AWS (AI) Summit highlights, Temu and Amazon converging, and robotaxi adoption

14 July 2024 | Issue #27 - Mentions $AMZN, $CRM, $ZI, $PDD, $AAPL, $PYPL, $BIDU, $UBER

Jul 15, 2024

Welcome to the twenty-seventh 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.

Highlights from the AWS Summit

This week I attended the AWS Summit in NYC to learn more about what’s happening in the ecosystem. Below are some interesting quotes from the keynote:

Interesting stats

“Generative AI is probably the single largest shift in how we’re going to interact with data and information and each other probably since the advent of the very earliest internet, and organisations that invested in that early internet, including Amazon.com, have gone on to experience pretty remarkable growth over the past 30 years. And it’s my estimation that customers that are investing in generative AI today are going to experience multiple Amazons of growth in the next 30 years, and it has never been a more exciting time to be a technologist and to be a builder. I have never seen such enthusiasm and energy and investment from customers, probably since the earliest days of cloud computing itself.
We have hundreds of thousands of customers today that are running their AI and machine learning workloads on AWS and Bedrock, our service for building Generative AI applications is one of the fastest growing AWS Services of the last decade.
Across of all AWS, machine learning and artificial intelligence is now a multi-billion dollar annual run-rate business for AWS. It is growing remarkably quickly, and it has been extremely invigorating to see such extraordinary growth in generative AI usage in enterprises as well as super successful startups.”
96% of all AI/ML unicorns run on AWS and 90% of the Forbes AI 50 run on AWS

Why do customers choose AWS?

Meet (and exceed) security requirements for GenAI and their data. They don’t use any of the data that moves through their paid AI services to improve the underlying models and they don’t have humans reviewing any of the the information. You can control where the data lies inside your VPC and the flow of that information within the network.
Broadest AI capabilities. Infrastructure layer for building, training and tuning foundational models (GPUs, Trainium, Inferentia, Sagemaker). Platform layer with Bedrock that makes it easy for anybody to build applications using GenAI. Application layer is Amazon Q, a GenAI-powered assistant that works across AWS applications.
New features at a rapid clip. Launched 326 GenAI features into general availability since 2023, more than 2x the other major cloud providers combined in the last 18 months.
Data already resides on AWS. There are many successful start-ups and enterprises in the exabyte (1M terabytes) club on AWS in terms of data stored.
Sustainability and renewable energy approach. Amazon matches all of the electricity used across the whole of Amazon data centres, grocery stores, fulfillment centres and corporate offices with 100% renewable energy, a goal they hit seven years early.

on multi-models

“And it seems intuitive to me that it would be impossible to address al of these (ideas), or even a small number with a small number of existing models, because there is no one model that is the best fit for everything. Instead what you want is to be able to have access to a very broad set of models, a broad selection of models where you have the choice to be able to pick the right model for your use case, and to be able to mix and match the models, such as you find the best fit which fits your use case. So some of these models are good at different things. Sometimes they’re really good at summarisation and natural language tasks. Others are really good at reasoning or code generation, and others just have different operational characteristics. Sometimes they’re really really fast. Other times they’re really really intelligent. Other times they’re just really really really low cost. And depending on your use case, you want to be able to lean into the advantage of each of these different models, such as it fits your use case, and then you’re able to cover a much broader set of use cases inside your organisation and drive innovation as a result, much more quickly and much more ubiquitously. And it’s this insight that led us to follow a slightly different strategy to some others.
…we make available by far the widest selection of models available anywhere.”
“The vast majority of customers, 97% in this survey of a couple of dozen of large enterprises that were spending multiple millions on generative AI every year, the vast majority are using not just multiple models, but are able to combine and build intelligent systems by pooling models from multiple developers all together. So this compounding multiplier of intelligence is only available when you have this broad model selection. And so multiple models is super important not just in isolation, but it’s incredibly powerful in combination. And it’s a big part of why customers are migrating and running on Bedrock as quickly as possible.”

Some of the top product announcements from the event here

Interesting tidbits from the floor:

Speaking to a Bedrock solutions engineer, he noted that large AI projects are starting to go into production from his customers. A notable one included a large bank deploying a tool for internal use. I mentioned reading an article (where I discussed here) that some AWS customers are going directly to OpenAI to use their models. He said this is done through Sagemaker, and by doing so you sacrifice latency, data privacy and the ability to quickly switch models for your use cases.

I also spoke to a Salesforce sales engineer who handles 90 enterprise accounts on the Service Cloud side. He said across his customers, they’re seeing a 10% reduction in seats / headcount because Einstein is making their customer service agents that much more efficient. There is a slight offset from revenue coming in on Data Cloud, but otherwise they haven’t shifted the pricing model on Service yet so they are slightly losing out. Einstein was released a year ago and they’re getting lots of interest from existing customers. It’s all they get asked about.

A ZoomInfo sales person highlighted to me interest from customers in its co-pilot product. It’s also been very popular internally with its own salespeople who are finding tremendous efficiencies and time-savings. One example was being able to ask the co-pilot to generate a summary of all previous interactions with a customer (across emails and call transcripts) to quickly get up to speed before having a follow-up conversation which saves hours of listening / reading transcripts.

It looks like GenAI is starting to deliver some impressive results when it comes to efficiency gains. Companies are seeing real benefits in streamlining processes and boosting productivity through AI-powered tools and systems. The challenge now lies in adapting pricing models to capitalize on this new paradigm. If businesses can successfully align their pricing strategies with the value created by GenAI, we could see a significant uptick on the revenue side as well.

Temu and Amazon business models are converging

The Information has provided excellent coverage on Temu and Amazon’s evolving e-commerce models where I discuss here. In a piece this week they detail how Temu is encroaching onto Amazon’s domestic business.

Chinese bargain site Temu has been a big hit with American shoppers in the past year or two, drawing increasing scrutiny from U.S. lawmakers over its approach to importing goods and attention from rival Amazon. Now, Temu is changing tack, encroaching more on Amazon’s turf.
Instead of only shipping items from Chinese factories directly to customers’ homes, avoiding paying U.S. duties along the way, Temu is signing up Chinese sellers that already have inventory in U.S. warehouses and can ship to homes from there, as well as U.S. merchants. Some of the sellers that Temu has signed up recently are also big on Amazon’s marketplace.
The new sellers now contribute roughly 20% of the gross merchandise volume on its U.S. site, according to two people close to Temu. Last year, Temu generated $17 billion gross merchandise value globally, about $10 billion of that in the U.S., estimates Bernstein analyst Robin Zhu.

$2bn in GMV might seem like a drop in the ocean compared to Amazon's numbers - it's just a fraction of the $238bn1 generated from Chinese sellers on Amazon, and even tinier when stacked against Amazon's worldwide GMV of $700bn. However, it's not the size that's catching attention here, it's the trend.

In its new approach, Temu signs up merchants with U.S. warehouses and lets the merchants handle their own delivery. That has allowed the company to broaden the range of items it can offer to include bulky items such as furniture.
“For Temu, there are some fears that there’s writing on the wall with the de minimis rule closure,” said Bernstein analyst Mark Shmulik. “So they want to make sure they’ve got a broader offering with local merchants and local fulfillment.”
While reliable shipping times could be an issue for U.S.-based sellers in the future, Temu may be able to handle any customer fallout, given how well established it is now.
Temu’s efforts come just as Amazon is moving in the opposite direction, testing a Temu-like discount section of its website that features goods shipped directly from warehouses in China and delivered between nine and 11 days, The Information reported last month. Amazon’s proposed bargain section will even utilize the de minimis provision used by Temu and Shein—a model Temu is now trying to move away from.
“Temu wants to be more like Amazon, to reduce supply chain risk and offer faster shipping, while Amazon is racing to replicate Temu because it cannot ignore the fact that its low prices trump Amazon’s convenience for some shoppers,” said Juozas Kaziukėnas, founder of Marketplace Pulse, an e-commerce consultancy.
A Temu spokesperson said in a statement that the company enables qualified sellers to manage their own logistics and ship products from local warehouses to help “meet different consumer needs with tailored fulfillment solutions.” “By using local warehouses, we can work with more sellers, offer a wider range of products and deliver orders faster,” the spokesperson added.
To attract sellers, Temu offered shipping subsidies to merchants. Temu has also created a dedicated in-house team that onboards U.S. sellers, according to a Temu seller presentation seen by The Information. It’s unclear how long the subsidies will last.
And Temu heavily promoted these new sellers’ listings. Its website and app in the past three and a half months has prominently featured their products, with a “local warehouse” label in green type plastered on every product photo to distinguish it from those shipped from China. Listings for items that ship from China also show similar products from local warehouses to entice shoppers to switch.
Products from local warehouses get delivered faster than similar items shipped from China—as fast as four days, compared to six to 22 days for Temu’s standard shipping and four to 11 days for express shipping. While there is a price difference, it’s usually marginal.
Several China-based Amazon merchants told The Information they are exploring joining Temu’s local warehouse program. These merchants had previously shunned Temu, which is known for keeping a tight grip on its Chinese suppliers, dictating what products they can sell and at what prices, The Information reported previously.
While Temu still retains the ultimate say in product pricing, it has been more lenient with the pricing it sets, according to one of the people close to Temu and various social media posts shared by Temu sellers.
The big Amazon sellers Temu has attracted include Anker, an electronics maker based in central China’s Hunan Province. Anker started selling on Amazon immediately after its founding in 2011. Its Amazon store was so successful that the company eventually went public on the Chinese stock market nine years later, and Amazon’s advertising site has showcased its growth story.
Anker joined Temu’s local warehouse program nearly two months ago, and has been offering some of the same products it also sells on Amazon at a lower price. An Anker bluetooth speaker currently sells for $25.99 on Temu and $36.99 on Amazon.
The same goes for Costway, a Fontana, Calif.–based furniture merchant operating on Amazon that joined Temu four months ago. A Costway three-piece patio rattan sofa set now sells for $135.99 on Temu, 32% cheaper than on Amazon.
This could contravene Amazon’s fair pricing policy, which says it can take steps including removing the “Buy Now” button from listings by merchants who set prices “on a product or service that is significantly higher than recent prices offered on or off Amazon.” Some sellers and the Federal Trade Commission have said the policy effectively blocks merchants from offering lower prices on other sites.
Amazon says the policy “applies to protect customers from harmful practices, such as price gouging, and does not apply to price competitiveness.”
Meanwhile, the new approach has hurt some of Temu’s China-based suppliers. Two of those suppliers told The Information they have seen sales stagnate or decrease as Temu diverted traffic away from their products after launching the local warehouse program.
The Temu spokesperson said local warehousing will improve shoppers’ experience, and the company’s goal is to “help all sellers reach more customers and grow their businesses with efficient shipping solutions.”

Temu's marketplace is presenting Amazon with a new breed of competitor. Amazon's journey from a first-party book seller to a vast third-party marketplace spanning countless categories has been impressive. They've traditionally outmaneuvered brick-and-mortar stores with competitive pricing, wider selection, and convenient delivery. As Amazon's sales skyrocketed, so did the proportion of Chinese sellers on their platform, capitalizing on US consumer demand. JPM's data shows Chinese sellers' GMV share grew from 34% in 2019 to 45% in 2023. Now, Temu is actively recruiting these sellers with lower take-rates and merchant subsidies - an attractive proposition given the historical lack of viable alternatives. Interestingly, it's unlikely that incremental sales are the main draw for these merchants. After all, it's hard to imagine a US consumer using Temu but not Amazon. This suggests that merchants can likely maintain their profit margins on Temu despite listing prices being about 30% lower, once you factor in the subsidies and lower fees. Amazon might be breathing a sigh of relief that Temu is shifting focus away from the US market. If Temu were to gain significant traction, it could have long-term implications for Amazon's take-rates and GMV growth. This situation underscores the evolving dynamics of e-commerce and the ongoing challenge for platforms to balance merchant attraction, consumer value, and their own profitability. It also highlights the critical role of execution in this competitive landscape.

Apple yields to the EU on payments

With the EU regulator dropping new probes into big tech left and right, it can be tough to keep up. This one on on Apple looks to have come to a close.

The European Union said that Apple’s AAPL 1.31% decision to let third-party mobile wallet and payment services use the technology behind its Apple Pay app fully addressed its antitrust concerns, ending a long-running probe that could have seen the iPhone maker hit with a hefty fine.
The European Commission, the EU’s executive arm, said it had made a series of commitments by Apple legally binding under the bloc’s antitrust rules, meaning that Apple now must comply with them by July 25.
European officials were concerned that the tech giant was stifling competition by keeping rivals out of its payments ecosystem. The bloc opened a formal antitrust investigation in 2020 to determine whether Apple’s policy to restrict access to so-called near-field-communication—the software and hardware that let mobile devices communicate with pay terminals—violated competition rules.
“From now on, competitors will be able to effectively compete with Apple Pay for mobile payments with the iPhone in shops,” said Margrethe Vestager, the EU’s competition chief. “Consumers will have a wider range of safe and innovative mobile wallets to choose from.”
In response to the EU’s concerns, Apple said earlier this year that it would allow other companies’ apps to make contactless payments on iPhones and other devices that use its iOS operating system.
Vestager said the commitments marked the end of the EU’s investigation into Apple Pay, though the company now has to make good on its promises or it risks a fine of up to 10% of its annual revenue or 5% of daily revenue for every day of noncompliance.
The company said users would be able to make a transaction with Apple Pay and Apple Wallet, or through another app on its iOS mobile operating system, meaning that consumers can choose and change their default contactless transaction app at any time.

This development is indeed a positive turn for digital wallet providers like PayPal, who've been hamstrung in their in-store payment game. Their current process of scanning QR codes is clunky compared to Apple's sleek double-click and tap method. The EU's decision could level the playing field, giving consumers more options at the checkout. It's not just about convenience; this could spark some healthy competition, potentially leading to perks or subsidies as companies vie for consumer loyalty. It's refreshing to see the EU chalk up a win after a string of regulatory misses. This particular probe seems to have hit the mark, potentially fostering innovation and competition in the digital payments space. The real winners here are the consumers. More choice usually translates to better services and potentially lower costs. It'll be interesting to see how quickly other wallet providers can capitalize on this opportunity and how Apple responds to the increased competition.

The bumpy road to robotaxi adoption

From the FT

“Well, this is one way to drive new tech growth. China’s answer to Google is betting big on robotaxis in China to revitalise its business. So far, that is working. Baidu has a head start on Tesla and the faster than expected take-up of its robotaxi Apollo Go has boosted shares in the search engine and artificial intelligence group. But Baidu’s progress faces risks at home. Baidu is leading the autonomous driving market in China. Its robotaxis are already in operation: its biggest presence is in the city of Wuhan where there is a fleet of 300 of the cars. Users can hail an unmanned, fully self-driving car for local taxi services. The 24-hour operations started this year. This business is just what Baidu needs. Its revenue grew at its slowest pace in more than a year in the March quarter up just 1 per cent, despite its lead in AI in the country. Monetising that technology will take time, while its ad revenue has long been cut into from competition from rivals such as TikTok owner ByteDance. Hong Kong-listed shares of Baidu are down more than 30 per cent in the past year
It will not be long, however, before Baidu starts to face stiff competition in this sector. In addition to local rivals Pony.ai, AutoX and WeRide, Tesla is reported to have been granted approval to test its advanced driver-assistance system in Shanghai, with its robotaxi expected to be unveiled next month So far, Baidu’s head start in mass production of robotaxis is giving it an edge. It had already partnered with Nvidia to develop autonomous cars and robotaxis in 2016 and launched Apollo, its open-source platform for autonomous driving, a year later. Those years of development and trial and error mean that it has managed to slash the cost of producing robotaxis: making its latest generation models costs less than half that of its previous version. But there is a flipside to Apollo’s unexpected popularity: an equally strong reaction from local taxi drivers. Robotaxis are typically cheaper than traditional taxis. Taxi drivers have been petitioning for Apollo to be limited. Social tensions in China have already been rising over the past year due to high levels of youth unemployment: the rate for those aged between 16 and 24 was at 14.2 per cent in May. Even as rivals pile in, that pushback looks a serious potential threat to this nascent sector’s growth.”

I suspect we'll see this theme play out repeatedly in the coming years as robotaxis expand their reach. It's reminiscent of the early days of ride-hailing, where local regulators in China and the US initially pushed back to protect the traditional taxi industry, despite clear benefits to consumers. With robotaxis, the stakes are even higher.

While I believe the technology will eventually be solved, mass deployment across countries will likely face significant delays. The challenge lies not just in the tech, but in coordinating infrastructure upgrades and navigating a complex web of regulations.

Local authorities will be juggling competing interests - the promise of efficient, futuristic transport against concerns about job losses, safety, and disruption to existing systems. Some cities might embrace robotaxis for their potential to reduce traffic and emissions, while others might resist, fearing the impact on local employment or struggling to adapt outdated infrastructure. This balancing act could result in a patchwork adoption, with robotaxis becoming common in some areas while remaining rare in others, even within the same country.

Waymo's rapid progress from zero to fully operational in San Francisco has been impressive, but I'm skeptical this pace can be replicated nationwide due to these varying local factors. China might have an edge with its centralized and unified government in decision-making, but even they’re facing pushback, as we’ve seen with the response to Baidu’s efforts.

Interestingly, this complex landscape might actually benefit ride-sharing leader Uber. Despite not having their own autonomous driving technology, the slower and uneven adoption of robotaxis could give them more time to adapt and maintain their market position.

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

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%↑ , CRM 0.00%↑ , ZI 0.00%↑ , PDD 0.00%↑ , AAPL 0.00%↑ , PYPL 0.00%↑ , BIDU 0.00%↑ , UBER 0.00%↑

According to JPM estimates

Tech takes from the cheap seats