AI: Year in Review for 2024. RTZ #586
...mostly continuing the key AI trends with acceleration into 2025
It’s the last day of 2024, so here’s the predictable look back at my 2024 predictions this time last year. And tomorrow, on New Year’s Day, some predictions for 2025 of course in these early days of the AI Tech Wave. Two years after OpenAI’s ‘ChatGPT moment’ on November 30, 2022, and several hundred billion in AI capex in by companies large and not so small. With trillions more to come potentially to Scale AI in predictable and unpredictable ways going forward.
In many ways, writing an AI roadmap to 2025 is more like Congress writing their seemingly never-ending ‘Continuing Resolutions’ at the last minute to extend the spending today to many tomorrows. It’s mostly the same ingredients kicked down the road.
At this early stage of the AI Tech wave, and at least for at least the next two to three years, Nvidia, and their fab partner TSMC will continue to enjoy the majority of the AI GPU chip market. Despite their best customers rushing to build competitive offerings with the help of other big semiconductor companies.
Every one of the the six boxes of the AI tech stack above will see Continuing Resolutions of their own extending their growth solidly in 2025 at least. With the new entrant this tech wave, the Power industry in Box 1 represented by Utilities large and small. I’ve been writing a fair bit about those in recent days.
So Boxes 1 through 3 will continue to see ‘Risk-on’ support from investors in 2025, unless of course unknown macro-economic, geopolitical, and US political factors potentially change Wall Street’s mind. With US/China ‘threading the needle’ dynamics of course playing a key, potentially throttling role in tech/AI and broader economic growth for the US.
The above is the continuing arc for broad 2024 AI trends continuing into 2025. Let’s go through some specifics from my outline this time last year for 2024. The six key points from last year held pretty well, and are still set to continue into 2025:
“AI still being invented and re-invented”.
This one has delivered this year, with AI tech rapidly evolving to AI Reasoning and Agentic capabilities, on their way to AGI however defined, and however many years away. I’ve written a lot about this year to date, and the pace continues to be ahead of Moore’s Law both in software and hardware terms. These are now recognized as ‘AI Table Stakes’, and the industry is ‘pedal to the metal’ going into 2025.
“Early days in Horse race”.
Can’t stress this one more. Unlike previous tech waves, this one is the least ‘baked’ in terms of the underlying technologies to create the AI devices, applications and services the world has imagined for decades via its sci-fi books and movies. From super capable personal agents/companions for billions, to humanoid robots catering to so many tasks capably, to of course those self-driving cars. The underlying tech even this year saw remarkable twists and turns that point to good things in the future. We just don’t know exactly when to ‘ready, aim, fire’. Hopefully in the right order.
“Chip Constraints Continue”.
It’s a bit better end of this year than beginning. We have a roadmap to go from tens of thousands to hundreds of thousands of new Nvidia AI GPU chips at $30,000 per, and likely seeing millions deployed by the end of next year. That is a good thing to prove out the AI Scaling thesis, or not. Can’t wait to be writing this piece same time next year with what we learned in 2025. The new constraints of course continue in terms of Power and Land for the AI Data centers, not to mention the AI Talent to build it and run it all. And of course the many, many billions to stake it all.
“Beware AI Veneers”.
This is a perennial one, where incumbents across every vertical and horizontal industry will add ‘AI’ stickers on their wares and try to charge extra. Other words like ‘Agents’, ‘Copilots’, ‘Full Self Driving’ and many others will be used. It’ll be a ‘buyer’s beware’ world, just like it’s been for generations. But there will be babies in the bathwater, and they’ll need to be plucked out. And it almost always takes longer than we think.
“Rising Regulatory Drumbeats”.
This one turned out to be somewhat less onerous in the US, and likely even less onerous with the coming new administration and Congress in 2025. There is more silicon valley/tech support for AI and tech going into 2025, especially framed as a geopolitical contest with China. But the palpable optimism there needs to be balanced so as not to have a premature ‘Mission Accomplished’ moment. Other locales like Europe, which is putting regulatory caution ahead of AI innovation, is succumbing potentially to premature ‘AI Fears’. And China of course is full throttle leaning into AI infrastructure, EXCEPT allowing unencumbered data collection and use in next gen AI models WITHIN China. So for now, the US has the most regulatory tailwinds going into 2025.
“Geopolitics trump globalized tech supply chains”.
This one is the most worrisome for 2025, when one looks at the AI/Tech outlook from a government perspective. Despite what I outlined in point 5 above. Everything from tariff negotiations, to immigration/mass deportations, to ‘friend-shoring’/re-shoring is done of global supply chains, could have meaningful, sudden impact on the outlook. So one to be watched most closely of all the items on the list. It’s the one with the biggest cluster of ‘unknowns’ going into the New Year.
There are a lot more factors to consider for the AI Tech Wave in 2025 of course. But the ones above are the big, overarching ones. Will continue to track these and many others in 2025 of course, just like have been doing daily for the last 586 plus days.
In the meantime, I want to thank you all for being readers and supporters here thus far this year at AI: Reset to Zero.
And wish you and yours the Happiest of New Year’s. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)