate, pushing trailing-twelve-month FCF negative to an outflow of $8.016 billion versus an $18.184 billion inflow a year ago. Despite that, shares jumped in after-hours trading.

The Street’s emerging view is that Amazon has offered the strongest proof so far that AI infrastructure investment is translating into cloud share gains and wider margins, even though AWS growth still lagged Microsoft Azure’s 43% and Google Cloud’s 82% this quarter. Amazon notched its fifth straight quarter of cloud share growth, reinforcing the view that among hyperscalers, CapEx discipline paired with reacceleration is what's earning investor conviction.

Apple’s Thursday print closed last week’s Mega cap earnings cycle on a cautiously negative note. The company beat nearly every major headline metric, but shares still declined after hours. This underscored that when expectations are this elevated, guidance must do more than simply meet the bar.

Apple reported record June-quarter revenue of $109.417 billion, up 16.4% y/o/y and slightly above the $109.039 billion consensus. EPS reached $2.02, up 28.7% y/o/y and ahead of the $1.89 Street estimate. Even excluding a $0.11 benefit from US tariff refunds, EPS still exceeded consensus. iPhone revenue increased 21.7% y/o/y to $54.252 billion, above the $53.742 billion estimate and now accounting for nearly half of total sales. Mac revenue rose 28.7% y/o/y to $10.352 billion, versus the $8.680 billion estimate. Services was the notable soft spot, increasing 12.1% y/o/y to $30.739 billion, below the $31.414 billion analysts expected. Regionally, Greater China revenue grew 22.4% y/o/y to $18.816 billion, broadly in line with the $18.771 billion estimate and signalling long-awaited stabilisation. Gross margin came in at 50.1%, lifted by roughly two percentage points from tariff refunds, though the underlying 48.1% still cleared the midpoint of guidance.

Despite the strong headline results, the stock declined in after-hours trading as investors focussed on guidance and supply-chain pressure tied to the global memory-chip shortage, which Tim Cook described as a ‘hundred-year flood’ and which has already led to price increases for Macs and iPads. The move also reflected a sell-the-news reaction after Apple’s strong July performance, its best month in four years, during which the company briefly reached a $5 trillion market capitalisation.

The report can also be viewed as a test of Apple’s position as the market’s AI safety play. Its capital-light approach, centred on licensing AI from Google rather than building hyperscale infrastructure, continues to stand in contrast with CapEx-heavy peers that face greater scrutiny. The quarter also carried symbolic weight, as it marked Tim Cook’s final earnings call as CEO before handing leadership to John Ternus.

Beneath the hyperscaler drama, the AI physical-infrastructure and semiconductor-equipment layers delivered some of the clearest evidence yet that capacity constraints, not demand uncertainty, remain the operative bottleneck. Lam Research posted record quarterly revenue and record EPS exactly in line with its own guidance, and raised its view of 2026 wafer-fab-equipment spending to above $140 billion. Vertiv beat on both revenue and adjusted EPS and raised full-year guidance again, reinforcing the view that data-centre power and thermal-management demand continues to outpace even elevated expectations. Seagate confirmed that nearline hard-drive capacity is effectively sold out through calendar 2027, an unusual multi-year visibility window that serves as evidence that the AI-driven storage buildout has structural legs well beyond the current fiscal year. Corning's Optical Communications segment once again outgrew the rest of the portfolio on hyperscaler fibre demand. Equinix delivered a cleaner AFFO print than the prior quarter, suggesting the debate over hyperscaler self-build displacing third-party colocation has not yet materially dented Equinix's bookings.

The week also delivered a decisive test of the defence, industrial and energy-transition-adjacent trade regime themes flagged the prior week. General Dynamics beat estimates on the back of continued Gulfstream and Marine Systems strength, extending the rearmament thesis that lifted Lockheed Martin and RTX the week before. Boeing's loss narrowed meaningfully and delivery pace accelerated toward the company's targeted 737 MAX production rate, a signal that the recovery is broadening beyond the aftermarket-services layer and into the airframe manufacturer itself. 

On the energy side, Valero's refining margins surged on the back of the continued Strait of Hormuz-related disruption to global refined-product supply, delivering one of the sharpest year-over-year EPS increases of the entire Q2 season. Chevron's results confirmed a strong sequential rebound in upstream realizations. Dominion Energy and Martin Marietta Materials each offered more nuanced reads. Dominion's data-centre-driven load growth continues to compound at the regulated-utility level, insulated from the volatility affecting hardware plays. Martin Marietta's diesel-cost headwind confirmed that not every industrial name is capturing tariff-era pricing power cleanly. 

Palantir's guidance raise, with full-year revenue now guided to $7.65-7.66 billion implying roughly 71% growth, extended the enterprise-AI-adoption thread that ServiceNow established the week prior. Robinhood's results, weighed down by softer crypto trading volumes, suggested a modest cooling in the most speculative corners of retail-investor risk appetite even as equities and options activity held firm.