LT Weekly: Mar 21, 2025

All eyes were on NVIDIA’s GTC event this week, and certainly the bar was certainly high. There were several standout updates and announcements from CEO Jensen Huang’s keynote (this was particularly impressive, as it was all on the fly without a teleprompter or notes!), which started with the notion that the industry vastly (understatement) underestimated the amount of compute power it needed. More specifically, Huang underscored that “this last year, this is where almost the entire world got it wrong… the amount of computation we need at this point as a result of agentic AI, as a result of reasoning, is easily 100 times more than we thought we needed this time last year.” The next step of AI will be enabled by improved training and knowledge work as well as teaching models to make AI models smarter and more efficient. In that context, Huang unveiled Blackwell Ultra, which is the next evolution of the NVIDIA Blackwell AI factory platform. It promises a +40x AI performance improvement and will be available in H2 of this year. Furthermore, the company’s chip roadmap includes its next generation GPU Vera Rubin coming in H2:26, followed by Vera Rubin Ultra’s expected launch in H2:27.

Also notable was Nvidia’s push into Physical AI with the introduction of GROOT N1, the world’s first open humanoid robot foundation model that brings generalized skills and reasoning to humanoid robots. Huang notes that this is important because “by the end of this decade, the world is going to be at least 50mn workers short,” hence capable humanoids will become essential (separately but related, there were a couple humanoid robot announcements out of China as well this week, as the country aims to fiercely compete in this tech area as well). Lastly, and also on the theme of Physical AI, NVIDIA is pushing harder into autonomous vehicles with the expanded partnership with General Motors. Importantly, both announcements further extend NVIDIA’s reach beyond the core datacenter business, positioning the company to capture more value throughout the AI stack across several industries.

See below for more of what captured our attention at this jam-packed event.

-> NVIDIA shares fell -3.3% this week and are down -12.4% YTD

-> If you are as interested in developments in the humanoid robotics area as we are, you might like this recent ~5min video by Boston Dynamics that showcases current capabilities: VIDEO LINK

Blackwell Ultra Is “Paving The Way For The Age Of AI Reasoning” (link /link)

• NVIDIA unveiled Blackwell Ultra, which is the next evolution of the NVIDIA Blackwell AI factory platform: Within Dynamo (more on that below), it promises a +40x AI performance improvement; This performance leap is particularly crucial for inference workloads, which Huang said is “one of the most important workloads in the next decade as we scale out AI”
  • Blackwell Ultra boosts training and test-time scaling inference, resulting in more compute during inference to improve accuracy, which is crucial for applications like AI reasoning, agentic AI & physical AI
    • The platform includes the GB300 NVL72 rack-scale solution & HGX B300 NVL16 system (see below)
    • “It’s a single versatile platform that can easily and efficiently do pretraining, post-training and reasoning AI inference”
  • GB300 NVL72: Enables AI models to access incr’d compute capacity, breaking down complex requests into multiple steps for higher-quality responses
    • It delivers +1.5x more AI performance than its predecessor and increases Blackwell’s revenue oppty by +50x for AI factories, vs with those built w/ NVIDIA Hopper
  • HGX B300 NVL16: This system offers 11x faster inference on LLMs, 7x more compute, and 4x larger memory vs previous Hopper generation; It’s designed for complex workloads like AI reasoning
  • Ideal applications: Agentic AI and Physical AI
  • Availability: Blackwell demand is in “full production” (following earlier comments about a “hiccup” in early production) and Blackwell Ultra-based products are expected to be available from partners starting from H2:25

Software Innovations Set To Reduce AI Bottlenecks… Dynamo Is The “Operating System Of An AI Factory” (link/link)

• Annc’d its new open-source Dynamo Inference software which will accelerate & scale AI reasoning models in AI factories w/ “leaps in throughput, faster response time and lower total cost of ownership”
  • This new AI inference-serving software is designed to maximize token revenue generation for AI factories deploying reasoning AI models
  • “Using the same # of GPUs, Dynamo doubles the performance and revenue of AI factories serving Llama models on today’s NVIDIA Hopper platform” and “when running the DeepSeek-R1 model on a large cluster of GB200 NVL72 racks, NVIDIA Dynamo’s intelligent inference optimizations also boost the number of tokens generated by over +30x per GPU”

-> Huang described Dynamo as “essentially the operating system of an AI factory

NVIDIA’s Roadmap For AI Computing Infrastructure Through 2027 (link)

• Blackwell Ultra NVL72 is coming in H2:25
• This will be followed by Vera Rubin, its next-gen AI chip (NVL144), which will be available in H2:26
  • Rubin will feature a new CPU that’s 2x as fast as the current Grace CPU, along with new networking architecture and memory systems

• It also unveiled Vera Rubin Ultra (NVL576) which is expected to be available in H2:27
  • This is an “extreme scale up” offering +14x more computational power than current systems; “You can see that Rubin is going to drive the cost down tremendously,” Huang noted, addressing concerns about the economics of AI infrastructure

Nvidia Is Super Charging Humanoid Robot Development (link)

• Unveiled Isaac GR00T N1, the world’s first open humanoid robot foundation model that brings generalized skills & reasoning to humanoid robots:
  • This model is based on a dual-system architecture inspired by human cognition:
    • “System 1” is a fast-thinking action model, mirroring human reflexes or intuition – it is trained on human demonstration data and massive amounts of synthetic data
    • “System 2” is a slow-thinking model for deliberate, methodical decision-making
  • What can GROOT do? Common tasks like grasping, moving objects with one or both arms, and transferring items from one arm to another – or perform multistep tasks that require long context and combinations of general skills
  • Highlighted applications: Material handling, packaging & inspection
  • Key partnership: The robot’s autonomous capabilities are the result of an AI training collaboration between 1X and NVIDIA
  • Which Cos have early access to GR00T N1? Agility Robotics, Boston Dynamics, Mentee Robotics and NEURA Robotics

• Introduced Newton, an open-source physics engine (available later this year) being developed in collaboration w/ Google DeepMind & Disney Research designed for robotic movements: This engine will help create more realistic and efficient robotic actions
  • Disney will be one of the first to use Newton to advance its robotic character platform that powers next-generation entertainment robots, such as the expressive Star Wars-inspired BDX droids

• Intro’d Isaac GROOT Blueprint for generating synthetic data: This lets developers generate exponentially large amounts of synthetic motion data for manipulation tasks from a small number of human demonstrations
  • NVIDIA generated 780k synthetic trajectories, the equivalent of 6,500 hrs, or 9 continuous months, of human demonstration data, in only 11 hrs; Combining the synthetic data w/ real data improved GR00T N1’s performance by +40% vs using only real data

-> “With NVIDIA Isaac GR00T N1 and new data-generation and robot-learning frameworks, robotics developers everywhere will open the next frontier in the age of AI”; This is important b/c “by the end of this decade, the world is going to be at least 50 million workers short,”

• China vs The US – Separately but related, there were a couple humanoid robot announcements by Chinese Cos this week, which also made a splash:
  • Chinese robotics startup Dobot annc’d that it has begun accepting preorders for its AI -powered humanoid robot (Dobot Atom), which is capable of dexterous manipulation and straight-leg walking for ~US$27k, w/ mass production reportedly planned for mid-2025; The robot can complete tasks such as preparing breakfast, pouring milk, toasting bread and carrying a box
  • China’s home appliance conglomerate Midea Group, which has acquired a German robotics firm, also unveiled a prototype of its humanoid robot capable of performing actions such as shaking hands, dancing, opening bottle caps, and driving screws, per media reports

NVIDIA Makes A Big Push Into Autonomous Vehicles Via A Partnership With General Motors (link)

• The partnership will enable GM to “build their future self-driving car fleet” – “The time for autonomous vehicles has arrived”
• Partnership covers 3 areas: AI for manufacturing, AI for enterprise, and AI for in the car
  • GM has been investing in NVIDIA GPU platforms for training AI models across various areas, including simulation and validation but this collaboration expands scope inyo transforming automotive plant design and operations
• NVIDIA also annc’d Halos, “a comprehensive safety system” for AVs

NVIDIA Partners W/ Mobile Players To Drive AI Native W-less Networks (link)

• The Co annc’d partnerships w/ T-Mobile, MITRE, Cisco, ODC, a portfolio Co of Cerberus Capital Mgmt, and Booz Allen Hamilton for the R&D of AI-native wireless network hardware, software, and architecture for 6G
  • AI will need to be fundamentally integrated into wireless networks to seamlessly connect hundreds of billions of phones, sensors, cameras, robots and AVs
  • What will AI enable? Enhanced svs for billions of users, set new standards in spectral efficiency and will enable “groundbreaking” performance and resource utilization while creating new revenue streams for telecom companies

Other Key Partnerships That Were Annc’d…

• IBM annc’d new collaborations with NVIDIA to accelerate AI at scale (link)
• Oracle and NVIDIA collaborate to help enterprises accelerate agentic AI inference (link)
• NVIDIA, Alphabet and Google collaborate on the future of agentic and physical AI (link)
• CrowdStrike collaborates with NVIDIA to advance Agentic AI in cybersecurity (link)
• NVIDIA and GE HealthCare collaborate to advance the development of autonomous diagnostic imaging with physical AI (link)
• NVIDIA to build accelerated quantum computing research center in Boston (link)

Source

After a failed attempt to purchase Wiz last summer for a reported $23bn (link) due to concerns related to antitrust matters, Wiz’s autonomy under Google Cloud, and the valuation (despite being well above Wiz’s $12bn valuation on a $1bn funding round earlier last year), Google persisted and finalized an agreement at a $32bn price tag this week. This would represent Google’s second largest acquisition behind its purchase of Motorola Mobility in 2011 for $12.5bn.

For context, Wiz is an end-to-end cloud native security company that serves both large and small customers big in an multi-cloud fashion (a feature that was highlighted numerous times by Google’s management on its webcast). Wiz will operate independently, though as a part of Google Cloud. The cybersecurity company reportedly is on track to reach $1bn in annual recurring revenue (ARR) in 2025, or double 2024’s level. This would also imply that Google is paying over ~+30x expected 2025 ARR, which is certainly a healthy multiple, illustrating the strategic importance of this transaction.

What is the rationale? Google talked about how AI is changing the security landscape and how companies need improved cloud security that spans multiple clouds. The acquisition of Wiz will help bolster the Co’s capabilities in this regard, while Google Cloud brings “the infrastructure, deep expertise, and a track record of security innovation.”

Regulatory approval remains the main wildcard. The fact that Wiz is multi-cloud, and will continue to operate as such, could help with the review process, but on the flip side and on an operational level, other cloud players may be less inclined to do business with Wiz, given it will be a sub-entity of a competitor, namely Google Cloud.

See below for more color on what we thought were the most important details regarding the transaction.

-> Alphabet shares fell -2.3% in reaction to the deal announcement and closed the week down -0.8%; YTD, Google stock is trading down -12.7%

• Deal basics –
  • All cash $32bn deal
  • Timing – Expected to close in 2026
  • Break-up fee – $3.2bn (link)
  • Wiz will operate independently as part of Google Cloud segment post deal close

• Financials?
  • Google did not disclose any financials about Wiz, but TechCrunch reported the Co currently has $700mn in ARR, w/ a target of $1bn in ARR this year (double 2024)
  • Valuation – This implies Google is paying over ~+30x the expected 2025 valuation for Wiz

• Wiz products/services:
  • Serves small to large Cos in public and private sectors
  • Connects to all major clouds and code environments (and will continue to operate in that fashion)

• Why the purchase? “The incr’dd role of AI, and adoption of cloud services, have dramatically changed the security landscape for customers, making cybersecurity increasingly important in defending against emergent risks and protecting national security”

• The combined Co can accel the ability of organizations to improved their security, lower the cost to do so, and spur adoption of multi-cloud and cloud computing

• The combined Co will offer a unified security platform across –
  • Threat Intelligence
  • AI threat Protection
  • Cyber Team Support
  • Measurable Defense

The China Tech sector has been on a hot streak as of late; however, this week saw a modest retrenchment within the space (KWEB has gained +22% since the beginning of the year but was down -3% this week). This dynamic was somewhat reflected in the market’s mixed reactions to Tencent and Pinduoduo’s (PDD) Q4 earnings results. Interestingly, Tencent’s stock price ended the week in negative territory after it printed better than anticipated headline numbers, while PDD’s share price rallied after it reported a broad deceleration in growth across all key financial metrics. Speaking to the former, Tencent’s business was buoyed by strength in its gaming business in Q4, but a significant step-up in CapEx during the quarter as well as plans to continue spending at a high level throughout 2025 to acquire GPUs for internal AI development and its Cloud Services business resulted in some squeamishness amongst investors. For PDD’s part, although the Co was challenged by fierce competition both domestically and in its global business in Q4, its long-term initiatives are starting to pay dividends and profitability still surprised to the upside, despite the difficult operating environment.

See below for more details on what we thought were the key points from Tencent and PDD’s Q4 earnings:

-> Tencent shares were up a slight +0.1% in reaction to earnings but ended the week down -3.4%; Tencent Music shares jumped +15.5% post-earnings, finishing the week up +11.6%.; YTD, Tencent stock is still trading up +23.6%, and Tencent Music stock is up +23.3%

-> Pinduoduo shares were up +4.0% in response to earnings, ending the week up +3.3%; YTD, Pinduoduo stock is trading up +30.6%

1. Tencent – See Below For Our Key Thoughts, Highlights, & Takeaways

A Robust Gaming Biz Drove Tencent’s Better Than Anticipated Financial Results

• Tencent’s headline results beat on all fronts: Total rev grew +11.1% y/y in Q4 (vs +8.1% y/y in Q3) and topped cons by +2.1%; Non-IFRS op profit was up +21.0% y/y (vs +18.6% y/y in Q3) and closed +3.5% ahead of cons; Non-IFRS net income incr’d +29.6% y/y (vs +33.2% y/y in Q3) and beat cons by +4.8%
  • Value-Added Svs (~46% of total rev) – BEAT: Q4 rev rose +14.4% y/y (vs +9.3% y/y in Q3) and exceeded cons by +2.4%; Gross margin incr’d +2ppts y/y to 56% (vs 57% in Q3), given a higher mix of domestic games rev, improvements in music subscription rev, and a lower mix of live streaming rev
    • Social Networks (~38% of VAS rev) – BEAT: Rev incr’d +5.7% y/y in Q4 (vs +4.0% y/y in Q3) and ended +3.8% above cons; Growth was driven by higher rev from app-based game item sales, music subscriptions, and Mini Games platform svs fees
  • Marketing Svs (~20% of total rev) – BEAT: Q4 rev was up +17.5% y/y (vs +16.6% y/y in Q3) and beat cons by +3.7%; Gross margin expanded +1ppt y/y to 58% (vs 53% in Q3) due to growth in high margin Video Accounts and Weixin Search rev
  • FinTech & Biz Svs (~33% of total rev) – BEAT: Rev grew +3.2% y/y in Q4 (vs +2.0% y/y in Q3) and finished a slight +0.3% ahead of cons; Gross margin incr’d +3ppts y/y to 47%, driven by a greater mix of high margin rev as well as improved cost efficiency in Cloud Svs

Domestic & International Games Received Strong Contributions From Both New & Evergreen Titles

• Domestic Games rev growth improved seq, though off weak comps: Q4 rev was up +23% y/y (vs +14% y/y in Q3) “against a low base qtr”; Cited contributions from evergreen games, including Honour of Kings, Peacekeeper Elite, and VALORANT, as well as from recently released games such as DnF Mobile and Delta Force
  • Several evergreen titles benefited from IP collabs and high-profile events: Resulting in higher DAUs and monetization
    • Honour of Kings incr’d gross receipts by a double-digit percentage y/y: Driven by higher DAUs as well as popular outfits based on the Detective Conan anime series
    • VALORANT’s gross receipts more than doubled y/y: Benefited from tie-ins w/ the World Championship winning team, Edward Gaming, and w/ Riot’s animated series Arcane Season 2
    • Fight of the Golden Spatula saw a double-digit y/y increase in gross receipts: Arcane-themed champions was a notable driver
  • The Co has been “seeking to nurture addt’l evergreen games” –
    • Delta Force generated RMB1bn+ of gross receipts from PC and mobile in China
    • Tencent also has a pipeline of “highly anticipated” titles: Including The Hidden Ones, Light of Motiram, GODDESS OF VICTORY: NEW HOPE, and VALORANT Mobile

• International Games rev growth also accel’d seq: Q4 rev grew +15% y/y (vs +9% y/y), or +16% y/y ex-FX; The Co saw “robust performances” from Brawl Stars, PUBG Mobile, and the early access release of Path of Exile 2
  • Brawl Stars’ gross receipts grew “several times y/y”: The game benefited from the Angels vs Demons season and redesigned battle pass; Across 2024, Brawl Stars was the third-highest mobile game by DAU outside of China
  • Path of Exile 2 ranked first among premium games by rev on Steam: Over the six weeks following the title’s early release in Dec
  • Warframe: 1999 achieved the highest level of Q4 gross receipts in its 11-yr history: The game received a major update during the qtr, boosting DAUs; Warframe’s gross receipts rose over +30% y/y in 2024

• Tencent highlighted three observable facts about its game biz –
  • The Co exited 2024 w/ deferred rev up ~+high-teens% y/y: This was mostly driven by the games biz and will flow through into reported rev throughout 2025 and impact some of 2026 as well
  • User behavior during Chinese New Year demonstrated “the health of the game industry”: During the 2025 Spring Festival, the Co’s five highest grossing games saw a y/y increase in DAUs, supported by gameplay and social activity initiatives
  • Games have benefited both directly and indirectly from AI tech improvements –
    • The direct benefit comes from game devs using AI to create content more quickly and serve users more effectively
    • The indirect benefit is that higher usage of AI will cause more people to want to express themselves through “high agency activities,” such as interactive entertainment, as opposed to more passive means of expression

FinTech & Business Svs Saw Accel’ing Growth… BUT Supply Constraints Limited Cloud Svs Rev Growth

• FinTech Svs rev “resumed ~+lsd% y/y growth” in Q4 (vs “largely stable y/y” growth in Q3): Driven by improved commercial payment volumes as well as increases in wealth mgmt and customer loan svs revs
  • Commercial payments rev improved seq to being “largely stable y/y”: Indicated the number of transactions in the segment “grew faster y/y”

• Biz Svs “grew modestly y/y” in Q4 (also saw y/y growth in Q3): The Co generated higher Cloud Svs rev as well as incr’d tech svs fees stemming from rising e-commerce transaction volumes
  • Cloud svs rev growth was constrained by supply issues: Tencent incr’d allocation of GPUs for internal use cases, including AdTech, foundational model training, and AI inference for Yuanbao as well as Weixin; This limited the provision of GPUs to the Co’s external clients
    • Increasing purchases of GPUs in Q4 was responsible for the “big step-up” in CapEx: The deployment of these GPUs is expected to accel the rev growth of Cloud Svs “over time”; In the meantime, Tencent has prioritized available GPUs toward high value use cases and clients
  • The Cloud Svs biz has growth vectors across the stack –
    • IaaS svs demand “is actually very strong”: The Co will be able capture more of this demand once it has more GPUs
    • In SaaS, Tencent Meeting and Tencent Docs doubled rev on a y/y basis: These are now the Co’s largest SaaS products in terms of rev
    • The Co has also been selling cybersecurity and audio/video communications PaaS software to its cloud clients
    • Yuanbao has seen strong growth momentum on the consumer-facing side: The chatbot takes a multi-model approach, leveraging Tencent’s HunYuan LLM as well as DeepSeek’s R1 model

The Marketing Svs Biz “Clearly Outpaced The Industry” In Q4

• Marketing Svs rev incr’d across most industries in Q4: Highlighted e-commerce, financial svs, fast-moving consumer goods, games, local svs, education, and healthcare as some of the categories where the Co saw y/y growth in advertising rev
  • Video Accounts rev growth was also similar seq: Video Accounts rev grew over +60% y/y in Q4 (~flat w/ Q3’s growth), driven by higher user engagement, AI enhancements, and incr’d consumer transactions within Video Accounts, resulting in more closed-loop, e-commerce advertisements
  • Mini Programs rev “incr’d rapidly y/y”
  • Weixin Search rev more than doubled y/y: Benefited from commercial queries as well as AI-optimized ad placements and ad formats, which boosted click-through rates

• The macroenvironment for advertisers remained “challenging” through the end of 2024: Still, the Co “feel[s] quite comfortable” w/ its advertising biz, as AI enhancements have driven “fairly organic growth” across “pretty much every industry”
  • The Co has cont’d to add LLM capabilities to improve its ad biz: These have facilitated more efficient approvals of ad content, led to more precise ad targeting, and have streamlined the ad creative process
  • More AI features will be added to Weixin moving forward: Tencent has started testing AI features in Weixin to enhance user experience such as for search, language input, and content generation

Addt’l Color On Tencent’s AI Initiatives

• Tencent sees AI as a “growth multiplier” both for its existing bizs and as a “new growth driver” –
  • The Co continues to invest heavily in its HunYuan Foundation Model: Views the model as “an important technology foundation for our consumer and enterprise-facing bizs” and in Feb, released the HunYuan T1 deep-thinking model, which performs comparably to other top-tier reasoning models
    • Touted HunYuan’s content generation capabilities: Highlighted that HunYuan’s image generation models achieved the highest score from FlagEval last Dec and that the model’s video output quality ranked first on Hugging Face; HunYuan has also pioneered text and image-to-3D generation
  • A multi-modal approach to consumer-facing AI products will provide the best AI experience: This strategy leverages all available models to serve different user needs; “A combo of various models can handle complex tasks better than a single model”
  • On the enterprise front, Tencent has been accel’ing AI integrations across its cloud stack:
    • Using its IaaS solutions customers can achieve higher performance AI training and inference at scale
    • For PaaS, the Co’s TI platform supports model fine-tuning and inference demands w/ flexibility
    • Tencent’s SaaS products are also increasingly benefiting from AI-powered tools, such as real-time transcription and meeting summarization functions

-> Tencent Music also released its Q4 results this week and reported headline numbers that broadly surpassed consensus estimates; The Co’s total rev rose +2.3% y/y in Q4 (vs +6.8% y/y in Q3) and beat cons by +2.2%; Gross margin of 43.6% (vs 42.6% in Q3) also exceeded cons’ 43.1%, and non-IFRS EPADS closed +20.2% ahead of cons; Online Music Svs rev was up +18 y/y in Q4 (vs +20% y/y in Q3), topping cons by +1.2% and w/ growth in music subscriptions rev benefitting from a stronger than anticipated ARPU; Notably, elevated audio effects and other enhancements drove seq growth in SVIP membership and boosted ARPU growth; Social Entertainment Svs rev fell -13.0% y/y (vs -23.9% y/y in Q3) and beat cons by +5.6%; Declines were due to adjustments to certain live streaming functions and more stringent compliance procedures

2. Pinduoduo – See Below For Our Key Thoughts, Highlights, & Takeaways
PDD Saw Decel’ing Growth Across All Financial Metrics

• Headline numbers were mixed, as top-line growth underwhelmed but profitability surprised to the upside –
  • Total rev – MISS: Grew +24.4% y/y in Q4 (vs +44.3% y/y in Q3) and missed cons by -2.6%
    • Online Mkting Svs + Others rev – BEAT: Rose +17.1% y/y in Q4 (vs +24.3% y/y in Q3) and topped cons by +1.0%
    • Transaction Svs rev – MISS: Grew +33.3% y/y in Q4 (vs +71.5% y/y in Q3) and closed -6.5% below cons
  • Non-GAAP op profit – BEAT: Incr’d +13.9% y/y in Q4 (vs +47.7% in Q3) and beat cons by +1.5%
  • Non-GAAP EPADS – BEAT: Up +16.3% y/y in Q4 (vs +60.1% y/y in Q3) and topped cons by +1.5%

The Core E-Commerce Biz Has Been Beset By Growing Competition

• Domestic competition and strategic investments were a drag on PDD’s short-term financials –
  • The Co experienced an “intensified competition landscape”: Indicated “some merchants are facing homogenous competition” in “certain industries,” which has squeezed profits and “undermines [the Co’s] innovation and investments”; Emphasized that competition is “inherent to the e-commerce industry” in China
    • Though there was some optimism regarding the broader economic situation: Highlighted a “steady macro economy recovery” in domestic mkts
  • “Significant ecosystem investments” were also a headwind: PDD has been focused on delivering a broader range of products to consumers and has also implemented “policies to support high-quality suppliers,” such as cutting fees as well as others intended to streamline operations
• BUT long-term investments “are now bearing fruit” –
  • New product offerings have “gained traction in both urban and rural mkts”: These products range from fresh produce, snacks, beauty, as well as baby care and have become “daily essentials for diverse groups”
  • Logistics supports measures have “spurred double-digit order growth in Western China”: These also helped drive “remarkable growth in several trending categories” from more rural Western regions in China during Chinese New Year

• Color on the Co’s promotional campaigns –
  • Addt’l savings promos served as a “new growth engine” during the Double 11 (Nov 11) shopping holiday: PDD launched multiple rounds of “super double savings” that were layered on top of discounts from its RMB10bn program and offered “excessive support to quality supply” during Double 11
    • Order volume achieved “a significant milestone”: During Double 11, order volume surpassed CNY 61mn+ during the first two rounds of the campaign; The number of participating brands and products also reached a record high
  • Double 12 (Dec 12) campaigns were “further enhanced”: PDD provided 10bn consumption vouchers to reinforce benefits to consumers
  • The RMB10bn program has been “effective in lowering operational costs for merchants”: New campaigns have accel’d brand building efforts and have driven supply chain upgrades

The International Biz (Temu) Has Been Grappling W/ Shifting Compliance Standards

• The global biz has faced a “fast-changing external environment” as well as “fierce competition”: Taken together, these external changes “will inevitably bring some challenges” to PDD’s global biz
  • The Co has had to adapt to changes in macro policies: PDD has been focused on ensuring full compliance in its operations and has been proactively communicating w/ external parties across the mkt to continuously refine its svs and compliance standards
  • PDD has also launched several corporate social responsibility initiatives: Through long-term investments and “consistently doing the right things,” the Co believes it can strengthen its core competitive advantages and build “long-lasting trustworthy relationships” w/ global merchants and consumers
  • The growth strategy moving forward: Consists of deepening the Co’s core strength in supply chain efficiency, exploring new biz models, experimenting w/ innovative localized supply chain solutions, deepening its mkt presence, and introducing new product categories & svs

Addt’l Notes On PDD’s Go-To-Mkt Strategy In 2025

• PDD plans to “redouble efforts to advance [its] high-quality development strategy” in 2025: The Co will leverage its digital capabilities and “amplify investments” across both its supply and demand chain; PDD also intends to accel the delivery of high-quality products to Western regions
  • The Co established the Merchant Rights Protection Committee in Jan: The aim of the committee, which formalizes a communication channel w/ merchants, is to “fully understand merchant needs, systematically collect feedback, optimize [merchants’] experience, and refine svs framework”
  • Other svs have also been released to “optimize merchants’ ROI” and respond to “extensive competition”: Including abnormal order alerts, low-price risk warnings, data analytics tools, and a noncompliance alert system designed to “intensify protections for merchants’ rights”

Supporting The Agriculture Industry In China Also Remains A Key Focus

• PDD’s long-term initiatives in the agricultural space have seen “encouraging results”: Highlighted that the six finalist teams in the fourth PDD Smart Agriculture Competition achieved breakthroughs in yield, efficiency, and energy consumption through digitally optimized new farmland
  • The Co will sustain investments in agricultural tech, supply chain innovation, and training farmers moving forward: These are intended to promote agricultural modernization, improve supply chain efficiency, boost farmers’ incomes, and generate local employment

This week marked another chapter in the ever-growing book of AI developments, and we wanted to highlight what we thought were some of the most incremental updates from the space. In terms of new models, Baidu announced ERNIE 4.5, which offers improved multimodal understanding and more advanced language abilities compared to prior iterations of its foundational model, among other enhancements. The China Tech company also intro’d ERNIE X1, a new reasoning model that delivers a performance comparable to DeepSeek’s R1 at half the price. Elsewhere, Meta revealed that its Llama foundational model reached 1bn downloads and started to roll out its Meta AI assistant across Europe after previously encountering some delays due to regulatory concerns. Some notable strides were also made in agentic AI, as Adobe debuted a host of new AI agents and OpenAI released a handful of new transcription and voice-generating models to its API. See below for more details:

Updates On Foundational Models + A New Reasoning Model Hit The Mkt

• Baidu introduced two new open-source AI models last weekend (link/link) –
  • ERNIE 4.5 is the latest version of the Co’s foundational “ERNIE” model: Per Baidu, ERNIE 4.5 comes w/ the following capabilities –
    • • Excellent multimodal understanding ability
      • More advanced language ability
      • Comprehensive improvements to its understanding, generation, logic, and memory abilities
      • High EQ, w/ the ability to understand memes and satirical cartoons
  • ERNIE X1 is Baidu’s new reasoning model –
    • The model “delivers performance on par with DeepSeek R1 at only half the price” and has “stronger understanding, planning, reflection, and evolution capabilities,” per Baidu
    • It is also the first deep thinking model that uses tools autonomously
  • These latest Ernie models will be free-to-use for individuals: Previously, the Co charged users a monthly subscription rate

  -> Baidu shares rallied +9.0% on the back of the news but ended the week up just +0.8%; YTD, Baidu stock is trading up +12.1%

• Meta’s Llama models hit 1bn downloads (link/link/link): This is up from 650mn downloads as of early Dec 2024, representing a ~+53% increase over a ~3-month period;
  • Llama has achieved widespread success: The model, which powers Meta AI, is also used in production by Spotify, AT&T, and DoorDash, among others
  • More Llama models are in the works over the next few months: These include reasoning similar to OpenAI’s o3-mini as well as models w/ natively multimodal capabilities
    • Some of these models will have “agentic” features, Meta CEO Mark Zuckerberg: Suggesting that they will be able action autonomously
  • More details will likely be revealed at LlamaCon, Meta’s first gen AI dev conference: This is scheduled to take place on April 29

-> On a related note, Meta started launching its Meta AI assistant across 41 European countries and 21 overseas European territories this week; The Co originally planned to roll out Meta AI to the region late last yr but took longer due to the need to “navigate its complex regulatory system”; Meta will start by incorporating its AI assistant into its messaging apps across Facebook, Instagram, WhatsApp, and Messenger; It will be available in six languages from the region, but its capabilities will be limited to text queries for now (link)

-> Meta shares ended the week down -1.9% but are still up +1.8% YTD

Agentic AI Also Remains A Central Theme

• Adobe launched a new tool and AI agents to enhance customer experiences (link/link): The new offerings were debuted this week at Adobe Summit, the Co’s annual user conference that highlights the latest capabilities within its user experience platform
  • The Adobe Experience Platform Agent Orchestrator is a portal for managing multiple AI agents: The tool trains AI agents on the Co’s enormous volumes of info stored in its customer data platforms; This significantly enhances their reasoning capabilities and makes them much more dynamic and capable
  • Adobe also unveiled a wide array of AI agents –
    • An account qualification agent designed to evaluate and advance new sales leads in B2B mkts
    • An audience agent that assists mkters by analyzing cross-channel segment data to identify and optimize high-value audience segments
    • A content creation agent that allows mkting teams to quickly produce high-quality, commercially safe mkting content for email, web, and paid media; Also offers dynamic personalization abilities
    • A data insights agent that analyzes multiple customer signals across the organization and surfaces valuable info and trends
    • A data engineering agent that performs tedious tasks related to cleansing, integrating, and securing customer and sales info
    • An experimentation agent that can help mkting teams working on personalized initiatives hypothesize and simulate new ideas while conducting impact analysis
    • A journey agent that assists in orchestrating cross-channel experiences

-> Adobe shares finished the week down -1.9% and are down -12.9% YTD

• OpenAI released upgraded transcription and voice-generating AI models to its API (link): These models can speak w/ a biz’s customers and fit within the Co’s broader “agentic” vision
  • The new GPT-4o-mini-tts text-to-speech model delivers more nuanced and realistic-sounding speech: The model is also more “steerable” than prior iterations, enabling devs to better instruct it on how to say things in natural language
    • The goal is to allow devs to tailor both the voice “experience” and “context”: Per Jeff Harris, a product staff member at OpenAI
  • The new GPT-4o-transcribe and GPT-4o-mini-transcribe speech-to-text models replace OpenAI’s Whisper transcription model: The new models are trained on “diverse, high-quality audio datasets” and can better capture accented and varied speech, even in chaotic environments
    • The new models are also less likely to hallucinate than Whisper…
    • … BUT accuracy varies depending on the language being transcribed: For example, Indic and Dravidian languages, such as Tamil, Telugu, Malayalam, and Kannada have a word error rate of 30%, meaning the model will differ from human transcription 3 out of every 10 words

Aside from the US vs China dynamic as it relates to humanoid robots as we highlighted in Theme #1, the race to electrify and autonomize the roads is charging ahead, with major US and Chinese EV and AV players making key moves this week. Top of the pack was BYD, which announced a new battery system that can add ~250 miles of range in just 5-8 minutes – nearly as fast as refueling a gas car. Set to launch in April with the Han L sedan and Tang L SUV, the technology arrives as Tesla’s position in China weakens, with February sales in the region plunging -49% y/y while BYD’s soared +161%. As a further blow to Tesla, Mercedes also introduced its new CLA sedan this week, featuring recharging speeds that also outpace Tesla’s.

Meanwhile, Tesla secured a California permit to transport passengers in company-owned vehicles as it moves toward ride-hailing services. However, the approval is limited to driving employees and does not allow for a full rideshare rollout or autonomous vehicle testing. Finally, Waymo gained approval to map San Francisco International Airport for future robotaxi service and expanding its autonomous driving zone in San Jose, which follows the company’s announcement last week that it was expanding across Silicon Valley. See below for more color on these key updates.

• BYD unveils battery system that charges EVs in 5-8 minutes (link) –
  • The new “Super e-Platform” battery and charging system can provide ~250 miles (400km) of range in ~5-8 mins in tests of its new Han L Sedan
    • That is about as fast as it takes to refuel a regular car
    • BYD will start selling vehicles with the new technology (the Hans L sedan and Tang L SUV) starting in April
    • BYD will also build 4k+ charging stations in China designed to accommodate the new technology
  • How does that compare to Tesla? Tesla’s currently take ~15 minutes to add ~200 miles (320km) to its range
    • That said, Tesla has a much larger network of 65k+ Superchargers worldwide
    • Tesla sales in China have taken a hit –
      • Tesla’s China shipments plunged -49% y/y in Feb to just 30,688 vehicles, the lowest monthly figure since in July 2022
      • On the other hand, BYD sold 318k+ cars last month, up +161% y/y and is now the top carmaker in China, the world’s biggest auto market, w/ a share approaching 15%

-> BYD shares soared +6% to a record high on the news Monday, raising its market value to almost $162bn; Tesla shares, by contrast, sank -4.8%.

-> Also this week, Mercedes unveiled its new CLA compact sedan, which has up to a 492-mile range and can recharge nearly half its battery and add ~200 miles (325km) to its range in 10 minutes (link)

Source

• Tesla gets approval for passenger transportation in California (link) –
  • Approved for a transportation charter-party carrier permit, which allows Tesla initially to drive its employees on a prearranged basis in company-owned vehicles
    • Approval does NOT allow Tesla to operate a rideshare svs
    • Approval also does NOT cover autonomous vehicle testing or deployment: Tesla’s application did NOT seek participation in the commission’s Autonomous Vehicle Passenger Service Programs in either a drivered or driverless capacity
  • Permit Tesla received is different from the one Uber and Lyft hold –
    • Tesla was granted a transportation charter permit (TCP), which means Tesla owns the vehicles and uses employees as drivers (as an example, a TCP is given to Cos that want to offer prearranged transportation services like roundtrip sightseeing)
    • Uber and Lyft hold transportation network company (TNC) permits, which applies to Cos that use an online-enabled application like a smartphone app to connect drivers using their personal vehicles with paying passengers

• Waymo robotaxis are now allowed to map San Francisco Airport (link): San Francisco and Waymo announced a 30-day temporary access agreement for Waymo to begin mapping roadways around San Francisco International Airport (SFO) ahead of possible robotaxi service to and from the airport
  • How will mapping work? The permit allows Waymo to utilize up to two of its vehicles at a time, with an actual human riding in the driver’s seat, to travel around the airport and train its fleet of driverless cars where and where not to go when taking passengers to and from SFO
    • The agreement began on March 14 and is valid through April 13, with an option for the airport to extend the term for another 30 days
  • Mapping is just the first of many steps: Once mapping is completed, Waymo will need to get permission from the airport to start test driving the vehicles at SFO – first w/ safety drivers in the front seat, and then with no drivers at all
  • Still some time before riders will be able to hail a Waymo to and from SFO: While Waymo is permitted to operate its vehicles on surrounding highways, the Co is still testing on and has yet to include freeways on its allowable routes

-> Also this week, Waymo annc’d that the California Department of Motor Vehicles has given the Co approval to expand the areas of the South Bay where it can operate fully autonomously, including almost all of San Jose; While the public won’t have access at this time, Waymo is “working closely with local officials, emergency responders, and communities to safely expand driving operations” (link)

The Recording Industry Association of America (RIAA) released its 2024 Year-End Revenue Report this week, highlighting growth trends across the music industry’s most prominent categories. Notably, the music industry hit a key milestone in reaching 100mn US paid music subscribers by the end of the year (UMG recently est’d a global TAM of 230mn paid music subscribers), and, unsurprisingly, streaming accounted for the vast majority of the recorded music industry’s revenues. However, growth in streaming has continued to slow, both in terms of subscribers and top-line trends. This is especially relevant as Warner Music, UMG, Spotify, and others are pushing towards more premium, artist-centric models to drive the next leg of growth in the industry.

See below for what we thought were the most incremental insights and takeaways from the report. Also, the full report can be accessed here: LINK

• Growth in total US retail recorded music revs decel’d on a y/y basis: 2024 US retail recorded music revs of $17.7bn were up +3.5% y/y in 2024 (vs +7.5% y/y in 2023)

• Streaming cont’d to account for most of recorded music revenues in 2024: Rev from paid subscriptions, ad-supported svs, digital and customized radio, social media platforms, digital fitness apps, and others rose +3.5% y/y to a record-high $14.9bn, collectively accounting for 84% of total revs for the third straight yr
  • Paid subscription svs revs comprised 79% of streaming rev: Incr’d +5% y/y to $11.7bn in 2024, which represented nearly two-thirds of total US recorded music revs
  • Ad-supported svs represented 12% of streaming rev: But declined -2% y/y to $1.8bn in 2024, accounting for 10% of total US recorded music revs; Includes YouTube, the ad-supported version of Spotify, Facebook, and others
  • Digital and customized radio music rev made up 9% of streaming rev: Rose +3% to $1.4bn in 2024; Includes SoundExchange distributions for revs from svs like SiriusXM and internet radio stations, plus payments directly paid by similar svs
  • Limited tier subscriptions revs accounted for 7% of streaming rev: But fell -2% y/y to $1bn in 2024; Note, these are svs limited by factors such as mobile access, catalog availability, product features, or device restrictions and include svs like Amazon Prime, Pandora Plus, streaming fitness svs, and others

• The avg number of US paid music subscriptions also saw decel’ing levels of y/y growth: Avg US paid music subscriptions of 100mn incr’d +3.3% y/y in 2024 (vs +5.7% in 2023)

• Trends within other categories –
  • Growth in rev from physical products slowed on a y/y basis: Physical product rev of $2.0bn was up +5% y/y in 2024 (vs +12% y/y in 2023)
    • Vinyl record revs comprised nearly three-fourths of physical revs: Incr’d +7% y/y to $1.4bn, marking the eighteenth consecutive yr of growth
  • Digitally downloaded music rev accounted for just 2% of total recorded music rev: Rev from digitally downloaded music of $336mn was down -18% y/y; In 2012, this category represented 43% of total US recorded music rev

Gen AI traffic to US retail sites is still “modest” in the grand scheme of things, but growth has been on fire and has been “doubling every two months since September 2024.”  Based on Adobe Analytics’ data this week, gen AI-powered traffic to US retail sites in February is up +1200% vs July 2024. Gen AI is increasingly being used to assist in compiling research before making a purchase and to help with travel plans as well as financial guidance. Conversion rates vs other traffic sources remain lower, but the gap has been narrowing. We certainly expect these trends to continue…

See below for our key takeaways from the report. (link to report)

• There has been a significant surge to US websites from gen AI sources –
  • Gen AI traffic to US retail sites first surged over the 2024 holidays (from Nov 1 to Dec 31), where traffic from gen AI sources incr’d +1,300% y/y
  • In Feb 2025, traffic from gen AI sources incr’d+1,200% vs July 2024

• 39% of respondents have used gen AI for online shopping, while 53% said they plan to do so: Based on Adobe’s survey of 5,000 US consumers, respondents are using gen AI to –
  • Conduct research – 55% of respondents
  • Receive product recommendations – 47%
  • Find deals – 43%
  • Get present ideas – 35%
  • Find unique products – 35%
  • Create shopping lists – 33%

• Consumers coming from gen AI sources show greater engagement on sites
  • They have +8% higher engagement and linger on the site for a longer period of time
  • These visitors also browse +12% morepages per visit, with a –23% lower bounce rate

• Conversion rates from traffic from gen AI sources are lower (but the gap is narrowing)
  • Gen AI traffic to US retail sites is -9% less likely to convert vs other sources of traffic (but this is an improvement from the 43% in July 2024)
    • Conversion rates from gen AI traffic are HIGHEST in electronics and jewelry
    • Conversion rates from gen AI traffic are LOWEST in the apparel, home goods & grocery
  • Supports the view that AI is being used for the research and consideration stage vs when consumers are hitting the buy button

• Vertical call outs:
  • In Feb 2025, traffic to US banking sites from gen AI sources incr’d+1,700% vs July 2024
  • In Feb 2025, traffic to US banking sites from gen AI sources incr’d+1,200% vs July 2024
  • In Feb 2025, traffic to US banking sites from gen AI sources incr’d+1,200% vs July 2024

The sports and sports media space were also front-and-center this week, headlined by the Boston Celtics’ agreement to a record-setting $6.1bn sale to Bill Chisholm. The price tag represents the highest amount ever paid for a North American sports franchise, eclipsing the Washington Commanders’ $6.05bn valuation when the NFL team was sold to Josh Harris in 2023. The MLB has also seen valuations skyrocket, as the Oakland Athletics are reportedly selling new shares to outside investors at a $2bn valuation, which is a notable step-up from the Baltimore Orioles’ $1.72bn valuation when the team was sold a year ago.

There were also some interesting updates pertaining to the sports media side this week. Along with reports that ESPN has re-engaged the NFL to purchase the league’s media assets, new sell-side estimates show that NBA rights moving from Warner Bros Discovery to NBC and Amazon could mark one of the biggest transfers of audiences and ad dollars in the industry’s history. See below for more details:

• The Boston Celtics agreed to a record-high $6.1bn sale (link): The Celtics ownership group, led by the Grousbeck family, agreed to sell the franchise to Bill Chisholm, the co-founder of PE firm Symphony Technology Group
  • The amount would be the most that a North American sports franchise has ever sold for…: Surpassing the $6.05bn valuation that the Washington Commanders received when purchased by Josh Harris in 2023
    • … And represents a +50% increase over the previous record for an NBA team: Also in 2023, Mat Ishbia acquired the Phoenix Suns at a $4bn valuation
    • The Grousbeck family originally bought the team for $360mn in 2002: This represents a +13.1% CAGR in value since the Grousbeck purchased the Celtics
  • Wyc Grousbeck will remain governor of the Celtics until the end of the 2027-28 season: This was at Chisholm’s request, per Grousbeck, though the current ownership group initially indicated that Wyc Grousbeck intended to remain governor of the team through 2028 when annc’ing the sale
  • Members of the new ownership group: Include Sixth Street, a PE firm that has agreed to invest $1bn+, Rob Hale, who is currently a minority owner of the team, and Bruce A. Beal Jr, the president of real estate firm Related Companies
  • Chisholm’s approach is “to win and raise banners”: Chisholm has been a lifelong Celtics fan and has already spoken to Celtics GM Brad Stevens about aligning their goals

• The Oakland A’s are reportedly seeking a $2bn valuation in a new share sale (link): Managing partner John Fisher is looking to raise $550mn to finance a new $1.75bn ballpark in Las Vegas by offering new shares in the Athletics (A’s) MLB team, per sources
  • Several investors have expressed interest in contributing $200mn+ at an enterprise value of $2bn: This valuation eclipses the $1.72mn paid by Dave Rubenstein for the Baltimore Orioles a yr ago; Notably, the A’s have been in the bottom five of attendance in the league almost every yr since 2015
    • Fisher led a group that originally purchased the A’s for $180mn in 2005: This represents a +12.8% CAGR in the team’s value since then
  • The Fisher family currently owns 95% of the team: Following the share sale, the Fisher family would still own more than 80% of the A’s
  • Financing for the Las Vegas ballpark is coming from a variety of sources: Along w/ the $550mn from outside investors, Clark County is contributing $350mn, $300mn will be from loans from US Bank and Goldman Sachs, and the Fisher family will be responsible for the remainder

• ESPN has resumed talks w/ the NFL to purchase some of the league’s media assets (link/link): After initially broaching the idea last yr, negotiations for ESPN to acquire several NFL media assets, including the NFL Network, RedZone, and NFL+, have started again
  • The acquisition could reportedly cost ESPN as much as $2bn: The NFL could also take an equity stake in ESPN as part of the potential deal
  • ESPN would acquire 8 NFL Network game broadcasts: This would add to ESPN/ABC’s coverage of the NFL; Last yr, ESPN/ABC carried 25 NFL games, including 21 games on Monday night and 2 playoff games
    • The deal could also help ESPN maintain rights to the NFL Draft: The NFL Network currently airs its own broadcast of the popular event alongside ESPN, and consolidating these rights would be another win for ESPN
    • Other NFL Media offerings would help ESPN attract fans: Such as the NFL’s fantasy products as well as shows like Good Morning Football and The Insiders

• NBA rights moving from WBD to Peacock and Amazon Prime Video could represent one of the largest transfers of audience and ad dollars in history (link): This was per estimates by an analyst at MoffettNathanson
  • Warner Bros Discovery (WBD) is projected to lose $1.1bn in TV ad rev in 2026: That would represent ~23% of its total ad rev this yr; A significant contributor would be the absence of NBA on its networks for the first time since 1989
  • Ad rev estimates for the next NBA rightsholders: Notably, NBC and Amazon have already started signing ad deals, while Disney has been more patient as the incumbent
    • Disney is expected to capture $1.25bn in annual ad rev from NBA games: The Co has an addt’l leg up on other rights holders b/c it carries the NBA Finals but will also pay the most for its rights at $2.6bn per yr
    • NBC could take in $1bn or more in ad rev: NBC is estimated to be paying the NBA $2.5bn per yr for its new package
    • Amazon could win up to $750mn in ad rev: The Co is expected to pay the NBA $1.8bn per yr for streaming rights
  • The rightsholders also see oppties to monetize the rights by tying them to other sports properties:
    • NBC is in talks w/ advertisers about linking commercials across the NBA, its 2026 Super Bowl b-cast, next yr’s Winter Olympics, and the next FIFA World Cup
    • Amazon intends to incorporate an NBA game w/ its emerging Black Friday franchise, which already uses an NFL game to drive ad deals around holiday gift-giving

Fox’s AVOD service Tubi (in conjunction The Harris Poll) published a consumer survey this week that included a few interesting stats that we wanted to flag for those of you that care about video streaming trends: (link/link)

• Consumers are spending $129 in aggregate per month on streaming svs AND paid TV subscriptions (I must be seriously overpaying!) – this is up +7.5% y/y
  • 56% of respondents monitor streaming services carefully so that they do not overspend
  • Gen Z viewers are particularly cost-conscious, as 76% say they have ended or would end their streaming subscription over incr’d price
• How do consumers feel about ads on streaming svs? They are willing to accept ads BUT not if they are paying for it
  • 79% said that if they are paying for a streaming svs, they expect no ads at all
  • 81% say that watching ads is a fair trade-off for access to free content on a streaming svs
• 81% of GenZ would take action after seeing interesting adds (81%), BUT 73% feel that the ads they’re currently seeing on streaming seem “misaligned” with their personal preferences -> the industry has more work to do here
• How many streaming svs do consumers have? On avg, almost 7, which is a mix of paid streaming svs (3.9 on avg) and free svs (2.6 on avg)
• What are current usage trends?
  • 57% of viewers stream TV and movies for 1-3 hours in one sitting
  • 38% stream for 3 hours or more in one sitting
• 80% of viewers say they’d rather spend their time watching a TV show/movie than scrolling on social media
• Related to what type of content viewers want to watch:
  • 63% say they are more exhausted than ever from sifting through all the choices available when looking for content to watch
    • 69% also say that if they don’t already know what they want to watch, it will take at least 10 minutes of searching to find something
  • 70% of viewers want to see more TV shows / movies on streaming that are independent or from smaller creators (up +4% y/y) –
    • 73% of Gen Z viewers say they prefer to watch original content over franchises/remakes
    • 72% of Gen Z viewers also wish they had more of a say in the type of content that gets made for streaming svs ->maybe there are opportunities to crowdsource?
  • 66% say they are enjoying discovering content that was originally released 10 years ago
    • 87% of Gen Z viewers agree that streaming svs should show them older content they’ve never seen before in addition to new shows/moviesRelated to what type of content viewers want to watch:
• Methodology: The survey was conducted w/in the US by The Harris Poll on behalf of Tubi from Oct 21, 2024, to Nov 1, 2024, representing 2,502 18 adults that stream video at least 1 hour a week

• Robinhood launches prediction markets hub to retail trading app, “giving customers the opportunity to trade on the outcomes of some of the world’s biggest events” (link/link) –
  • Launched in partnership w/ Kalshi, a CFTC-regulated prediction market platform which is already registered to list a wide range of prediction markets, including political outcomes, economic indicators, and entertainment events
  • The new feature allows users to trade contracts on various events,including the Federal Reserve’s target interest rate in May and the results of the men’s and women’s College Basketball Tournaments
  • More coming soon? “We have been in close contact with the CFTC over the past several weeks and look forward to continuing to work with them to promote innovation in the futures, derivatives and crypto markets”

• Apple is reportedly losing $1bn annually on Apple TV+ streaming svs (link/link) –
  • Apple is reportedly spending ~$4.5bn on content annually, down from $5bn per year it has been spending since launching Apple TV+ in 2019
    • For reference, Apple generated $391bn in revenue and posted a net profit of $93.7bn for its fiscal year ended in September 2024
    • The streaming svs is understood to be the only Apple subscription service that isn’t profitable
  • Apple TV+ also reportedly has ~45mn subscribers
    • That said, it’s unclear how many are paying the full $9.99/month or $99.99/year subscription fee and how many are on discounted or free accounts
    • Per research and measurement firm Antenna, Apple TV+ added ~2mn subscribers in February, largely due to the popularity of its original show, Severance

• More satellite mobile updates this week –
  • Charter and Comcast launch satellite connectivity for mobile devices (link): Charter’s Spectrum Mobile and Comcast’s Xfinity Mobile have launched satellite-based services through a collaboration with Skylo, a non-terrestrial network (NTN) service provider
    • Will provide customers with access to emergency messaging services when their device is not connected to either a cellular or WiFi network.
    • Now available on Samsung’s Galaxy S25 and Google Pixel 9 devices; Skylo plans to make its services available on additional devices in the future
    • The ability to send and receive SMS text messages over satellite service on Galaxy S25 and Google Pixel 9 devices is expected to be added in the coming weeks
  • Verizon launches free satellite texting svs for select Android devices (link/link): Verizon announced a “significant” expansion of its satellite texting capabilities, enabling its customers to become the first in the US to send text messages to any other customer device via satellite when outside the reach of terrestrial cellular networks
    • Comes at no extra charge to Verizon customers
    • Available on Google Pixel 9 or Samsung Galaxy S25 series phones
  • Starlink and Bharti Airtel sign deal to explore bringing satellite internet to India (link): The deal will “explore how Starlink can complement and expand Airtel’s offerings, and how Airtel’s expertise in the Indian market complements SpaceX’s direct offerings to consumers and businesses”
    • As part of the agreement, the two Cos will explore offering Starlink equipment at Airtel’s retail store network, Starlink services through Airtel to business customers, and using each other’s network infrastructure to enhance coverage across India
    • Airtel, which is India’s second largest telecom operator, said that the deal with Starlink depends on whether it can receive government approval to begin operating in India