MBI Daily Dose (July 14, 2025)
Companies or topics mentioned in today's Daily Dose: AWS growth, Interest rate in AI boom scenario, LLM Adoption at work
As Anthropic scaled its annualized revenue from $1 Billion in December 2024 to $4 Billion by mid-2025, people are starting to wonder what it means for AWS revenue growth. Morgan Stanley (MS) recently published a note estimating Anthropic’s potential impact on AWS revenue.
MS assumed Anthropic’s revenue will increase from $4 Billion in 2025 to $10 Billion in 2026 and $19 Billion in 2027. If you assume Anthropic’s gross margin is ~60% and three-quarter of Anthropic’s cost of revenue is spent on AWS, Anthropic alone could contribute $5.6 Billion revenue in 2027. That would still make it only LSD revenue mix of overall AWS, but given how many investors are super focused on betting on stocks that are about to experience “accelerating” topline, even such incremental contribution can matter for the stock in the near term.
How about the long-term impact given that’s much more important for long-term shareholders? It’s no denying that Anthropic’s recent momentum has been pretty impressive, but long-term questions still feel quite murky to me. Coding has been a a very popular use case in Anthropic, but if there are multiple SOTA models around in 5 years and companies such as Cursor can still offer multiple model alternatives to choose from, will model developers cede more gravity to the layer above them over time? Cursor is the big name in this space, and their CEO Michael Truell’s recent interview at Stratechery had some interesting angle in terms of how Cursor thinks about this space in the long term. Notice this below excerpt:
Ben Thompson: Is that a real sustainable advantage for you going forward, where you can really dominate the space because you have the usage data, it’s not just calling out to an LLM, that got you started, but now you’re training your own models based on people using Cursor. You started out by having the whole context of the code, which is the first thing you need to do to even accomplish this, but now you have your own data to train on.
Michael Truell: Yeah, I think it’s a big advantage, and I think these dynamics of high ceiling, you can kind of pick between products and then this kind of third dynamic of distribution then gets your data, which then helps you make the product better. I think all three of those things were shared by search at the end of the 90s and early 2000s, and so in many ways I think that actually, the competitive dynamics of our market mirror search more than normal enterprise software markets.
One other interesting bit is at one point Truell mentioned in that interview, “We’ve actually kind of already seen that where some of the most enthusiastic and loud users of these tools are people who are slightly technical or not really technical”
Shopify’s Head of Engineering pretty much corroborated that in a recent interview:
“We’ve been using AI tools a long time in engineering. We were the first to use GitHub Copilot outside of GitHub in 2021, we weren’t charged for 2 years and in exchange we gave them lots of feedback.
We then deployed Cursor 1 year ago. We’re trying all these things to see what’s working and what’s not and whatever works we let more people use it.
The most interesting thing about Cursor is that the growth in Cursor at Shopify is happening a lot outside of engineering and R&D. Finance, Sales, Support, those are the teams using Cursor.”
Of course, Anthropic has Claude Code, and Google just “licensed” Windsurf which was supposed to be originally acquired by OpenAI. Google also already has Gemini CLI and with Windsurf in the wing, I am expecting a vigorous attempt by Google to dominate this space. I’m sure OpenAI will still launch something on their own.
The opportunity for these tools may be quite large given the technical bar to develop a software will perhaps exponentially go down; nonetheless, it remains a bit hazy to think how the competitive dynamics will be settled here in a couple of years. So, it remains to be seen whether it makes sense to be enthusiastic about Anthropic’s momentum that may result into AWS growth acceleration in the next few quarters.
Seth Benzell at “Empiricrafting” examined whether a potential AI-driven economic boom could make large government debts more sustainable. His simulation suggests that for the US, the resulting GDP growth could help manage debt, but this is complicated by rising interest rates and negative impacts on developing nations. Ultimately, he concludes that the high potential of AI makes it even more crucial to reduce government debt to avoid "crowding out" the private investment needed to realize that potential.
It’s a bit speculative piece, but I think is worth highlighting that interest rate for govt financing could continue to increase if AI boom actually materializes (I have noticed plenty of people assume the opposite). Some excerpts from the piece:
In 2017 the government’s borrowing costs were at record lows (1.2% of GDP), but have exploded upwards since. In 2024, the US spent over 3% of GDP paying interest on the national debt.
This line item will not get smaller with the AI productivity boom. Instead, the opposite. This is because an AI boom will make it costlier for the government to borrow money.
The countries that benefit the most from AI will be the ones with (currently) high wages and TFP, and low costs of capital — the US is the lead example here, with high wages, a well developed financial system, and low corporate taxes.
…when the marginal cost of capital is high, that means there are productive uses for it that are being unexploited! Therefore, it’s even more important to save when interest rates are high.
In addition to "Daily Dose" like this, MBI Deep Dives publishes one Deep Dive on a publicly listed company every month. You can find all the 60 Deep Dives here. I would greatly appreciate if you share MBI content with anyone who might find it useful!
A recent paper looked at LLM adoption at work, and the pace has clearly been picking up:
In our own nationally representative surveys of U.S. workers, we find that LLM adoption at work among survey respondents above 18 has increased rapidly from 30.1% as of December 2024, to 43.2% as of March/April 2025, and to 45.9% as of June/July 2025, a substantial increase in 2025 which we find to be largely due to an increase in ChatGPT and Generative AI use.
…Among those who report to use Generative AI at work, about 33% claim to use it every workday, while selectively using it for a limited number of hours per week. Generative AI seems to decrease the time spent on a task by an average of one hour, resulting in tripling the productivity in tasks where this technology is used.
While more and more people are using LLMs at work, only mid-teen percentage of people use it daily. As you know, these data tends to be Rorschach test. The AI optimists focus on accelerating adoption whereas skeptics probably would highlight DAU being only mid-teen percentage of users despite broad awareness. It is an open question what percentage of workers will religiously use AI if it requires as much agency as it currently does on users’ part (formulating proper prompts, uploading files, verifying claims/data etc.)
I do think DAUs will be lot higher than mid-teens in 3-5 years, but a lot of unlock may need to happen through lowering agency required to become core part of people’s workflows. That’s precisely why OpenAI (and others) wants to launch browser and hardware products.
Current Portfolio:
Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.