AI Results, and opportunities ahead
Quarterly results and AI
So Microsoft had a strong quarter — “Microsoft’s third calendar quarter earnings were the most robust in the company’s five-decade history,” with significant growth in Azure due in part to AI. Google's results were strong as well, attributing it to the cloud and AI. And the AI transformation is just starting — significant usage and revenue wins from AI are still in front of us. Much of this early momentum is from trials and developers starting to build the mainstream use cases; we aren’t even close to mainstream use yet.
AI usage is going to accelerate when it can help all of us with our personal information overload. I have 20 years of email and documents about our house — the initial plans, email exchanges with the architect about every aspect of design, email exchanges with the county around permits, the permit docs, receipts for construction items, as-built plans, documentation from the contractors, years of receipts for maintenance, email exchanges about maintenance items, etc. It’s a wealth of information but very unwieldy — a mix of PDFs, scanned images, word documents, excel documents, text files, etc.
I want to easily be able to ask questions like “Who last maintained our septic system, and how do I contact them?” Or “Do we have a conduit between the garage and house that I could pull an ethernet cable through?” Or “When was our well water last tested, and what were the results?” All the info is in there.
I have tried in the past to sort and sift through all my info and clean it up, but that is incredibly time-consuming. And isn’t that exactly what computers are for? I’ve built a couple of versions of a RAG application for my use case using llamaindex. They worked, but it was still a bit of a headache, and the results weren’t perfect. And the models and frameworks are moving so fast that my apps are obsolete even before I write them.
I just want to point at a directory of files and have an agent automagically created, informed by these docs. I’m not unique in this desire — Scott Hanselman, for instance, wants the same thing. This usage model applies to home records, medical records, estate planning, bills, etc. And even more business use cases.
This is the obvious thing for Microsoft or Google to do. They already have all my docs in their storage solutions. The latest OneDrive release doesn’t seem to be it. Google’s NotebookLM is getting closer — I can pull in 50 docs and query across them. But I have thousands of docs. Once Google or Microsoft crack this (or likely both of them), it will generate a ton of use. And Microsoft and Google have a huge advantage — I don’t need to move anything anywhere.
NVidia has done very well with the AI wave so far, but history (for instance, the PC revolution) suggests that more of the wave’s value will accrue to software and applications than to hardware. Microsoft and Google have great opportunities.
AI comes for energy
It is fascinating how the computing needs for AI are driving the tech and energy industries together.
Google is investing in nuclear, Amazon is investing in nuclear, Microsoft is investing in nuclear twice over, Microsoft is investing in geothermal, and Meta is investing in geothermal. Watching the relatively fast-moving tech industry culture collide with the slower-moving regulated energy industries will be fascinating. Perhaps dedicated commercial-use power plants can bypass some of the regulatory slowness.
AI and cloud energy use is still small compared to overall household energy use — maybe 2-3% of total USA energy consumption vs 21% for total household energy consumption. But it is probably the fastest-growing use, and on the margin, it will drive many power capacity decisions.
Moore’s law comes for every industry, one way or another — either it eats up your industry or becomes the dominant consumer of your industry’s output.
Working at Microsoft
Microsoft seems like a great place to spend part of one's career these days. Microsoft prides itself these days on having a culture focused on learning, and you need to be learning everyday in your career. I think I’d like working at the company again these days.
My own personal ride at Microsoft was a learning firehose. I joined the company in 1988, and I had a choice of a couple positions when I joined. The first was in the Languages Business Unit, working on what would become Visual Basic. A great team, and the product turned out to be a total winner. I kind of understood the product challenge; I was a user of language tools and had done all the usual things in college, like build a compiler and build an interpreter. So, it seemed like a reasonable fit. But I was changing industries and career paths; moving to Microsoft was a reset in a lot of ways.
So I decided to take the more challenging and uncomfortable route, a position in a business that was totally new to me. The Network Business Unit also made me an offer, working on server products, and I had never really worked on networking products in the past. I understood really low level concepts like signal processing, I understood the physical layer of networking, but everything else was new to me. So I took the job that offered the greatest potential learning opportunity.
I was also given the choice of being a product manager or a program manager. I hardly knew what these were, but the program managers seemed to spend a lot of time talking to developers, and the developers seemed like very smart people, so that was the path I chose. I spent the next 18 months drowning in learning. I generally felt like the dumbest person in the room; a great way to learn is to surround yourself with experienced, competent, intelligent people.
Very early in my job at Microsoft, as I was learning my way, and generally feeling like an imposter, my manager noticed my tentativeness and took me aside. He was (and is) a baseball fan, and he explained the job to me in a baseball metaphor. He told me I needed to get up and swing. And that I would miss a lot. And that is ok, because the business is a batting average game. Most of my swings would be failures. But each success would make up for that, and I would learn a lot from each failure, and the company was tolerant of learning by doing (as long as you learned!).
The Microsoft of today is a very different company than the Microsoft of 1988, but with its aggressive moves in AI, its cultural commitment to learning, its willingness to take on risk, and the business opportunity ahead of it – it seems like a great place to grow. The company may be 50 years old, but it seems like a pretty youthful 50 year old.
Comments ()