Hyperscalers forced to play chicken, not all AI revenue is equal and EU regulations

8 July 2024 | Issue #26 - Mentions $AMZN, $MSFT, $ACN, $META, Shein, $PDD

Welcome to the twenty-sixth 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.

AI capex chicken

As another week goes by, we public market investors get another datapoint on the biggest driving theme at play in the market right now. I had tweeted about this early in the week, but the WSJ wrote a piece on Amazon's data centre investment plans. They managed to snag some notable quotes from John Felton, Amazon's CFO, that shed light on the company's strategy.

Amazon is planning to spend more than $100 billion over the next decade on data centers, an impressive level of investment even for a company known for its spending ways. The Seattle company is now devoting more investment money to its cloud computing and AI infrastructure than to its sprawling network of e-commerce warehouses.

Amazon Web Services, the arm that manages Amazon’s cloud business, has opened data centers for years, but executives said there is a surge in investment now to meet demand triggered by the excitement around AI.

“We have to dive in. We have to figure it out,” said John Felton, who took over as AWS’s chief financial officer this year after spending most of his career in Amazon’s retail fulfillment operations.

The company’s financial commitment reflects the importance and high costs of AI. Felton said building for AI today feels like building that massive delivery network in years past. “It’s a little uncertain,” he said. AWS is expanding in Virginia, Ohio and elsewhere.

Amazon’s spending on data-center capital expenses, including leases, compared with total capital expenditures hit a decade-high last year of 53%, according to market-research firm Dell’Oro Group. Amazon said it expects AWS infrastructure spending to remain high this year, and the company has announced many AWS investments in recent months.

The capex figures seem to back up my initial thoughts on the current data centre investment cycle at big tech companies - it's looking like a pull forward. Amazon's Q1 24 capex hit about $15bn, and if we assume over half of that goes to data centres as the article suggests, we're looking at a run-rate of around $33bn. Let's round that up to $40bn for 2024, accounting for growth throughout the year. John Felton's statement that the company expects to spend over $100bn in the next decade is interesting. We can deduce it'll likely be under $200bn (or they'd have said as much), which implies Amazon plans to spend an average of at most $19bn yearly on data centres over the next decade. That's less than half of this year's projected spend. Meanwhile, consensus estimates have Amazon shelling out nearly $550bn in total over the next seven years, with potentially $360bn of that going to data centres (assuming the split leans more towards data centres versus logistics). These numbers don't quite add up, suggesting there might be some discrepancies between Amazon's stated plans and market expectations.

Source: Koyfin

I can see where Felton's coming from, though. As CFO of one of the world's most crucial divisions right now, they need to project a willingness to invest in a platform capable of hosting cutting-edge AI tech. But he's also got to walk a tightrope, balancing these statements with ROI considerations and managing investor expectations to show some fiscal responsibility. Amazon's reported multibillion-dollar revenue run-rate from AI in their latest earnings seems like small change compared to the $100bn+ they're planning to spend. Yet they can't afford to be perceived as lagging in AI. They're the leader in hyper-scale computing with nearly 50% market share, but their slow start in GenAI has cost them. They've lost ground to Microsoft and Google as companies gravitate towards infrastructure that's seen as more "AI forward". It's a delicate balancing act for Amazon - staying competitive in the AI race while not overcommitting financially. This disconnect between stated plans and market realities highlights the complex challenges tech giants face in the rapidly evolving AI landscape.

In a sense, Amazon's caught in a game of capex chicken. Even though AI isn't bringing in much revenue yet, this massive data centre build-out is like an entry fee to keep their core hyper-scale business in the game. And right now, investors are currently rewarding companies for this kind of spending. I've written before about how Microsoft's market cap has actually benefited, despite revenue and EPS estimates being lower than pre-ChatGPT levels. In DCF speak, investors are discounting the potential cash flows from these AI investments at a lower rate than the overall market, even with the higher capex. It's as if they're betting on a bigger future payoff. It's a classic case of "dance while the music's playing" in the tech world. With AI being the hottest tune right now, these companies feel they have no choice but to join the dance floor, even if the cost of admission is steep. The fear of being left behind seems to outweigh the immediate financial considerations, at least in the eyes of the market.

These next 12-18 months will be crucial. Amazon needs to show significant AI-related revenue growth, or they'll have to dial back their capex guidance. Investors will be parsing every word from execs with a fine-toothed comb. Interestingly, Dario Amodei mentioned in his "In Good Company" interview that while training models costs about $1bn now, it could skyrocket to $10bn or even $100bn by 2025-2027. If that's the case, someone might want to give Felton at AWS a heads-up - after all, they're Anthropic's cloud provider of choice.

Related: Why AI is so expensive

Not all AI revenue is created equal

I touched on Accenture's GenAI leadership a couple weeks back when they hit that impressive $3.6bn revenue run-rate. While the company described these as smaller-scale projects with clients dipping their toes into AI, Benedict Evans' podcast shed some more light on the actual scope of these initiatives.

“No one wants to spend on the old stuff while they work out what the new stuff is. But meanwhile, Accenture, people have spent... Well, last 12 months, $2.25 billion.

Current run rate, $4 billion. Now, so there's a bunch of like, let's see what the hell this is. Now, meanwhile, I think BCG said this will be 20, 25, 20% of their revenue this year.

And so two observations on this, and I'm concerned I'm on logging, but two observations. One of them is that it essentially gave another number in the August quarter, at which point they'd done 300 million. And a little bit later in the call, they said, we've done 300 projects.”

The project sizes really do point to this being an experimental phase. With consulting fees typically eating up 50-70% of a project's budget, there's not a whole lot left for hardware and software in these $1m projects. It suggests we're looking at strategy outlining and maybe some model fine-tuning, rather than full-scale implementations. We're still a long way from seeing large-scale production deployments that could become steady revenue streams for Accenture. But it's worth noting that we're not entirely sure how much of Microsoft's AI-related revenue is experimental either. Modest Proposal on Twitter took a stab at breaking down the numbers using various online sources, and it's possible that a good chunk of their API and OpenAI on Azure revenue could also be from companies in the experimentation phase.

The EU doesn’t like chess

It looks like Meta's attempt at 4D chess with its "consent or pay" model to sidestep EU privacy regulations has hit a snag. I previously discussed how this strategy seemed like a clever workaround, but it appears the regulators aren't buying it.

Meta Platforms (META.O),  was charged by EU antitrust regulators on Monday for failing to comply with landmark tech rules as they took aim at the U.S. company's newly introduced pay or consent advertising model, already the target of privacy regulators and activists' ire.

The tech giant launched the no-ads subscription service for Facebook and Instagram in Europe last November, saying users who consent to be tracked get a free service which is funded by advertising revenues. Or they could pay for an ad-free service.

The European Commission, which acts as the EU competition enforcer, said the binary choice breaches the bloc's Digital Markets Act (DMA) which seeks to rein in the power of Big Tech, as it sent its preliminary finding to Meta.

It said the binary choice forces users to consent to the combination of their personal data and fails to provide them a less personalised but equivalent version of Meta's social networks.

"We want to empower citizens to be able to take control over their own data and choose a less personalised ads experience," EU antitrust chief Margrethe Vestager said in a statement.

Meta said its model complied with a ruling from Europe's top court.

"Subscription for no ads follows the direction of the highest court in Europe and complies with the DMA. We look forward to further constructive dialogue with the European Commission to bring this investigation to a close," a Meta spokesperson said.

Ben Thompson wrote a good piece here on why the EU might be overstepping here.

This fight between big tech and the EU is definitely one to watch. As these companies see their revenue and profit pools from the region shrink in importance, the threat of fines up to 10% of global revenue could start tipping the scales. There might come a point where these tech giants decide the EU market isn't worth the hassle. If it came to that, the ramifications would be massive. We could potentially see the rise of EU internet champions, similar to what's happened in Russia or China. It's probably not going to go that far, but it's an entertaining scenario to consider.

More countries cracking down on cheap cross-border goods

From the FT

Brussels is drawing up plans to impose customs duties on cheap goods bought from Chinese online retailers including Temu and Shein in an effort to stem a surge in what the EU says are substandard items coming from China.

The European Commission later this month will suggest scrapping a current €150 threshold under which items can be bought duty free, three people briefed on the matter told the Financial Times. The main platforms being targeted were China’s online marketplaces Temu and AliExpress and clothing retailer Shein, one official said.

Last year, 2.3bn items below the duty-free €150 threshold were imported into the EU, according to the commission. Ecommerce imports have more than doubled year on year, topping 350,000 items in April — or almost two deliveries per household, commission data shows.

China benefits from subsidised postage costs, meaning it is cost-effective to send cheap goods by air.

The provisions would apply to any online retailer shipping to EU customers directly from outside the bloc. US-based Amazon typically uses sellers based in Europe.

There's been a growing trend of countries either nixing duty-free thresholds on imported goods or slapping on new tariffs. The Economist recently put out a good piece that helps put China's situation into perspective.

To reverse an economic slowdown and cement its control over global supply chains, its leaders have launched an investment spree in high-tech goods, such as batteries, electric vehicles and other green devices. Weak domestic demand for traditional products, such as cars, chemicals and steel, mean they are also flooding global markets. The average price of Chinese manufactured exports fell by nearly 10% from 2022 to 2023. China’s export volumes have surged to near-record levels.

Countries are tightening the screws on imports to stem the tide of cheap, potentially low-quality goods and shield local businesses. They're turning to tariffs as their weapon of choice. But here's the kicker - even with these extra charges tacked on, some Chinese products are still managing to undercut the competition.

This emerging-market attempt to lower trade barriers with the West is happening at the same time as they are being raised with China. Officials see this as necessary to protect domestic manufacturers until China’s subsidy wave subsides. “In the [late 2000s], Mexican companies would ask for protections and the government would tell them…‘well, you have to learn to compete’,” says Mr Guajardo. “That is no longer the case.” Mexico raised tariffs on 544 products in April. It has slapped an 80% levy on certain steel imports.

Yet some Chinese goods are so cheap they have the lowest prices even with sky-high tariffs. Moreover, some products sneak past levies because they are packaged in third countries. That is why non-tariff barriers and import bans are also proliferating. India has launched anti-dumping probes into a variety of products, including unframed glass mirrors and fasteners, which it says will protect its small and medium-sized businesses. It has also filed the most anti-dumping cases of any country in the world. China is retaliating. Sumant Sinha, boss of ReNew, an Indian green-tech firm, says it is even quietly blocking India’s access to solar equipment.

The EU's legislative push to tackle the influx of potentially dangerous goods comes at an interesting time, especially for companies like Shein and Temu.

Shein's gearing up for an IPO, while Temu's shifting focus away from the US market. This EU crackdown puts both in a bit of a pickle, as they can't afford to ignore the European market.

The number of dangerous products reported by EU countries jumped more than 50 per cent from 2022 to 2023 to more than 3,400. Cosmetics, toys, electrical appliances and clothes were among the products with the most safety issues.

The EU toy industry has accused Chinese retailers of shipping dangerous toys to Europe. Toy Industries of Europe, an industry group, said in February it bought 19 toys from Temu and found that none complied with EU standards, while 18 presented a real safety risk for children.

Temu said that “all 19 product listings are no longer available on our EU website”. It added that “product safety is of paramount concern to us and we have strengthened the monitoring of this product group and its associated requirements”.

Christel Delberghe, director-general of EuroCommerce, a retail lobby group, said: “We want an EU-level playing field in online retail for all players targeting EU-based consumers, no matter where they are established.”

She said existing legislation was sufficient but it needed “an effective and efficient cross-border enforcement strategy”.

Threads (and this newsletter) turns one

It's hard to believe it's been almost a year since I penned my first post. I initially started writing because I disagreed with the internet's premature declaration that Threads was dead after just 10 days. Looking back over the year, I realize some of my best posts have been those where I challenged the consensus view. While I'm still figuring out the long-term direction for this newsletter, I'm incredibly grateful to all of you who've taken the time to read and provide feedback. Your engagement has been invaluable in shaping this journey.

I enjoyed Casey Newton's interview with Adam Mosseri about Threads' first year. The app's definitely not dead, but it's clear there's room for improvement in user engagement. By my calculations, the app's DAU retention after a year still places it in the top 10% of apps. However, its DAU/MAU ratio falls short of the impressive 67% seen in Meta's core Facebook and Instagram apps. Threads is still in the red financially, and it hasn't reached the scale where implementing ads would be more beneficial than further enhancing Instagram. This puts the app in a tricky position where growth is crucial. That said, it's worth noting that Threads has amassed nearly the same number of MAUs¹ as ChatGPT in just a year {footnote: According to the website referenced, ChatGPT had approximately 180 million MAUs as of July.}. This suggests the app has some staying power, even if it's not yet living up to its full potential.

“I would love to do it sooner rather than later,” said Mosseri, who also oversees Threads, in an interview last month. “It’s just really a question of opportunity cost. Is that the best way to drive business versus making Instagram ads a little bit better on any given month? But it’ll happen, and hopefully sooner rather than later.”

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

This is a free publication but if you’d like to support my work, please consider buying me a coffee. 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: AMZN 0.00%↑ , MSFT 0.00%↑ , ACN 0.00%↑ , META 0.00%↑ , Shein, PDD 0.00%↑

1. According to https://explodingtopics.com/blog/chatgpt-users, ChatGPT had approximately 180 million MAUs as of July.