Apple is behind in AI, more TikTok woes, and IT budget losers

24 March 2024 | Issue #19 - Mentions $AAPL, $GOOG, TikTok, $UMG, $META, $MSFT, $ACN

Mar 24, 2024

Welcome to the nineteenth edition of Tech takes from the cheap seats. This will be my public journal, where I aim to write weekly on tech and consumer news and trends that I thought were interesting.

Let’s dig in.

Apple is behind in AI

The week started off with a bang with rumours of a deal between Apple and Google.

Apple Inc. is in talks to build Google’s Gemini artificial intelligence engine into the iPhone, according to people familiar with the situation, setting the stage for a blockbuster agreement that would shake up the AI industry.
The two companies are in active negotiations to let Apple license Gemini, Google’s set of generative AI models, to power some new features coming to the iPhone software this year, said the people, who asked not to be identified because the deliberations are private. Apple also recently held discussions with OpenAI and has considered using its model, according to the people.
If a deal between Apple and Google comes to fruition, it would build upon the two companies’ search partnership. For years, Alphabet Inc.’s Google has paid Apple billions of dollars annually to make its search engine the default option in the Safari web browser on the iPhone and other devices. The two parties haven’t decided the terms or branding of an AI agreement or finalized how it would be implemented, the people said.

Although currently mere speculation, the validity of these rumours remains uncertain until potentially confirmed at the upcoming WWDC in June. Recent reports from the Wall Street Journal, however, lend additional weight to these speculations.

Apple has held preliminary talks with Baidu about using the Chinese company’s generative artificial-intelligence technology in its devices in China, the latest example of the iPhone maker’s efforts to widen its AI capabilities.
The U.S. tech giant has been exploring using external partners to help accelerate its AI ambitions. It has held discussions with companies including Google and OpenAI about using their technology to power its mobile features.
In China, Apple has been looking for a local generative AI model provider, mainly because China requires such models to be vetted by its cyberspace regulator before being launched to the public, people familiar with the matter said.
Since the authorities introduced the rule in August, Beijing has approved more than 40 generative AI models, including Baidu’s Ernie Bot. No models made by foreign developers have been approved yet, and it couldn’t be determined whether any foreign companies have sought government approval.
Leading generative AI models including OpenAI’s ChatGPT and Google’s Gemini aren’t available in China. Apple rival Samsung’s newest Galaxy smartphone uses Gemini outside of China and Baidu’s Ernie in China to power some AI features.
Apple’s discussions with Baidu are still exploratory, the people said. It couldn’t be determined whether Apple has engaged with other Chinese generative AI companies.

This deal with Baidu would make sense as the article notes Samsung's latest Galaxy smartphone uses Gemini outside of China and Baidu's Ernie in China to enable AI features, since Google isn't permitted to operate in China. Stratechery offered insightful commentary on the Apple and Gemini partnership.

Although there are some unknowns about the deal, the decision by Apple seems like a defensive strategy to retain market share. Despite generative AI features and use cases still being in development, they add a "new" factor to these devices, potentially attracting consumers looking to upgrade their phones. Apple, similar to Samsung, has developed its own generative AI model, though its technology will be restricted to smaller scale, on-device capabilities. As Thompson points out, Apple started training their models late and hasn't built the necessary infrastructure to cater to its user base. Collaborating with Google allows Apple to quickly address this gap in their capabilities and prevent iPhone users from switching to a different handset provider for generative AI. Licensing Google's technology means less software integration for iPhones, but Apple doesn't have much choice here. It might not need tight integration if responses are personalized for small scale, on-device use, and latency is low for larger scale models. Perplexity, the AI-search engine startup, is doing well even though it's essentially a wrapper for Google, Bing, OpenAI, and Claude. It's uncertain whether Apple will ever develop its own large scale model for cloud-based inference given the substantial investment required, uncertain return prospects, and their current lag.

This leads to the next consideration - economics. We don't know the specifics of the licensing agreement, but inference cost isn't free. Apple and Samsung will have to pay GCP for compute costs. Currently, Samsung seems to be absorbing the cost as they're not charging customers an additional fee for using Gemini Pro, which is likely not too expensive since it's only available for summarizing select Samsung apps (Notes, Voice Recorder, and Keyboard). As more compute-intensive tasks emerge, handset companies will need to monetize this, either through device cost increase or a subscription fee.

For Google, they'll be receiving additional revenue with approximately 60% gross margins (Anthropic has 50-55%, Google likely has lower inferencing costs due to TPUs, the figure could be as high as 90% in the future with scale using software gross margins as a comp). Whether these generative AI queries will cannibalize Google search queries is yet to be determined. Google's Chief Business Officer, Philipp Schindler, stated at a recent conference that they saw positive search query growth in all major markets over the last 12 months ending in February. This suggests that GenAI chatbots have had minimal impact on the market so far and that GenAI is an incremental addition to general search. Crucially for Google, this deal helps to maintain their position at a strategic distribution point, and they could be paid for it.

PS: This agreement underscores Google's prowess in AI, especially noteworthy considering Apple's tendency to prioritize consumer-centric partnerships. It's worth noting, however, that Samsung remains open to exploring alternatives, as per the insights from the WSJ article.

The South Korean firm plans to continue cultivating Gauss for its own purposes. At the same time it will keep working with Google—or any other AI partner—to enable generative AI on Samsung’s phones, said (Won-Joon) Choi, the Samsung executive who heads research and development for mobile.
“Right now it’s Google. But it could be Microsoft or Meta in the future. The goal is to provide the best solution,” Choi said.

TikTok headwinds in the US

From the WSJ

As TikTok fends off a new attack by Congress, it is also battling headwinds that have sapped its U.S. business of its mojo.
For the first time in TikTok’s history, its user growth is stagnating, according to people familiar with the matter. Ad sales are hitting growth targets but not exceeding them, and it is at war with the world’s largest music company. Its bet on e-commerce as a new revenue source has angered some users by muddying the app’s stream of short videos. And TikTok’s main competitor, Instagram Reels, has been gaining ground.
…
In Washington, the House voted overwhelmingly Wednesday to approve a bill that requires ByteDance to sell off TikTok or face a ban, contending among other things that the collection of American user data by its owners poses a national-security risk. TikTok has said it hasn’t shared and won’t share the data with the Chinese government even if Beijing demands it.
The Chinese government, meanwhile, has signaled that it won’t allow a forced sale of TikTok, limiting options for the app’s owners as potential buyers begin talking about a bid for its U.S. operations.
And then there is the business itself. TikTok is still gaining new users, but the number of people quitting the app has grown to the point that the total number of American users has stalled, the people familiar with the matter said. In the past, TikTok consistently added more users than it was losing.
U.S. average monthly users ages 18 to 24 declined by nearly 9% from 2022 to 2023, according to the mobile analytics firm Data.ai. Some users in their 20s say they have gotten off the app entirely to focus more on life and work.
Those trends have complicated TikTok’s relationship with U.S. advertisers, which have long been drawn to its young audience. TikTok met its ad-sales growth targets for the back half of 2023 but didn’t exceed them, according to people familiar with the matter.
At a congressional hearing in January, TikTok Chief Executive Shou Zi Chew said that one of the app’s fastest-growing demographics is people over the age of 35. Whether to say that in public was a subject of internal debate, people familiar with the matter said.

In 2023, TikTok generated $20bn in ad revenue, with the US accounting for $16bn of this amount. The Information reported in December that the company requested its advertisers to increase their spending for a global revenue target of $30bn. Advertisers have been attracted by the platform's highly engaged young audience, who reportedly spend almost an hour per day on the app according to eMarketer's findings. This has enabled TikTok to target specific demographics and gain some ad share (increasing by 7 percentage points in 2023). However, as user growth slows and demographics shift, TikTok is exploring other monetisation methods beyond ads.

Last August, I covered TikTok's ecommerce ambitions here, but it appears that some users are finding its shopping features bothersome.

As part of the rollout, TikTok flooded users’ For You pages with videos of influencers promoting products available on the app.
Some users have said they find the shopping videos to be obnoxious and that they have ruined the watching experience on TikTok. The addition of the shopping feature has made the app more messy and cluttered, some users have said.
“How do we opt out of TikTok Shop videos? I don’t want to see them anymore,” Brody Wellmaker, a TikTok creator with 21 million followers, said in a video on the app. “I’m not going to shop on TikTok Shop. I’m going to shop on Amazon. I want my For You page back the way that it was.”

Livestream shopping has gained significant popularity in Asia, but the success has not been mirrored in the West. Just two years ago, TikTok abandoned its ecommerce expansion plans in Europe and the US following an unsuccessful UK launch. It appears that consumer preferences and behaviors differ significantly. Instagram's decision to remove the shop tab and concentrate more on shop ads seems to be a wise move, as it prevents clutter and doesn't detract from the app's primary purpose of entertainment.

The last thing I’ll highlight from the article is TikTok’s battle with Universal Music Group.

TikTok’s licensing deal with Universal—whose artists include Taylor Swift, Drake and Ariana Grande—expired in December. After signing a one-month extension, the two sides couldn’t agree on how much to pay artists for use of their songs on the social-media app, where music and dancing have been core content.
Last month, all music from Universal’s artists had to be wiped from TikTok, affecting not only artists but also creators who use snippets from their songs in videos.
Universal transferred several promotional campaigns for artists to Instagram, including for Kacey Musgraves, the Grammy-winning country singer who had a new album come out Friday.
Before lawmakers introduced the latest bill seeking a sale or ban of TikTok, the company had reached out to Universal to restart negotiations to get the music from some of the world’s biggest pop stars back on the app. No meaningful progress has been made, according to people familiar with the negotiations.

TikTok has emerged as a significant platform for discovering new music. Songs that trend on TikTok frequently land on the Billboard 100 or Spotify's top 50. As a result, artists are now creating TikToks to promote their new music, and music marketers are collaborating with influencers on the platform for song promotion. Aware of its influence in the music industry, TikTok has even considered vertically integrating the entire music distribution process1. Losing access to a major music label's catalog is a significant setback and substantially diminishes its ability to compete with other short-form video competitors. So far, its efforts here are only a test as it is mindful of pissing off an important partner (the music labels)

The way TikTok will navigate these challenges will be intriguing to observe. The internet sector has witnessed substantial changes in just four years. Notably, in 2020, President Donald Trump attempted to ban the app; now, he's advocating for its preservation.

The acquisition that wasn’t

From The Information

Microsoft doesn’t want its plan to hire two of Inflection AI’s co-founders and most of its 70-person staff to be seen as an acquisition. But it’s still writing a hefty check to the two-year-old artificial intelligence startup.
The software giant has agreed to pay Inflection approximately $650 million, mostly in the form of a licensing deal that makes Inflection’s models available for sale on the software giant’s Azure cloud service, according to a person involved in the transaction. The startup is using the licensing fee to help provide its investors with a modest return on their capital, according to a second person who was briefed on the arrangement.
The financial details reveal previously unknown aspects of the unconventional arrangement that Microsoft and Mustafa Suleyman, Inflection’s co-founder and CEO, unveiled this week. The mass exodus of staff will leave Inflection, which has raised $1.5 billion in less than two years to develop Pi, a chatbot, with few of the AI researchers that made it valuable. The much smaller company is instead fashioning itself as an “AI Studio” to help other businesses train and fine-tune their AI models, and has named former Mozilla executive Sean White as CEO.
The deal immediately drew questions from AI practitioners and venture capitalists about whether Microsoft structured it this way to avoid a lengthy review from antitrust regulators. U.S. officials have said they are already concerned about the concentration of AI power among a few large technology giants. A Microsoft spokesperson declined to comment on these questions.

There are several elements to consider here. According to an article from the Information, Microsoft CEO Satya Nadella approached Mustafa Suleyman a few months ago to hire him and his team of AI engineers and researchers. This occurred after the OpenAI incident where Sam Altman was ousted as CEO. When combined with Microsoft’s investment in Mistral a few weeks ago, it indicates the company’s hedging away from OpenAI. I wrote in November about how the company was offering Models as a Service in Azure - hedging and making multiple models available on its enterprise service makes the most sense as an infrastructure provider. It broadens options for customers and ensures higher utilisation datacenter footprint.

However, the appointment of Suleyman as Microsoft AI’s CEO is surprising. A philosophy dropout from Oxford, he co-founded DeepMind, worked as a VP in AI Product Management and AI Policy at Google (after being fired from DeepMind for bullying), and briefly worked as a VC at Greylock before co-founding Inflection AI. Despite his diverse experience, Suleyman lacks a successful track record in leading a consumer division, which is his new focus at Microsoft. He is known as ‘decel’ that led teams at DeepMind in deploying AI into healthcare use cases and optimised energy consumption for Google’s data centres. Pi - Inflection AI’s consumer chatbot failed to gain traction and was bailed out. Perhaps it was launched too late compared to alternatives without much differentiation. This further highlights the importance of having distribution. It’s possible that Mustafa realised this. Anyway, while this move by Microsoft was creative, I have my doubts on whether they have the right man for the job.

Finally, there's speculation that Microsoft's acquisition method could serve as a blueprint for other tech giants to avoid FTC scrutiny. However, this method may only be effective in specific cases, such as with Inflection, which was overvalued and lacked a clear product-market fit. Microsoft primarily wanted the team, not Inflection AI’s technology, making this a relief for the company's investors who got bailed out.

Budget winners vs budget losers

Accenture, a leading IT consulting company worldwide, is considered an AI winner due to the potential benefits from clients spending money on formulating and implementing an AI IT strategy. This week, the company reported second quarter results that met market expectations but lowered its full-year revenue guidance from 2%-5% to 1%-3%. Although generative AI new bookings contributed $1.1bn in the first half, the firm cited macroeconomic uncertainty as a factor leading to constrained IT budgets among its customers.

I thought these comments from Accenture’s CEO Julie Sweet were interesting

So there's 2 things. So first of all, it's about prioritization, right? So their overall constraint on spending, right? So you make choices, right, as opposed to it being additive.
So they're not able to allocate extra budget. They're prioritizing their budget. So you're seeing more of a substitution right now as opposed to, "Hey, we need to do this. Let's add to the budget." And that's tied to the uncertain macro that's putting people constrained. I had one banker say that corporates have put themselves on a diet given the macro, right?
The second thing, Bryan, is you have to remember that you can't just jump to the great data foundation, right? You need to be in the cloud. You've got to have modern platforms. And so what you should read into the higher client -- the clients doing these higher bookings, right, is that they're doing the big transformations often times to be ready to put in the data foundation, right?
There's only still 40% of workloads are in the cloud, 20% of those roughly haven't been modernized, right? Many of our clients haven't put in the platforms. If you don't have the major ERP platforms that are modern, you don't create a data foundation to fuel GenAI in isolation.
So you've got to build the digital core. And as we've said, there's a lot more to go. And that's what's driving these larger complex transformations. Like people don't like to do these big transformations in the sense of they're big, they're hard, they're complicated, and they need to do them in order to ultimately be able to use the AI not just in a part of the business or as a proof of concept, but really to transform and get the value they now see.
And so it's -- again, you can't jump to AI. You've got to put all the pieces, and a lot of clients aren't there yet, which is our opportunity.

This seems to corroborate with what KPMG consultant Todd Lohr said in this article from The Information.

“Very few large companies are going to use generative AI to automate customer service any time soon because there are still questions about how the AI will behave,” said Todd Lohr, a principal at consulting firm KPMG who advises multinational companies on automation. (KPMG resells Microsoft software to its customers.)
Most large companies are finding the cost-benefit analysis of using AI tools “challenging” because of the high cost of the software and the difficulty of measuring productivity gains, Lohr said. Instead, large firms are usually drawing from their existing software budgets rather than devoting new money to slowly expand their AI usage, he said.

We’re still at a point where the size of the IT budget pie is not expanding and any spend going towards generative AI is being pulled from lower priority categories. Despite all the hype and enthusiasm of productivity benefits from AI, customers are reluctant to embed these assumptions into their IT budgets. Until there is clear and measurable ROI, something else needs to be sacrificed. According to the CIO survey from Piper Sandler in December, the lowest priority software categories look to be HCM, Database and Marketing.

Another noteworthy comment from Sweet pertains to the constraints of modernizing the tech stack. Some customers wish to implement generative AI, but they first need to upgrade their data foundation. They must undergo significant transformations of their ERP platform. This is promising for cloud service providers and certain software providers, who benefit as customers expedite their migrations to utilize GenAI.

That’s all for this week. If you’ve made it this far, thanks for reading. If you’ve enjoyed this newsletter, consider subscribing or sharing with a friend

Share Tech takes from the cheap seats

I welcome any thoughts or feedback, feel free to shoot me an email at portseacapital@gmail.com. None of this is investment advice, do your own due diligence.

Tickers: AAPL 0.00%↑, GOOG 0.00%↑, TikTok, $UMG, META 0.00%↑, MSFT 0.00%↑, ACN 0.00%↑

So far, its efforts here are only a test as it is mindful of pissing off an important partner (the music labels)

Tech takes from the cheap seats