美国互联网行业数字广告2Q26：AI身份危机

Digital Ads 2Q26: AI Identity crisis Rarely, correction never, have we gone into an earnings season with almost no attention paidto core business performance. Such is the gravity of the AI trade: Trim companies that couldget disrupted by AI and those writing the checks and buy the companies receiving thosechecks while giving fewer and fewer widgets for those same checks. It shouldn’t come asa surprise that for the large Internet stocks to work, investors are asking Meta to put on aNeoCloud’s clothes and Google to offer evidence that there’s an NVIDIA hiding inside ofGoogle Cloud. Mark Shmulik+1 917 344 8508mark.shmulik@bernsteinsg.com Wenhuan Chang+1 917 344 8546wenhuan.chang@bernsteinsg.com Deeksha Pandey+1 917 344 8447deeksha.pandey@bernsteinsg.com Momentum is everything?For much of 2Q, investor interest in Internet sector was non-existent, but inbounds picked up meaningfully over the past few weeks as the AI hardwaretrade lost some steam and attractive relative valuations offered a potentially compellingentry point - that is if, of course, Meta and Google could offer durable ROIC commentaryagainst the ever-increasing CAPEX number. With some crumbs around model progress,willingness to rent out infrastructure even temporarily, and rising TPU-sales businessexpectations, there may be some hope yet for the Internet megacaps. The generational ad run continues in 2Q… can it sustain?Under a stable macro, adchannel checks were strong for Meta and Google, with Meta’s performance acceleratingthrough the quarter, while Google Search’s offered stable month-over-month performance.Smaller platforms also saw encouraging trends with retail returning to Pinterest and Redditbringing on more advertisers and expand budget. Looking ahead, the comps get tougherfrom fading FX tailwinds for the larger players, while Pinterest and Snapchat face a mucheasier second half with room to improve. CapEx remains front and center.While 2026 CapEx could move modestly higher acrossboth Meta and Google (~$5B-$10B) to absorb BOM inflation and supply-chain costs,investor focus has already shifted to 2027, where spending could step up materially -Meta estimates already appear in the $225B-$250B range and Google potentially ~$325B-$375B+. Encouragingly, AI monetization levers are also beginning to emerge,from compute rentals, third-party chip sales and infrastructure partnerships to a growingpipeline of AI-driven products/revenue streams. Details and economics of these revenuestreams will likely be front and center during the earnings call. Return of retailer ad spending, especially the Chinese ones.As retailers lap the tariff-related disruptions of 2025, we are seeing signs of ad budgets shifting back toward theU.S. While Temu's global ad spend remains down -14% Y/Y in 2Q26, Shein's total spendis up +9% Y/Y, and both companies are reallocating budgets toward the U.S. market up389% Y/Y for Temu and 89% Y/Y for Shein. Stick with the big?Amazon and Meta remain our top picks, with the ability to deliver KPIbeats while gaining control of their AI narrative. Google Search and Cloud look strong,although expectations & valuation are relatively high within the group. Questions onmodel competitiveness, infrastructure bottlenecks, and tougher search comps keep uson the sidelines. PINS could see revenues inflect as retailer spending returns but hard tounderwrite, while questions remain over Reddit’s engagement, and SNAP remains a show-me story. BERNSTEIN TICKER TABLE INVESTMENT IMPLICATIONS META: Risk-reward skews positive here.Core performance remains impressive though we note tougher comps ahead in3Q - if Meta can continue to beat and raise, we have our tablestakes to get into the heart of the debate. We all know Capex isgoing higher and with commentary around capacity expected to come online in 2027 and BOM pricing trends across memory,networking, and XPUs, we can land in a $225B-$250B range. We could also see this year’s CapEx move higher once again byanother $5B-$10B largely to account for BOM inflation. Unlike last quarter that was mostly mum on new AI initiatives, we expectthis earnings call to be a lot more specific on Meta AI progress, Muse Spark’s monetization potential, short-term infrastructuredeals, and a few more things for good measure. It’s a start. We don’t expect investors to buy in overnight, especially with fearsaround a rising stock price into an equity issuance to fund these initiatives a la Google last quarter. Ultimately, Meta only needsto get one idea to hit product-market fit to shift investor sentiment. The valuation remains attractive relative to mega-cap peers,supported by multiple AI-related optionality vectors. We bring up our revenue and CapEx estimates.We remain Outperformon Meta, with a PT of $850/Share.We value Meta using a 50/50 combination of 2027e EV/Sales multiple of 7x, and aDCF using an WACC of 10% and a terminal growth of 3.5%. We consider Meta primarily a digital advertising business andbenchmark valua