LT Weekly: Aug 21, 2026

As always, when new 13F filings are disclosed, we like to start with taking a look at Berkshire Hathaway’s portfolio changes in the period, and this past quarter marked a notable shift under new CEO Greg Abel as the firm was a sizable net buyer of equities for the first time in 14 qtrs, purchasing $23.5bn of stock while selling just $3.7bn. The firm’s most notable purchase was a large ~83% expansion of its Alphabet position to ~$38b, making it Berkshire’s 3^rd largest equity holding, up from the 5^th largest last qtr. The firm also incr’d its positions in Macy’s (+142%) and Delta Air Lines (+44%), both of which were initiated last qtr, and further incr’d its positions in Lennar (+30%) and The New York Times (+4%). Berkshire’s only new position was a very small ~$0.6mm investment in D.R. Horton, which was a re-entry into the homebuilder after most recently exiting the position in Q3 2025. On the selling side, Berkshire decr’d its positions in Capital One (-58%), Nucor (-52%), Kroger (-22%), Ally Financial (-7%), and DaVita (-5%), while further reducing its stake in Bank of America (-6%) for the fifth consecutive qtr. The firm also sold out of its position in Constellation Brands, while maintaining its positions in SiriusXM, Occidental Petroleum, Kraft Heinz, Coca-Cola, Chubb, and Chevron. As of quarter-end June, the fund’s five largest holdings were Apple (19.7%), American Express (14.3%), Alphabet (10.6%), Coca-Cola (9.8%), and Bank of America (8.4%), with the top five collectively representing ~63% of the total portfolio.

Where Are The Top Hedge Funds Investing, In Aggregate?  

Each quarter, WhaleWisdom tracks the stocks that were most bought and sold during the quarter across 150 of the top hedge funds. It ranks the “hottest” stocks based on a formula that takes into account the number of buyers adding and initiating new positions vs sellers, the change in average ranking that the stock had in the portfolios, and the number of times the stock appears in the top 10 holdings of the portfolios. The biggest takeaways were:

• TMT’s favorability continued….8 of the top 10 “hottest” stocks were in TMT this qtr, up from 7 in Q1: They were Intel (#1), Western Digital (#2), Applied Materials (#3), Micron Technology (#4), Snowflake (#6), CoreWeave (#7), Analog Devices (#8), and TTM Technologies (#9)
  • Interestingly, only 2 of the 5 MAANG stocks remained in the Top 100 – they were Meta, which decreased q/q (#49 in Q1 -> #80 in Q2) and Alphabet, which increased q/q (63 in Q1 -> #33 in Q2)

While most US TMT companies are past earnings season, Alibaba and Baidu both reported this week and their results reinforced a consistent theme also evident in US tech. Namely, AI cloud demand remains very robust and AI-related revenues are scaling rapidly, but the necessary capital expenditures to meet demand are materially increasing and pressuring FCF.

More specifically, our main takeaways from these China Tech giants are that 1) AI cloud growth accelerated y/y at both companies w/ Alibaba’s AI External Cloud revenue growing +45% y/y (a 22-qtr high) while Baidu’s AI Cloud Infra revenues grew +50% y/y (though decelerated from +79% y/y in Q1 and missed consensus), with GPU cloud revenue growth at Baidu accelerating to +283% y/y; 2) Both companies are ramping capx aggressively to meet the AI compute demand given that Alibaba spent CNY67.7bn (~US$10bn), up +75% y/y, and Baidu spent CNY11.39bn (~US$1.7bn), +200% y/y, resulting in significant FCF pressure in each case; 3) Proprietary AI chips are gaining traction at both companies (Alibaba’s T-Head and Baidu’s Kunlunxin) which will be key margin levers as they replace commercially procured GPUs; 4) Specific to Baidu, weakness in its advertising business (42% of total revenue) due to a delay in AI search monetization was a negative focus (and that pressure will continue in H2) along with a step down in autonomous ride volumes; and 5) Specific to Alibaba, its China E-Commerce revenue declined -8% y/y (and surprised on the downside), though its Quick Commerce business more than offset (+45% y/y).

See below for more details on these most important updates from Alibaba and Baidu’s results this week.

-> BABA slightly rose a little over 1% on the back of its results while BIDU materially fell almost -13% in response to its print; YTD, both companies have been under a lot of pressure with BABA down almost- 20% and BIDU down almost -30%

Alibaba’s AI External Cloud Accelerated To +45% Y/Y (22-Qtr High) W/ Expanding Margins BUT AI Capx Surged, Driving A Large FCF Loss

• Alibaba delivered a solid FQ1 with in-line total rev and better adj EBITA vs Street expectations (the non-GAAP EPADS missed due to below-the-line items)
  • Revs grew +9% y/y, in-line w/ cons
  • Adj EBITA fell -30% y/y but beat cons by 7.8%
  • Non-GAAP EPADS fell -42% y/y and missed cons by -16%

• Higher capx was a main sticking point with investors and FCF was under pressure…
  • FQ1 capx grew +75% y/y to CNY67.7bn (~US$10bn) given cont’d AI infrastructure investment
    • The Co has now spent CNY190bn of its CNY380bn 3-year investment plan (annc’d Feb 2025), broadly in-line w/ expectations
  • FQ1 FCF loss was CNY44.7bn, which was up meaningfully from a loss of CNY18.8bn in the same period last year…this is mainly attributable to higher cloud infrastructure expenditure
  • Going forward…Mgmt cautioned not to annualize this qtr’s figure as hardware deliveries follow different procurement cycles w/ quarterly volatility

• …and the higher capx has also led to an even lower buyback rate
  • FQ1 repurchases totaled US$162mn, vs ~US$1bn in all of FY26 and a MUCH higher US$9bn in FY25

• BUT mgmt tried to soothe concerns with a detailed AI capx ROIC framework
  • At the current avg gross margins, AI-related capx can breakeven in ~3 years
    • At a 3-year payback period, keeping growth below 33% would enable positive FCF, BUT that is “not our strategic choice”, given AI is still early-stage
  • As gross margins improve and the proprietary chip substitution rate increases, the payback period is expected to shorten to 2.5 years or less
    • At a 2.5-year payback, the Co can pursue growth of 40%+ while also maintaining positive FCF
  • Three levers enhance ROIC:
    • Continuing to develop state-of-the-art models to improve AI product gross margins and expand the higher-margin MaaS business
    • Ramp proprietary T-Head chip deployment to replace commercially procured chips
    • Co-build data centers w/ partners and collect prepayments for compute services
  • Mgmt indicated that older GPUs (V100 from 2018, A100) are still running at full capacity, hence the useful life and positive cash flow generation extends beyond the 3-year payback

• Also note that the Co changed segment reporting starting this qtr
  • Alibaba E-commerce Group:
    • Unified domestic + cross-border e-commerce (China E-comm, Quick Commerce, Int’l E-comm, Global Wholesale)
  • AI Cloud & Compute Services:
    • Combined Cloud Intelligence Group + T-Head
  • AI Labs & Applications:
    • Consolidated AI model labs, Qwen Consumer Business Group, and QwenWork (previously in All Others)
  • All Others:
    • Alibaba Health, Hujing Digital Media & Entertainment Group, Amap, Lingxi Games, other tech businesses

• FQ1 AI Cloud & Compute Services revenue accelerated to +45% y/y w/ external revenue also growing +45% y/y (to a 22-qtr high) & mgmt expects further acceleration as compute demand continues to outstrip supply: Growth was “broad-based” driven by AI compute, storage, MaaS, and AI applications
  • Model + application services ARR (incl. MaaS) surpassed CNY16bn as of August
    • Outlook – mgmt remains confident in achieving the YE target of CNY30bn ARR
  • FQ1 AI-related product rev reached CNY12.4bn (annualized at CNY 49.5bn or US$7.3bn), marking the 12th consecutive qtr of triple-digit y/y growth
    • AI-related products now account for 35% of external cloud rev and generate “significantly higher gross margins” than the avg cloud portfolio
    • Outlook – mgmt indicated that next qtr’s annualized AI-related product rev is expected to approach US$10bn

• Alibaba’s proprietary chips (T-Head) are seeing broad commercial adoption…the CEO calls them “definitely the best among domestic Chinese chips”
  • Adoption: “Hundreds of companies” are leveraging them via Alibaba Cloud across embodied AI, autonomous driving, and large model companies
  • As mentioned above, scaling these chips will be a key margin lever

• Adj EBITA for AI External Cloud rose a strong +133% y/y and margins expanded to 12% from ~7.2% y/y, driven by improved economies of scale and stronger pricing power for AI-related products in a supply-constrained market
  • Outlook – Mgmt expects EBITA margin to “further expand steadily in the coming quarters” through “improved resource utilization, optimized model portfolio, and innovating new scenarios”

• When asked about the long-term cloud target, mgmt expressed high confidence in reaching ~US$100bn in external cloud rev by 2030 w/ good visibility to a 20% gross margin …this is based on:
  • Industry supply-demand dynamics (compute shortage expected through at least 2030)
  • Their full-stack competitive advantages
  • Pricing power from their proprietary models

• AI Labs & Applications loss widened significantly y/y but improved seq
  • Revs grew +16% y/y & the adj EBITA loss dramatically incr’d to CNY13.9bn from CNY3.2bn in the same period last yr which was driven by incr’d investment in AI capabilities and higher inference costs related to the Qwen App
    • But to note, the losses narrowed q/q due to reduced marketing spend on the Qwen App
  • Outlook – mgmt expects the segment loss to narrow over the coming quarters, driven by improving efficiency in both model training and marketing spend

• The Co’s Qwen model ecosystem continues to scale
  • Downloads: The Qwen model series has been downloaded 3bn+ times globally w/ 300k+ derivative models built on it
  • Monetization plans: Current API-based monetization of LLMs is a “short-term transitional approach” and mgmt sees the ultimate business model as delivering actual products, results, and operations as the industry approaches AGI

• China E-Commerce remains under pressure BUT it is more than offset by China Quick Commerce scaling & improving unit economics
  • China E-Commerce rev fell -8% y/y and customer mgmt revenue (CMR, which is ad + commission revenue) decr’d -7% y/y primarily due to weaker transaction activities
    • Ex-the contra revenue impact from a new business development program, CMR would have grown +1% y/y on a like-for-like basis
    • 88VIP membership grew double digits y/y to ~64mn members as of June 30
    • Strategic focus going forward: Strengthening core supply capabilities while leveraging AI to enhance the shopping experience and improve operational efficiency
  • China Quick Commerce rev grew a strong +45% y/y driven by Freshippo and Taobao Instant Commerce (TIC)
    • TIC cont’d to improve unit economics q/q while maintaining market share, driven by higher AOV and enhanced fulfillment logistics efficiency
    • Quick Commerce losses narrowed substantially q/q
    • The Co is accelerating the integration of Tmall Supermarket into Quick Commerce and mgmt expects non-food transaction volume to surpass food within the next fiscal year
    • Freshippo maintained robust y/y growth in orders and rev, driven by expanding footprint into emerging cities/counties and deeper collaboration w/ TIC
    • Outlook…Quick Commerce is expected to achieve overall profitability in FY29 & mgmt believes it has the potential to contribute 30% of the platform’s total GMV long-term, becoming the “second growth curve” for e-commerce
  • International E-Commerce rev fell -1% y/y due to tariff policies and the geopolitical environment pressuring growth
    • BUT the business delivered significant improvement in profitability while maintaining growth in transaction volume
    • AliExpress achieved operating profit this quarter
    • Regionally: Local e-commerce platforms in Turkey and Middle East are growing rapidly; Southeast Asia operating efficiency continues to improve

Baidu Paints A Mixed Picture W/ Decelerating AI Cloud Infra Revenue, Accelerating GPU Cloud Revenue, Worse Than Expected Ad Revenue Results, & A Step Down In Autonomous Rides Volumes

• Baidu reported mixed Q2 results…worse on revenues and better on profitability
  • Total revs MISSED cons by ~1% (down -4% y/y), driven by:
    • Broad softness across online marketing services (MISSED by ~4%, -19% y/y as mgmt deliberately held back on AI search monetization) and AI-native marketing services (MISSED by ~20%)
    • AI Cloud Infra rev grew by +50% y/y but decelerated from +79% y/y in Q1 and MISSED cons by ~4% despite GPU cloud rev accelerating to +283% y/y; Revs fell -17% seq
    • Partially offset by legacy business outperformance (BEAT by ~9%)
  • Adj EBITDA BEAT cons by ~6% (20% margin vs cons 18.1%)
  • Non-GAAP EPADS MISSED cons by ~23% almost entirely below the operating line

• Capx ramped significantly to CNY11.39bn (US$1.68bn), which is up +200% y/y due to the Co’s aggressive investment in AI infrastructure…though no specific forward guidance was provided; FCF was in deficit
  • “We are still in an AI investment cycle and our commitment to that investment is unwavering”
  • Investments are driven by clear demand from both customers and internal business and much of the investment is expected to begin contributing to rev relatively soon
  • Q2 FCF was negative CNY7.95bn due to the capx ramp

• Even though AI Cloud Infra rev growth of +50% y/y disappointed the Street, mgmt believes growth will remain strong in H2 with potential for further acceleration
  • GPU cloud rev nearly quadrupled y/y, growing +283% and accelerated from +184% in Q1 (4th straight quarter of triple-digit growth)
    • GPU cloud is growing as a % of total AI Cloud Infra rev and carries a more attractive margin profile
  • AI Cloud Infra profit and margins both incr’d y/y in Q2

• Cloud demand drivers remain broad-based across both training and inference workloads, with computing supply still constrained across the market
  • Existing key clients (online gaming, e-commerce, lifestyle content) continue to increase usage and spending
  • The customer count is growing rapidly with new clients at varying sizes
  • Demand is broadening across verticals: Internet, embodied AI, autonomous driving, smartphones, financial services
    • Embodied AI rev grew ~6x y/y within the vertical mix

• The MaaS (model-as-a-service) Qianfan platform showed strong momentum…rev from external customers’ token usage grew 9x+ y/y, driven by rapid growth in daily avg token consumption
  • The platform offers one of China’s most comprehensive model libraries covering Baidu’s Ernie family and virtually all of China’s leading models
  • MaaS still a relatively small share of AI Cloud Infra rev today BUT mgmt sees it becoming an increasingly meaningful margin contributor over the longer term as usage scales and unit inference costs decline

• The Co’s proprietary AI chips (Kunlunxin) are gaining commercial traction and the listing of this business is in process, but there was no new news on that front
  • A growing number of customers are adopting Kunlunxin chips for an expanding range of AI workloads
  • The Co realized improved inference throughput and overall compute efficiency in Q2

• AI search is driving more engagement and user satisfaction but also weighing on ad revs as mgmt. prioritizes product over monetization
  • Online marketing services rev MISSED cons by ~4% and decl’d -19% y/y
    • AI-native marketing services rev MISSED cons by ~20% (~flat y/y)
  • The y/y decline reflects holding back on monetizing AI search
    • OUTLOOK…mgmt expects the ad business to remain under pressure in H2
  • Competition for users’ time and attention has “intensified further”
    • The priority is still getting the product and user experience right…monetization opportunities will “emerge naturally” as model capabilities, user experiences, and task completions improve
    • AI search quality is improving meaningfully as answers are more reliable, better structured, and presented in richer formats; Hallucination rates also remain low
  • The integration of Ernie Assistant is also deepening the search experience
    • Ernie Assistant DAU grew +83% y/y in June; Daily avg conversation rounds more than tripled y/y
• Baidu App MAUs also fell: 644mn in Q2 (vs 655mn in Q1) but no additional color was provided on the reasons for the decline
• The Co’s AI Applications portfolio continues to expand…Digital Humans, Vibe Coding & Agents are gaining ground
  • Digital Humans use cases are expanding both domestically and overseas
    • E-commerce livestreaming
    • Digital human videos
    • Real-time interactive digital humans
    • New video podcasts
    • Example of success…a well-known Chinese Internet company expanded its digital human livestreaming deployment to ~2.5x the previous level after just one quarter of use
  • Miaoda (Vibe Coding Platform): With the launch of 3.0 last qtr, users can now generate standalone mobile apps for Android and iOS using natural language
    • MAUs incr’d +67% June vs March; Adoption is expanding across technology, education, healthcare, manufacturing, financial services, and logistics
    • Users are moving beyond one-off experimentation to continued development
  • Famouu Agent 2.0: This enterprise AI agent is solving complex operational problems
  • DuMate (General Purpose Agent): In Q2 introduced enterprise version and expanded proprietary skills and specialized toolkits
  • While new products and engagement is gaining traction, monetization has not ramped yet as AI Applications rev at CNY2.50bn was up only 3% y/y in Q2 (inline w/ cons)

• Driverless rides volumes were negatively impacted in Q2 but it is re-gaining momentum
  • Apollo Go ride volume were down materially vs Q1…it conducted ~1mn fully driverless rides in Q2 vs 3.2mn in Q1
    • Why the decline in volumes? Ride volume was temporarily affected by operational adjustments in certain domestic cities due to regulatory considerations but operations in affected cities have begun to resume in Aug and mgmt is “confident that ride volume will regain momentum over the coming quarters”
    • Cumulative rides: Reached 23mn+
    • Global footprint: Now reaches 28 cities (up 1 from Q1’s disclosure)
    • Accumulated autonomous kms by the fleet: 350mn+ (240mn+ fully driverless) vs 330+mn (220mn+ fully driverless) in Q1
  • International is a big oppty: The market outside of the US and China is “larger than the domestic China market”…the addressable opportunity is “quite substantial”
    • And in overseas markets with higher ride prices, Baidu’s low-cost vehicles and proven operating model have the potential to deliver even stronger unit economics than domestic markets
  • International expansion highlights:
    • HK: Received HK’s first permits for fully driverless testing in June; Began testing on Airport Island in July
    • Dubai: Launched fully driverless commercial ops in July; They are now operating at the largest scale among fully driverless robotaxi services in the city; Rides are available through both Apollo Go and Uber apps
    • London: Apollo Go began open-road testing in partnership with Uber and Lyft
    • Switzerland: Conducting open road testing in partnership with PostBus
    • Kazakhstan: Signed MOU with Turlov Private Holding to jointly explore autonomous ride-hailing services
  • “Industry-leading safety performance”: Fully driverless vehicles averaged ~1 airbag deployment per 14.4mn km
  • H2 priorities: further enhance safety, advance global expansion, scale fleet/ride volumes, bring more cities to unit economics breakeven

-> Also related to autonomous cars, Uber, Verne, and Pony AI launched autonomous rides on the Uber app in Zagreb (the capital of Croatia), marking the 1st time European riders can book an AV through Uber; Service will cover key areas including the city center and coverage will expand over time; A licensed operator will initially monitor behind the wheel before transitioning to fully driverless operations (link)

There were two media entertainment updates that we thought were worth flagging this week. First, a wave of CNBC executive interviews included some interesting perspectives on the future of cable/tv, streaming, sports, and more.

Also, Disney’s D23 event (the largest show floor in its history) added more color around its full content (theatrical, streaming, and gaming) and experiences slate and disclosed some early KPIs on the Co’s Verts short form product (which were surprise surprise, positive), and a Formula 1 partnership expansion was also announced, among other updates.

See below for more of a drill down.

Industry Executives Opine On Key Trends In TV, Streaming, Sports, & More (Per CNBC Interviews)

• Who participated in CNBC’s interviews?
  • Chris Winfrey (Charter CEO)
  • Jeff Zucker (RedBird IMI CEO)
  • Charlie Collier (Roku Media President)
  • Rashida Jones (Uncensored CEO)
  • Jimmy Pitaro (ESPN Chairman)
  • Anjali Sud (Tubi CEO)
  • John Landgraf (FX Chairman)
  • Jeffrey Hirsch (Starz CEO)
  • Brian Fuhrer (Nielsen SVP of product strategy and thought leadership)
  • Debra OConnell (Disney Entertainment TV Chairman)

• In Three Years, Will We Have Hit A Floor On Cable TV Subscribers? The consensus is that subscriber declines will continues, driven by rising retransmission costs and streaming bundle substitution, though the pace has been slower than previously forecasted
  • Winfrey:
    • Expects dramatic declines with cable TV subscribers, noting retransmission costs now exceed $30/customer for content that’s “essentially free”; Sees streaming bundles replacing traditional cable bundles
    • “Netflix is essentially a big cable programmer that could end up being bundled together with the other streaming apps to provide more choice, more value, and more utility for customers over time”
  • Zucker:
    • “I don’t think we’ll have a floor. I do think it will continue to decline, and it’ll probably do so every year until sports rights eventually disappear from cable. But, I think that’s at least a decade off”
  • Collier:
    • “Nothing truly goes to zero” but the direction is clear
  • Jones:
    • “We haven’t been right as an industry to date. It hasn’t been quite as fast and as aggressive as maybe we were forecasting even four or five years ago”

• What’s One Thing That Will Become A TV Industry Standard That Doesn’t Exist Today? Personalization, frictionless commerce, AI-powered language toggling, and immersive sports formats were the most-cited bets on what becomes mainstream
  • Pitaro:
    • Ubiquitous personalization where networks deliver the right content to the right user at the right time; Content will be promoted and content will be created and tailored based on a user’s preferences
    • He also flagged seeing a lot more frictionless commerce integration w/ deep-linking to partner sites to make a purchase
  • Sud:
    • There will be hyper-personalized ads that feel useful rather than interruptive, mirroring the social media ad experience which she believes has less friction
  • Landgraf:
    • Day-and-date global releases will become the norm for the biggest TV shows
  • Zucker:
    • More podcasters and livestreamers will be licensed to cable/broadcast networks at scale
  • Hirsch:
    • Reiterating what he said 3 years ago, which was that there will be major advances with AI enabling consumers to watch content in their native languages
    • Subbing and dubbing will go away and consumers can select whatever language they want
  • Winfrey:
    • Immersive sports programming via 8K will be more pervasive, citing Charter’s Spectrum Front Row partnership w/ NBA and Apple Vision Pro as a proof of concept…”I think that’ll be a unique and new experience”

• Will There Be A Major Gov’t Action To Prevent A Big Tech Co From Getting Bigger In The Entertainment Industry? Most respondents expect growing scrutiny but doubt the genie goes back in the bottle as the tech-entertainment convergence is already too far along
  • Zucker:
    • “Big Tech companies will be in for a lot of scrutiny…but that depends on what happens in the next two elections, in ’26 and ’28”
  • Jones:
    • We will see “consumer and an industry backlash to some of these deals that will then kind of change the course”
  • Sud:
    • The convergence between Silicon Valley tech and Hollywood, it’s already happened; Tech platforms are taking the highest share of time spent and attention (like YT); Instagram is now doing vertical videos on TV, the Oscars are going to be on YouTube, you have creators putting movies out in theaters in the box office
    • I don’t think you can put the genie back in the bottle

• Are We In A Sports Viewership Ratings Bubble? There was strong consensus that live sports remain structurally durable, though some caution that measurement improvements are inflating comparisons and saturation may eventually cap growth
  • Pitaro:
    • The answer is a fast “no”
    • Ratings measurement is getting better
    • “Every time the industry expects a downturn, numbers continue up and to the right”
  • Jones:
    • “This is a slightly adjacent industry for me — but I think every industry has a ceiling and available audience”
    • “I think you get to a ceiling. You get to a peak, and I think we’ll start to see some of that drop down”
  • Brian Fuhrer
    • Methodological improvements, especially expanded out-of-home measurement, are driving increases which will continue, but the big y/y bumps will level off
  • Winfrey:
    • “I think you’re going to continue to see ratings go up. Maybe that’s a little bit of a mix of short-form content, long-form content, immersive capabilities of bringing you into the stadium”
    • “I think it’s going to continue to drive ratings inside the system”
  • Zucker:
    • “People want live events, they want live news, and they particularly want live sports”
    • “Whether ratings go up or down depends on matchups and championships, but overall I think sports are going to continue to be very, very strong”
  • Collier:
    • “Live sports remains one of the last true mass reach experiences in our culture”
    • “I suspect premium sports remain incredibly valuable, perhaps even more than today”

• There’s Broad Agreement That The Biggest Streaming Services Will Probably Still Be The Biggest In Three Years. What’s Another Service That You Think Will Gain Significant Market Share? There were some far ranging answers to this question…from FAST platforms (Tubi, Pluto, Roku Channel) to gaming (Epic Games) to a yet-to-emerge super-aggregator…but some skepticism that any unknown entrant breaks through in long-form video
  • Debra OConnell (Disney Entertainment TV Chairman):
    • It is likely a service that we’re not thinking of today…like Instagram’s new TV push
  • Winfrey:
    • There is a real opportunity to be a new aggregator that bundles services at a discount, which is what was cable’s original value proposition.
    • “I think that’s the opportunity for a new entrant into the space”
  • Zucker:
    • “We see TikTok videos getting longer and longer, and I think that’s a trend that will continue”
    • Niche casting is also going to continue
  • Fuhrer:
    • Platforms like Roku Channel, Tubi, Pluto will continue to gain share…that won’t slow down
  • Pitaro:
    • Epic Games will gain share
    • “The possibilities to connect gaming experiences with both entertainment and sports content, including potentially live events, are endless”
  • Landgraf:
    • “Do I really think that a streamer we’ve never heard of is going to be meaningfully competitive in long-form video three years from now? No, I don’t think so”
  • Hirsch:
    • “Starz is my answer”

• Link to full CNBC article

Disney’s D23 Event Provided Color Around Its Robust Multi-Year Content Slate + Some Incremental Updates

• Disney’s D23 event this week was reported to be its largest show floor yet in its history…there were a couple updates that we thought were worth flagging…(link/link/link/link/link/link/link)

• First, if you are wondering what the full content (theatrical, streaming & games) and experience slate that Disney has in mind over the next several years, look no further…see the list below:

• Other incremental reveals at the D23 Event
  • The Co Reiterated Its 13 Ship Expansion Plans + Provided Color On Its New Class Of Ships Debuting Starting In 2029: The new class of 3 ships sits in size between the classic Magic-class ships and the newer Dream- and Wish-class vessels, with the first two targeted for delivery in 2029 and the third in late 2030
    • The ships will be “bigger than the classic Magic-class ships (Disney Magic & Disney Wonder) and smaller than the Dream- and Wish-class vessels”
    • This “gives us some amazing opportunities to visit new places and reach brand new audiences we haven’t before”
  • Disney’s Verts Short Form Video Is Off To A Strong Start… the Co disclosed Verts KPIs:
    • On the ESPN App: Verts is now the primary driver of short form video consumption, w/ more than 24mn total users generating over 1.5bn minutes of time spent since the Aug 2025 launch
    • On Disney+: ~40% of mobile users have already interacted w/ Verts, and the Co has seen increases in avg min per visitor, clips per session, and watchlist interactions since the March launch
    • Plans for Verts looking ahead: The team is exploring ways to evolve Verts beyond content discovery and into creator communities, original storytelling, microcontent, and podcasts
  • Disney’s Formula 1 “Fuel the Magic” deal was extended & there are 2 new consumer events
    • Terms: The deal was extended through 2028
    • Mundo Pixar Experience annc’d its first ever US tour launching in ’27, after runs in the UK, Mexico, Brazil, Spain, Belgium, Seoul and Tokyo, w/ details to follow
    • Disney Theatrical Group will premiere a new stage adaptation of The Greatest Showman at Theatre Royal Drury Lane in London’s West End in spring ’27…this follows a sold-out world premiere preview engagement at The Bristol Hippodrome
  • The Co Also Plans To Invest More In Local International Content On Disney+: Over the next 3 yrs, Disney plans to roughly 3x the number of local original series on Disney+

Growing concern over the impact of social media on young users has become a global regulatory movement with 20+ countries having now enacted or proposed bans on minors using these platforms (and 40+ if you include those considering bans – link). This regulation was led by Australia (a ban has been in place since Dec 2025), followed by the UK, Canada, France, and others in rapid succession over the past year. In the US, that pressure is now escalating in the courtroom. Meta recently lost a landmark child safety case in New Mexico (the first time a US state successfully sued the Co over the issue) which is resulting in almost $1bn in penalties.  However, that amount would pale in comparison to potential penalties from a trial brought by a coalition of 29 states against Meta claiming the Co harmed teens’ mental health, which just kicked off this week in a CA federal court.

Concurrently, OpenAI is trying to take a more proactive approach after controversy on its platform as well. This week it launched a dedicated “ChatGPT for Teens” experience with built-in safety protections and parental controls.

See details below.

• The trial against Meta over claims that Facebook and Instagram were designed in ways that harmed young users’ mental health began on Tuesday in a California federal court (link/link)
  • A coalition of states are starting to make their case: Attorneys for Colorado, California, New Jersey and Kentucky are leading a bipartisan group of 29 states
    • All 29 states accuse Meta of improperly collecting and using children’s personal data in violation of federal law
  • What kind of damages are being sought? Meta has said the penalties could be as high as $1.4 trillion BUT lawyers representing the states reportedly said last week that $200bn is a more likely amount
  • Reminder that Meta lost a case recently in New Mexico that will require some service changes and a payment of almost $1bn in total

-> This week, META shares fell -6.6% while SNAP shares fell -3% in sympathy; YTD, META is down ~-17% and SNAP is down ~-35%

• Also on this theme this week, OpenAI launched “ChatGPT for Teens” (link/link)
  • What is it? A dedicated experience for users aged 13–17 that activates automatically based on age at signup or behavioral/account signals
  • Features of ChatGPT for Teens:
    • Study Mode: It guides students through problems w/ questions and step-by-step explanations rather than providing direct answers
    • Study Hours: This allows teens or parents to set Study Mode as the default during certain time periods
    • Stronger defaults on content safeguards: Regarding self-harm, violence, eating disorders, dangerous activities, and explicit sexual content
    • Other: The chatbot is barred from using language suggesting romantic feelings or claiming sentience/personal feelings
    • Parental controls: Linked adults can manage settings, set Quiet Hours, and receive safety notifications in high-risk situations, but cannot read or monitor their teen’s conversations
  • Timeline: It rolled out globally on free and paid personal plans starting Aug. 18 and full availability in Australia is expected by Sept. 8

We’ve been following the development of the drone delivery market for some time and while it still remains niche, momentum is building. Both Amazon and Uber announced notable expansions on this front this week. In some cases, this new form of transportation will take delivery to only minutes…. 

See details below.

 Amazon Prime Air Drone Delivery Is Expanding to ~500 US Cities by YE 2026

• Amazon is scaling its Prime Air drone delivery to nearly 500 cities/towns by YE 2026, an ~6x increase from its current footprint, reaching tens of millions of customers (link)
  • Current mkts: 11 US locations in 10 metro areas (each Prime Air site serves an area of ~175 sq miles)
    • Phoenix, Arizona
    • Tampa, Florida
    • Kansas City, Kansas
    • Baton Rouge, Louisiana
    • Detroit, Michigan
    • Omaha, Nebraska
    • Houston, Texas
    • Dallas, Texas
    • San Antonio, Texas
    • Waco, Texas
  • Planned new launches “soon”:
    • Chicago, Illinois
    • Syracuse, NY
    • Cleveland, Ohio
    • Atlanta, Georgia
    • Boise, Idaho

• Some key specifics about Amazon’s drone delivery service…
  • What is eligible for drone delivery? Nearly all items that weigh less than 5 lbs that fit in a large shoebox qualify; This covers 60%+ of the most frequently purchased items on Amazon and includes “millions” of items in groceries, electronics, cosmetics, medications, household products
  • How fast will it be? As fast as 30 min but most orders will arrive ~60 min after checkout
  • How many products have been delivered by drone? “Hundreds of thousands” of packages YTD
  • How much does it cost? Prime members get free drone delivery on orders $50+
    • It’s a $2.99 fee on orders less than $50 & non-Prime customers pay $4.99
  • How loud are the drones? The sound level during drop-off is below that of an idling delivery truck (takes ~30 sec to drop off)
    • During flight at altitude, it’s comparable to a window fan on low and is typically not heard indoors

• Amazon is confident in the safety aspect of its drone delivery
  • Prime Air holds FAA Part 135 certification which is the same framework used for commercial air carriers and the highest level of FAA oversight for drone delivery ops
  • The drones have an “industry-leading” Detect-and-Avoid system that allows drones to independently monitor airspace and make real-time decisions in flight
  • The drones are fully electric w/ zero exhaust emissions

• The Co is expanding drone svs internationally
  • The 1st drone delivery site outside the U.S. launched in Darlington, England

Source: link

Uber Is Partnering W/ Zipline To Launch Drone Deliveries On Uber Eats

• Key details on Uber/Zipline’s announcement… (link/link)
  • Markets/timing: Will launch in Dallas and Houston by year-end
  • Longer term targets: 1mn drone deliveries/day by the end of 2029
  • Uber is also investing in Zipline: But the amount is undisclosed (the Co was valued at ~$7.6bn earlier this year)
  • Deliver time? As fast as 5-10min but most orders should be delivered in less than 30min
  • This is a big step forward w/ Uber’s drone initiative
    • The Co previously tested drone delivery in 2019, sold that aviation arm, and has been running smaller programs with Flytrex and Manna
  • Competitive moves:
    • DoorDash recently received FAA air carrier certification for its own in-house drone program
    • Wonder (GrubHub parent) has its own Zipline partnership
    • Amazon and Walmart are active in drone delivery as well (see above for more on Amazon’s recent expansion)

OpenAI and Anthropic have been in a head-to-head battle as it relates to philosophy, technology, customers, and going public, and this battle escalated again this week. While OpenAI had a headstart in the marketplace, Anthropic’s quarterly revenue in Q2 surpassed OpenAI’s for the first time, more than doubling seq to $11.5bn+ while posting a small adj operating profit. In contrast, OpenAI’s Q2 revenue grew just +18% seq to $6.7bn while losses widened. Anthropic’s public IPO filing is expected as early as the end of this month, while Open AI’s CFO this week told employees that the Co will be public in 2027, and “possibly sooner.”

Meanwhile, public unease regarding AI is ticking up, especially with younger US adults, per a new Pew study.

See more details below.

Anthropic Surpasses OpenAI In Revenue For The First Time In Q2 As It Grows Sequentially Much Faster

• Anthropic’s annualized revenue run rate hit $65bn at end of July (confirmed by CNBC)… (link /link/ link/link/link)
  • How does that compare?
    • This is up from the reported ARR of $47bn in May
    • And is up 7x from a year ago
    • OpenAI reportedly recently reached $40bn ARR (up ~2x the end of 2025)
  • Q2 preliminary numbers (per press)
    • Revenue exceeded $11.5bn+, up from $4.73bn in Q1 (+140%+ seq) and vs $787mn in Q2 2025 (up ~14x y/y)
    • Adj op profit (excl SBC) was ~$559mn
    • See next section for OpenAI’s comparable numbers
  • Forward expectations: Investors expect Anthropic to finish 2026 between $100bn-$120bn in total revenue and the Co is reportedly projecting ~$190–200bn in 2028 revenue

-> Per Bloomberg this week, Anthropic is preparing to file publicly for its potential mega-IPO as soon as “the end of this month” (link)

• In contrast, OpenAI told investors that its Q2 revenue was $6.7bn, up +18% seq from $5.7bn in Q1…which was a pace that “disappointed some shareholders” who expected faster progress vs. Anthropic (link)
  • Its Q2 operating loss (incl. SBC) widened to $12.3bn from $9.3bn in Q1
  • OpenAI reportedly told investors that Q3 growth has accelerated, partly driven by the GPT-5.6 launch and subsequent price cuts

-> OpenAI CFO Sarah Friar also told employees this week at an all-hands that the Co “will be a public company in 2027,” or possibly sooner if “our business continues to inflect” (link)

All This At The Same Time That Public Skepticism Of AI Deepens

• 52% of Americans now say that they are “more concerned than excited” about the incr’d use of AI in daily life…that is up materially from 37% in 2021, per a new Pew Research study (link)
  • The growing concern is even more pronounced with young adults (see chart below)

Profitability remains a defining theme across the streaming landscape with price hikes as a key driver…and another one is on the way as Peacock is raising pricing for all of its tiers (this is the 4th hike in four years). At the same time, they are also looking to layer on a loyalty program (in beta) to offset churn risk.  Meanwhile Paramount+ is now testing a free ad-supported tier as an alternative to traditional trials and in the background, streaming hit 48.5% of total TV viewing in June and nearly a third of U.S. homes now have no linear TV access at all (per new data from Nielsen and ARF).

See more color below.

More Rate Hikes Hit The Streaming Industry W/ Peacock Next On Deck, Which Also Comes With A New Loyalty Program As Well

• Peacock raised prices across all tiers effective this Aug 18^th …this is its 4th price hike in four years w/ the last one in July 2025 (link)
  • Price changes:
    • Select: $8.99/mo (from $7.99)
    • Premium: $12.99/mo (from $10.99)
    • Premium Plus: $19.99/mo (from $16.99)
  • Annual plans also up:
    • Select: $89.99/yr (from $79.99)
    • Premium: $129.99/yr (from $109.99)
    • Premium Plus: $199.99/yr (from $169.99)
  • New vs existing subscribers
    • The new pricing applies immediately to new subs
    • Existing subs transition on the next billing date on or after Sept. 17^th, and promotional pricing is honored through promo period

• Separately, Peacock launched a beta for a new Membership Program, available to select Premium and Premium Plus subs (link)
  • There will be three tiers based on how long they have been subscribers:
    • Gold (3 mos–1 yr)
    • Platinum (1–3 yrs)
    • Diamond (3+ yrs)
  • What are the rewards while in beta?
    • NBCU Shop discounts (25%)
    • Tiered Bravo Shop discounts
    • Complimentary 1-yr Instacart+ membership
    • A Pizza Hut promo (unlocked after streaming 3 movies)
  • Other potential rewards include:
    • Universal Parks perks
    • NBCU content previews
    • Early access to live events

Paramount+ Tests a New Free Tier

• Paramount+ began testing a limited free tier on Android devices…users can browse the app and stream up to 2 episodes from select programs w/o a paid subscription; The free tier is ad-supported and does not replace existing paid plans (link/link)
  • What about iOS devices? There is no confirmed timeline for this feature on iOS or other platforms
  • This is part of a broader product experimentation push at Paramount+: That includes short-form video, micro dramas, vertical video podcast segments, and expanded multiview for live events

Streaming Nears 50% Of Total TV Viewing & Nearly a Third of U.S. Homes Are Digital-Only, Per New Reports This Week

• Streaming share rose y/y to 48.5% of total TV viewing per Nielsen’s June 2026 Gauge (link)
  • Broadcast share rose to 19.8% from 18.5% last yr period
    • Boosted by FIFA World Cup on Fox/Telemundo and ABC’s NBA Finals
  • Cable share fell to 19.5% from 23.4% last yr period
  • Among streaming platforms…all gained shr except for Netflix
    • YouTube – gained share to 13.5% from 12.8% last yr period
    • Prime Video – gained share to 4.2% from 3.6% last yr period
    • Roku Channel – gained share to 3.0% from 2.5% last yr period
    • Peacock – gained share to 2.3% from 1.5% last yr period
    • Netflix – lost share to 7.9% (from 8.3%)
  • But note that some analysts expect the recalibrated Gauge (using ARF DASH-based universe estimate this fall) to show streaming w/ somewhat less dominance vs. broadcast/cable than recent reports suggest

Source: Nielsen

• 30.5% of U.S. households now get TV content only through streaming, w/ no access to linear TV…this is up +1ppt from 29.5% for the full-year 2025, per ARF DASH midyear 2026 data (link)
  • Antenna-only homes also rose +1ppt to 17.1% over the same period
  • Linear and pay-TV penetration cont’d to decline, though the pace of decline appeared to slow

• More push-back on data centers…Pennsylvania Governor Shapiro signs an executive order implementing new restrictions on data center developers (link)
  • What does the executive order say? It directs the Dept of Environmental Protection to only review permit applications if developers have made a legally binding commitment to meet the Governor’s Responsible Infrastructure Development (GRID) Requirements and have received local approval
    • What are GRID requirements? They demand that developers meet strict standards on energy affordability, environmental protection, workforce and economic development, transparency, and community engagement
  • The EO also removes all AI data center proposals from the Fast Track permitting process and prohibits the use of nondisclosure agreements for data center projects

• TikTok is moving into P2P payments….the Co is apparently developing a feature that would enable users to send money to each other through direct messages (link)
  • How would it work? The feature would use TikTok Pay, which is already operating in Southeast Asia, to process transactions
  • How did this leak? Reportedly code was discovered in the current version of TikTok’s US iPhone app that indicated these plans
  • Other potential new entrants: X (formerly Twitter) has also moved into financial services under Elon Musk’s leadership; Some creators on X have tested new features such as “X Money” to send payments to one another

-> Shares of U.S. money transfer PayPal and Western Union were relatively flat after the news was reported

• Walmart added new in-store perks to its Walmart+ subscription program (link)
  • Photo Center benefit in Walmart stores
    • Members get 25 free 4×6 photo prints/mo at photo centers in Walmart stores
    • Maximum of 300 prints/yr
  • Money Center benefit at Walmart Money Centers: Quarterly perk includes fee-free:
    • Check cashing
    • Money orders
    • Money transfer sends
    • Bill payments
    • eCash loading
  • For background…Walmart+ was launched in 2020 and mgmt has previously said that Walmart+ members…
    • Generally spend 4x more than nonmembers overall
    • Have 7x more e-commerce visits each year