What the people building AI said would happen, and whether it did.

Inference revenue will end this year well over $200 billion.

“Yeah, exactly. I think we end this year well over 200 billion in inference revenue, well over.”

The big unsolved problems of physics, such as quantum gravity and unified field theory, will be solved by another method within half a year.

“I don't think they'll be solved this way, actually. I think they'll be solved another way, but I don't think it'll be more than half a year.”

OpenAI will become profitable on revenue minus compute costs within two or three quarters.

“And OpenAI gets there, you know, within the next couple of quarters, maybe two or three quarters.”

Within a year, AI agents in enterprises will expand beyond web-based use to more product surfaces, such as Teams, Asana and Confluence, where they can be triggered.

“ai agents are going to continue expanding to product surfaces to expect like any single surface you have today to be able to interact with your core ai agents whether if it's teams whether if it's asana whether if it's you know confluence whatever you're working with being able to trigger them from there”

Waymo will expand internationally and into other weather conditions and highways over the next 12 months.

“And I think that is what the next 12 months is going to be, more of it, international, other weather conditions, highways, and everywhere scaling up with this high safety bar we have”

Within a year, some medical AI tools will reach retrospective clinical trials, and possibly prospective trials.

“I mean, I think what we'll start to see is some of these tools could potentially reach these clinical trials, at the very least, definitely like retrospective trials, potentially even prospective trials.”

Somebody will claim to have demonstrated error correction or mitigation on a running quantum computer by the end of 2027.

“And so I'll predict you'll see somebody make claims by next year.”

One or two companies will make progress on quantum error correction, though not at scale, by 2027.

“you'll find one or two will make progress by 2027 error correction not at scale but error correction”

U.S. CapEx, including data centers, chips and the wider supply chain, will be on the order of $2 trillion next year.

“So, like, I mean, if you look at, like, next year, CapEx across the U.S. will be on the order of $2 trillion, not just data centers and chips, but also all the rest of the supply chain that people are investing in.”

OpenAI and Anthropic together could reach $1.5 trillion in revenue by the end of 2027.

“I'm really in that you can get to, you know, one and a half trillion, maybe even by 2020 end of 2027 with just those two companies.”

Frontier model training runs will cost on the order of ten billion dollars going into next year.

“Going into next year, I think order 10.”

Within 12 to 18 months, governments will regulate data center access and usage, allocating GPUs, inference and chips to sectors such as defense and financial services.

“What I think is going to end up happening, and this is a prediction no one may like, but I do think we're going to end up in a world where governments are going to regulate data center access and usage.”

Pick-and-place and packaging tasks on assembly lines that traditional robotics could not solve will be solved within the next two years.

“A lot of these pick and place, these packaging tasks that you see on assembling lines that had to be manipulation bots that couldn't be solved by traditional robotics, we will get there within the next year or two.”

Western open-weight models as capable as the Chinese ones will come out within the next year or two.

“But I think that over the next year or two, I'm very doubtful that there will not be great models that come up in the West that are also open weight and that are as capable as the Chinese.”

AI with a virtual cell will propose in silico small-molecule drug interventions that cure the top 5,000 diseases within two to three years at most.

“there are in silico proposed say small molecule drug interventions that cure the top 5,000 diseases thanks to a virtual cell I think we're going to see that in the next two to three years maximum.”

AI will solve the search problem of planning intervention paths from a diseased state to a healthy state within the next two to three years.

“How do I, you know, turn left and then turn right, take this pill, then take this pill, then climb this, you know, hill to get from A to B and the AI will solve. That's just a search problem. So AI will solve the search problem in, I think, next two to three years.”

There will be a massive compute shortage in the next three or four years.

“we are probably going to see a massive, massive, massive compute shortage in the next three or four years.”

By the end of the decade, annual AI infrastructure build-out will probably be between $3 trillion and $4 trillion, closer to $4 trillion.

“I said a couple of years ago that as we exit this decade, infrastructure build-out will probably be somewhere between $3 to $4 trillion a year.”

Within five years there will be 100 times more people writing 100 times more software, about 10,000 times more software overall.

“Well, I think you're going to have, you know, a hundred times more people writing a hundred times more software.”

First drafts of cellular atlases of health and disease will exist within two years, and a pretty comprehensive map within five years given a focused coordinated effort.

“first drafts of that within the next one or two years, you know, within five years with a really focused coordinated effort, we can have, you know, a pretty comprehensive map there.”

Enterprises will consolidate from hundreds of software vendors to five to ten main platforms within six years.

“I believe that's going to play out whether it takes six months or six years.”

An AI takeover, via controlling government or physically overpowering militaries, will not happen within the next five to ten years.

“Both those pathways, I think, are quite implausible to me in the next, you know, like five year, 10 year horizon or something.”

Within ten years, AI will leave no use for people in a lot of the things people do now, eliminating many jobs.

“I think there's going to be no use for people for a lot of things that people do right now.”

Mistral's next model, ML5, will be able to do meaningful work for a few days, up from a few hours now.

“Tomorrow with ML5 is going to be a few days, for sure, of meaningful work.”

Bots will make up 99.999% of online traffic very soon, rather than humans.

“I expect that to be 99.999% very soon will actually be bots, not humans.”

Very soon, OpenAI's backlog of AI solutions to significant open problems will be published, hundreds of them, mostly unchecked by any human.

“but but my my expectation is that at some point you know very soon in fact you know all these things are going to be published all right like you know we're going to start seeing hundreds of solutions to significant open problems that in most cases no human has gone through”

Anthropic will likely release a new model in the coming weeks, ahead of its IPO.

“So I think, you know, Anthropic will likely release a new model in the coming weeks, you know, for their IPO.”

OpenAI will likely release a new model in the coming weeks.

“I think OpenAI will likely have to release a new model in the coming weeks as well, just because it's such a competitive environment.”

Agents will potentially drive a 10 to 100x increase in the volume of queries hitting databases.

“So I think there's going to be potentially a 10 to 100x increase in volume and the number of queries that are going to hit up against databases.”