Excellent, just excellent. This is the piece that made me a subscriber. Fewer than 10 people in discretionary active management could have written this. I know I couldn't and I've been at this for 30 years and in the AI space for 3 years. I've been building AI analytics, and here is the launch of positive revisions for your NVDA example, right on cue. In of course, .json :)
date": "2023-01-05",
"transcriptLabel": "NVDA JPMorgan CES Tech/Auto Forum",
"keyTopics": [
"Introduction of new Ada architecture GPUs (4090, 4080, 4070 Ti) driving the gaming cycle",
"Integration of cloud gaming (GFN) with automotive applications",
"Continued advancements in data center platforms and automotive solutions leveraging Orin"
],
"metrics": {
"ProductAdoption": "High demand for 4090; robust DLSS 3 implementation"
},
"notableQuotes": [
"Our 4090 is tight in availability and 4080 pricing is marked up relative to MSRP, reflecting strong demand."
]
},
{
"date": "2023-02-22",
"transcriptLabel": "NVDA Q4 2023",
"keyTopics": [
"Stable data center growth driven by H100 uptake and generative AI emphasis",
"Mixed segment performance with recovering Gaming and record Automotive revenue",
"Launch of NVIDIA AI cloud services (DGX Cloud) and integrated AI platforms across public/private clouds"
],
"metrics": {
"QuarterRevenue": "$6.05 billion",
"FullYearRevenue": "$27 billion",
"DataCenter": "$3.62 billion",
"Gaming": "$1.83 billion",
"Automotive": "$294 million"
},
"notableQuotes": [
"H100 is as much as 9x faster than the A100 for training and delivers up to 30x faster inferencing for transformer-based models.",
"Generative AI applications will help almost every industry do more faster."
Drew! This is massive. It looks like you are already working on this. And I didn't go into it in the post, but does it change how we write research reports? Do reports get built (possibly in JSON), for other AI's to read? Do we need to optimise for AI instead of human use? You see it already with Google Search. I keep looking at the AI summary and do less clicking. Graham never looked at the source documents the AI was using.
Hi Erik, I am working on this, and here is a tidbit of what I see in the space. Incumbent firms know something is afoot with AI but have no idea what to do next. So, there are slide decks on AI on execs desks built by first-year MBAs who got assigned the job to "figure out how to do AI." I think the default position is to wait until one of the major terminals figure it out. Meanwhile, the frontline analysts are Grokking, and Perplexity-ing which is good and bad. I have a prosumer app called DoTadda Knowledge that uses a corpus of 350,000 conference call transcripts that gets updated in real time. It works just like a chatbot. It can output formatted text for a human, code, or .json. For the .json response above here was my prompt:
"Read all of the transcripts from 2021 until April 2023 for NVDA. Tell me what you see and what topics are changing at the margin. Put your results in chronological order with the date in .json format."
Now imagine that prompt running at scale! Happy to chat more on this fascinating topic. FYI, you're correct, the machines will read the inference tokens that the AI produces from the everchanging corpus of information.
Erik, I've been re-reading your recent pieces. This sentence hit me like a thunderbolt: "Having a large labor force in 2030 is like having a factory in Ohio in the 70’s, or having a nationwide store footprint in 2010."
We will see many headlines or, at least, hear many rumors and recruiter "shop talk" about similar phenomena at other tech firms and eventually, everywhere.
Thank you. Any maybe the AI effect on the labor force won't be so much headlines like '500 laid off and replaced with AI', but more that the job openings never came up in the first place. There will just be fewer human roles and companies will be more selective about who they hire for them and it will just feel really difficult to get a job. Kind of like how it feels now, but worse.
I agree as long as earnings hold up. The labor market is frozen. But if we have another 2022-style swoon, I would not be surprised to see mass layoffs. And as AI edges out humans, it'll be premiumization for the well-off (human interaction as luxury good) and "South Africanization" for everyone else.
Excellent, just excellent. This is the piece that made me a subscriber. Fewer than 10 people in discretionary active management could have written this. I know I couldn't and I've been at this for 30 years and in the AI space for 3 years. I've been building AI analytics, and here is the launch of positive revisions for your NVDA example, right on cue. In of course, .json :)
date": "2023-01-05",
"transcriptLabel": "NVDA JPMorgan CES Tech/Auto Forum",
"keyTopics": [
"Introduction of new Ada architecture GPUs (4090, 4080, 4070 Ti) driving the gaming cycle",
"Integration of cloud gaming (GFN) with automotive applications",
"Continued advancements in data center platforms and automotive solutions leveraging Orin"
],
"metrics": {
"ProductAdoption": "High demand for 4090; robust DLSS 3 implementation"
},
"notableQuotes": [
"Our 4090 is tight in availability and 4080 pricing is marked up relative to MSRP, reflecting strong demand."
]
},
{
"date": "2023-02-22",
"transcriptLabel": "NVDA Q4 2023",
"keyTopics": [
"Stable data center growth driven by H100 uptake and generative AI emphasis",
"Mixed segment performance with recovering Gaming and record Automotive revenue",
"Launch of NVIDIA AI cloud services (DGX Cloud) and integrated AI platforms across public/private clouds"
],
"metrics": {
"QuarterRevenue": "$6.05 billion",
"FullYearRevenue": "$27 billion",
"DataCenter": "$3.62 billion",
"Gaming": "$1.83 billion",
"Automotive": "$294 million"
},
"notableQuotes": [
"H100 is as much as 9x faster than the A100 for training and delivers up to 30x faster inferencing for transformer-based models.",
"Generative AI applications will help almost every industry do more faster."
]
}
]
Drew! This is massive. It looks like you are already working on this. And I didn't go into it in the post, but does it change how we write research reports? Do reports get built (possibly in JSON), for other AI's to read? Do we need to optimise for AI instead of human use? You see it already with Google Search. I keep looking at the AI summary and do less clicking. Graham never looked at the source documents the AI was using.
Hi Erik, I am working on this, and here is a tidbit of what I see in the space. Incumbent firms know something is afoot with AI but have no idea what to do next. So, there are slide decks on AI on execs desks built by first-year MBAs who got assigned the job to "figure out how to do AI." I think the default position is to wait until one of the major terminals figure it out. Meanwhile, the frontline analysts are Grokking, and Perplexity-ing which is good and bad. I have a prosumer app called DoTadda Knowledge that uses a corpus of 350,000 conference call transcripts that gets updated in real time. It works just like a chatbot. It can output formatted text for a human, code, or .json. For the .json response above here was my prompt:
"Read all of the transcripts from 2021 until April 2023 for NVDA. Tell me what you see and what topics are changing at the margin. Put your results in chronological order with the date in .json format."
Now imagine that prompt running at scale! Happy to chat more on this fascinating topic. FYI, you're correct, the machines will read the inference tokens that the AI produces from the everchanging corpus of information.
The AI/JSON is China/Container analogy is genius!
Erik, I've been re-reading your recent pieces. This sentence hit me like a thunderbolt: "Having a large labor force in 2030 is like having a factory in Ohio in the 70’s, or having a nationwide store footprint in 2010."
To wit, https://www.businessinsider.com/shopify-ceo-tobi-lutke-employees-prove-ai-job-2025-4
We will see many headlines or, at least, hear many rumors and recruiter "shop talk" about similar phenomena at other tech firms and eventually, everywhere.
Thank you. Any maybe the AI effect on the labor force won't be so much headlines like '500 laid off and replaced with AI', but more that the job openings never came up in the first place. There will just be fewer human roles and companies will be more selective about who they hire for them and it will just feel really difficult to get a job. Kind of like how it feels now, but worse.
Here’s another big tech company with an AI-first policy (a/k/a/ shadow hiring freeze): https://www.linkedin.com/posts/duolingo_below-is-an-all-hands-email-from-our-activity-7322560534824865792-l9vh?utm_medium=ios_app&rcm=ACoAAAIJQcgBEWcfdcvjzhXmCsHSQFEyqEMI4MY&utm_source=social_share_send&utm_campaign=copy_link
The comment that is like going mobile first in 2012 is powerful.
I agree as long as earnings hold up. The labor market is frozen. But if we have another 2022-style swoon, I would not be surprised to see mass layoffs. And as AI edges out humans, it'll be premiumization for the well-off (human interaction as luxury good) and "South Africanization" for everyone else.
Oh my goodness!! Human interaction as a luxury good. I LOVE it! And South Africanisation for everyone else. Sheesh!
Erik, thank you, this is a great thought piece and analogy for rethinking processes in the world of AI. So good, I just upgraded to paid, thank you.
Thank you very much. I think there is still a lot we can figure out.