What this covers

This is Nate B Jones working through Mary Meeker's 340-slide AI trends deck, with his own commentary layered on top of her charts. The first part of the walkthrough stays on adoption, capital, and compute. ChatGPT reached 100 million users in two months, where the internet took years to get there. Its 800 million users are spread globally, with North America a smaller share than he expected, and mobile-first access in South Asia, Africa, and Latin America helps explain that. He also points to roughly 20 minutes of daily use per person, and to computing patent filings climbing in a way that resembles the 1995–2003 internet years.

The middle of the talk turns to economics. Inference costs have fallen sharply while training costs keep rising, and he reads Capex (Capital expenditure)) as something only cash-rich incumbents can sustain. He argues that Nvidia's developer base was growing well before ChatGPT launched. Meeker's framing of AI as a meta-technology that helps incumbents like Google and Microsoft pursue their existing missions is flagged as contested, as is the claim that OpenAI's revenue is outrunning serving costs but not its losses. Model proliferation and a growing set of cloud providers point him toward a Multimodel World.

The remaining ground is more skeptical or more specific. Data-center efficiency is a recurring example: tokens per megawatt-year rose from 1.3 billion to 65 trillion, a figure he says energy headlines rarely mention, even as aggregate demand raises questions about new nuclear capacity. On medicine, he describes AI shortening R&D timelines by 30 to 80 percent, with drug-target discovery as an underrated path to profit. He questions whether training data will run out and whether future gains will come from algorithms instead of raw scale. He also sets today's text generation against Meeker's 2030 projections of full-length films and human-level understanding, noting the unsolved problems in memory, adaptive learning, and intent. Measuring productivity across industries, he adds, will be as hard for AI as it was for the internet.

Sharpest takeaway

Mary Meeker's AI trends deck illustrates that AI adoption is accelerating faster than internet adoption across nearly every metric, driven by proliferating models, plummeting inference costs, and aggressive capex investment by incumbents, but critical gaps remain between current capabilities and claimed future possibilities.

  • ChatGPT reached 800M users and 100M in 2 months, vastly faster than internet or mobile adoption timelines
  • Inference costs have dropped 99.7% in 2 years while training compute costs continue climbing 2400x over 8 years, creating a revenue-capex mismatch
  • A proliferated multi-AI world is emerging due to low-cost token generation and distributed data centers, not a winner-take-all scenario

The claims · ranked97 claims · weighted by value

This asset isn't compiled yet

You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.

0.79

ChatGPT's text-heavy, image-capable form factor accessible via cell phone has enabled adoption across India, sub-Saharan Africa, South Asia, Middle East and North Africa, Latin America and Caribbean, where mobile-first internet is the primary connectivity mode.

factualhigh valueestablishednovelty 2/4durability 3/4· Nate (Host/Analyst)

“much of the uh like India, subsaharan Africa, South Asia, um Middle East and North Africa, Latin America and the Caribbean. These are places where if you have a cell phone, you're online and you can get chat GPT. And so it's actually relatively easy to stack up adoption into a very very large overall approach.”

0.75

Data center power use has decreased 43% over 8 years while per-unit energy efficiency has improved 50,000x, and tokens per megawatt per year have scaled from 1.3 billion to 65 trillion, demonstrating dramatic improvements in compute efficiency that are underreported in energy consumption headlines.

factualhigh valueestablishednovelty 2/4durability 3/4· Nate (Host/Analyst)

“Material improvements in GPU performance. Basically everything's gotten vastly better. Like data center power use is down 43% over 8 years leading to a 50,000x greater per unit energy efficiency. You do not hear that a ton in the headlines about power usage. And I'm not saying power usage isn't relevant. But the fact that we've gotten 50,000 times more efficient makes it's a big deal. Uh we've scaled from 1.3 billions tokens per megawatt year to 65 trillion per megawatt year.”

0.75

Data center construction is scaling rapidly, with individual data centers consuming electricity equivalent to the needs of ~418 US homes, and aggregate AI data center energy consumption will drive infrastructure investment in new nuclear power in the US and other countries.

factualhigh valueestablishednovelty 2/4durability 3/4· Nate Mastersonn (channel host)

“Data center growth, existing capacity, new capacity, just going super fast. Um, how big is a data center? I love this little illustration. It's very, very large. It fits 418 US homes...And so she talks about overall energy consumption. Even though we're getting vastly more efficient, which is great, still consuming a lot of energy, we're still going to need to address that. Um, one of the things that I think is interesting is that this has catalyzed a downstream revolution in nuclear power in the US where we're now greenlighting a lot of nuclear power.”

0.75

Medical R&D timelines have experienced 30 to 80% reductions due to AI, and the speaker expects to see more gains with a push toward the 80% reduction figure.

factualhigh valueestablishednovelty 2/4durability 3/4· Nate Mastello (inferred)

“Research uh 30 to 80% reduction in medical R&D timelines which is wild. This one doesn't get publicized enough. We're going to see a lot more like this. We'll probably see more gains here. More of the push to 80%.”

0.75

Unlike the internet, which generated significant GDP growth but made it difficult to definitively measure productivity improvements across industries, AI faces a similar challenge of demonstrating measurable utility across enough different industries to clearly see its impact, despite potentially being tremendously impactful.

causalhigh valueestablishednovelty 2/4durability 3/4· Nate (Host/Analyst)

“It's hard to prove positively how the internet affected GDP growth and productivity growth in particular. Uh you can see internet companies generating revenue but proving that productivity increases drove that GDP growth. That's a little bit sketchier. It's one of the notorious loopholes um in technologies. It's very difficult for a generally available technology to demonstrate its utility enough across enough different industries in a very measurable way for us to see the impact very clearly.”

0.75

Emerging AI applications in cancer detection, robotics, and drug discovery through protein folding and novel drug target identification via chemical structure search and academic literature analysis are scaling rapidly, with drug targets being a particularly underappreciated path to profitability.

factualhigh valueestablishednovelty 2/4durability 3/4· Nate (Host/Analyst)

“Emerging applications are accelerating. Cancer detection, robotics, she doesn't talk about drug pipelines, but that's another big one. Um protein folding kind of gets at that a little bit. But really, drug targets are a new one where you search through past academic papers and search through the chemical structure of the of the drug to sort of identify novel targets and novel use cases for existing drugs, which can be a simpler path to profitability off of existing drug testing.”

0.75

Data center power consumption has become more efficient by 50,000x in terms of tokens generated per megawatt-year, scaling from 1.3 billion tokens/megawatt-year to 65 trillion tokens/megawatt-year, which is rarely reported in discussions of AI energy consumption.

factualhigh valueestablishednovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“data center power use is down 43% over 8 years leading to a 50,000x greater per unit energy efficiency. You do not hear that a ton in the headlines about power usage. And I'm not saying power usage isn't relevant. But the fact that we've gotten 50,000 times more efficient makes it's a big deal. Uh we've scaled from 1.3 billions tokens per megawatt year to 65 trillion per megawatt year. That's insane.”

0.75

ChatGPT has achieved faster user adoption than any previous technology, reaching 100 million users in 2 months compared to the internet taking 11 years to reach equivalent penetration, indicating AI is on track for horizontal technology status similar to the internet.

causalhigh valueestablishednovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“User growth up into the right. Subscribers, revenue up into the right. This is the bullcase for Chat GPT right here. like they they are hitting absolute vertical on 800 million users. I would expect them to hit a billion by the end of the year. They are thanks to the power of an internet that's already available, they are very very rapidly hitting uh the same search mark that Google took 11 years to hit”

0.75

Demonstrating productivity and GDP impact from AI will be difficult because it is a generally available technology that operates across many industries, making it hard to isolate and measure causation, similar to the difficulty in proving internet productivity gains despite obvious GDP correlation.

causalhigh valueestablishednovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“It's very difficult for a generally available technology to demonstrate its utility enough across enough different industries in a very measurable way for us to see the impact very clearly. So basically it may be tremendously impactful and it may be hard to see like the internet. And that's essentially what she's suggesting is that the AI era is like internet on steroids.”

0.73

ChatGPT's revenue is scaling but not as fast as their compute losses, creating a fundamental unit economics disconnect that the company must resolve through either increasing revenue per user, reducing compute costs per query, or improving model efficiency.

causalhigh valueestablishednovelty 1/4durability 3/4· Nate Mastersonn (channel host)

“One of the things that Mary is essentially calling out is that Chad GPG's revenue is scaling but not as fast as their losses are scaling from serving all the compute. There is a disconnect and they're going to have to figure out how to close that”

0.70

Medical R&D timelines are being reduced by 30-80% through AI application to drug targeting and novel drug use case discovery, where AI searches through past academic papers and chemical structures to identify novel targets for existing drugs, providing a simpler profitability path than developing entirely new drugs.

factualhigh valueestablishednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Research uh 30 to 80% reduction in medical R&D timelines which is wild. This one doesn't get publicized enough. We're going to see a lot more like this. We'll probably see more gains here. More of the push to 80%. Okay. Uh rising rapidly across age groups. What she doesn't get at, by the way, yes, usage is rising, but the difference in estimates versus what Pew surveys say is notable. And Pew is also the company that has been or the survey company that has been reporting the most pessimism about AI usage among American adults. And so I'm a little bit shocked that we see this big a disparity. I guess people lie on surveys all the time. Um, and I think she doesn't do us a service by not calling out how pessimistic Pew survey respondents are about AI. But this is definitely a factor. Again, quantitatively, people are using AI a ton, whatever they say.”

0.70

ChatGPT's user base is globally distributed, with North America representing a surprisingly small proportion of the 800 million total users, with more users in Europe and the developing world than in the US, making it a truly global product rather than US-centric.

factualhigh valueestablishednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“I will say I did not realize that chat GPT app users in North America, that blue section there, I didn't realize that was such a small proportion of overall users. If you had tapped me on the shoulder and you would said, Nate, out of 800 million chat GPT users, how many are in the US? You know what I would have guessed? 150 million. Maybe half the country. Nope. Apparently, it's a lot smaller than that. And they're stacking stuff up in Europe. There's it's truly a global product.”

0.69

Computing-related patent applications spiked sharply in the 2020-2025 period (comparable to the 1995-2003 internet expansion), indicating that we are at the beginning of a new technological revolution where innovation is rapidly expanding in breadth and creating new opportunities for disruption.

factualhigh valueestablishednovelty 1/4durability 3/4· Nate Mastersonn (channel host)

“This is one of the eye openeners for me...at the end of the day, this is indicating we're at the beginning of a new revolution. Like, people are doing new things and the innovation is really popping. And I had no idea that the pop for computing related patents was this high. This is super exciting.”

0.69

Extracting 20 minutes of daily usage from individual users across millions of people is a significant achievement, and this engagement is inflecting upward as model capabilities improve and new modalities (images, etc.) are added.

factualhigh valueestablishednovelty 1/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“And I think that's definitely a factor. Again, quantitatively, people are using AI a ton, whatever they say. Like, you can actually see this. Um, and it is really hard to get minutes out of someone's day. So, being able to carve out 20 minutes from someone's day on average across millions of users is huge. And it's inflecting up. This is inflecting up with better models. It's inflecting up with images, etc.”

0.69

Tech companies have loaded cash reserves and are recycling free cash flow into AI compute spend for training and serving models, particularly through data center capex, indicating that cash-rich incumbents can sustain AI investment indefinitely regardless of unit economics.

factualhigh valueestablishednovelty 1/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“and of course they're loaded with cash. They to no one's surprise Apple, Nvidia, Microsoft, Google, Amazon and Meta make a ton of money. Um and they make uh lots of money that they then recycle into investments. So free cash flow just continues to grow. Uh, and as they continue to grow, they're using it for AI compute to spend to train and run AI models. So, what are they spending it on? This is one of her larger points, right? They're spending it on data centers, right?”

0.68

AI is a meta-technology that enables major companies like Google, Facebook, and Microsoft to optimize their operations and better fulfill their core missions rather than fundamentally disrupting them, because these incumbents have been aggressive about keeping up with AI trends and maintaining their dominance.

causalhigh valuecontestednovelty 2/4durability 3/4· Nate (Host/Analyst)

“This is a meta technology that enables us to do a lot of other things more effectively including use the internet, including uh make it easy to do business everywhere, including organize the world's information. Like you can see the efforts of these major companies as essentially efforts to optimize against a meta technology that makes it easier to fulfill their mission.”

0.68

North American ChatGPT users represent a surprisingly small proportion of total users (the speaker expected ~150 million or roughly half the US population), indicating that AI adoption is genuinely global rather than US-centric, with significant user bases in Europe and developing regions.

factualhigh valueestablishednovelty 2/4durability 3/4· Nate Mastersonn (channel host)

“I didn't realize that chat GPT app users in North America, that blue section there, I didn't realize that was such a small proportion of overall users. If you had tapped me on the shoulder and you would said, Nate, out of 800 million chat GPT users, how many are in the US? You know what I would have guessed? 150 million. Maybe half the country. Nope. Apparently, it's a lot smaller than that. And they're stacking stuff up in Europe. There's it's truly a global product.”

0.68

Developer growth in the Nvidia ecosystem shows a 6x increase, with timing that predates ChatGPT's release and coincides with crypto/blockchain discussion period and early machine learning hype, suggesting the developer boom was already underway before generative AI became public, setting up the infrastructure for the subsequent AI explosion.

causalhigh valuecontestednovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Developer growth in the Nvidia ecosystem 6x like just an absolutely stunning gauge. And you know what's interesting? It gapped up around the time when we started to talk about crypto around the time when we started to talk about late Zerp era some AI but not Chad GPT release. And so it's a little disingenuous to see AI as driving this trend. when Chad GPT came out here. It's more accurate to say that developer ecosystem has been booming for Nvidia for a long time and arguably set up AI to be a super boom, which I think is an interesting insight that doesn't get reported a ton.”

0.68

ChatGPT's revenue is scaling faster than it should given the costs of serving the compute, but not fast enough to close the gap between revenue and losses, creating a structural revenue-capex mismatch that the company must resolve.

causalhigh valuecontestednovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Chad GPG's revenue is scaling but not as fast as their losses are scaling from serving all the compute. There is a disconnect and they're going to have to figure out how to close that”

0.64

Training data set size has been scaling rapidly since 2010 and reached 1e13 (10 trillion) words, but there is a substantial unresolved question about whether data will run out and whether we can continue scaling training data sets at the current log-scale pace without finding alternative scaling laws.

factualhigh valuecontestednovelty 2/4durability 3/4· Nate (Host/Analyst)

“Training data set size. This is super controversial. Yes, it's been scaling really fast since 2010. I think the question is can we keep scaling from here? Is there a limit? We are one e to the 13 on uh number of words in the data set. Does it make sense to keep going up? Is there more to it?”

0.64

200% annual growth over nine years of compute gains from better algorithms, combined with raw compute scaling, demonstrates that future AI improvements will likely come increasingly from algorithmic progress and better use of existing compute rather than just scaling raw compute or pre-training data over the next 5-10 years.

causalhigh valuecontestednovelty 2/4durability 3/4· Nate (Host/Analyst)

“200% annual growth over nine years of compute gains from better algorithms has led to uh sort of tremendous gains in AI intelligence. Uh it's also led to gains in effective compute. Basically, you can scale compute, you can scale algorithmic progress. The thing I take away here is that this is one of the clearest graphs I've ever seen of the difference between scaling algorithms and scaling just raw compute.”

0.64

There is a significant gap between what ChatGPT can do today (generate text) and what Meeker claims it will do by 2030 (create full-length films, achieve human-level understanding and speech), with multiple unsolved breakthroughs needed in adaptive learning, context handling, memory, and intent modeling that native human cognition handles intuitively.

causalhigh valuecontestednovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“There's a massive gulf between these two slides. It's a gulf not just in terms of assembling new tokens, but in terms of several breakthroughs we haven't seen yet that I don't think Mary does a good enough job teasing out. And maybe that is just not in the scope for this deck because it's a phenomenal deck overall. One of them is how these systems adaptively learn in the wild, how these systems handle context, how these systems handle memory, how these systems handle intent over time. I could go on and on. There's a lot of things that are native to humans, especially adult humans who have been educated for information work or even creative work.”

0.64

The number of new large-scale AI models being released is explosively increasing, illustrating rapid model proliferation that cannot fit on a single slide, demonstrating that the AI landscape is moving toward a multimodel world rather than consolidation.

factualhigh valueestablishednovelty 1/4durability 2/4· Nate (Host/Analyst)

“Number of new large scale models. This is absolutely explosive. This is illustrating proliferation to me. Like she can't even fit this on the slide. This is one of the money slides.”

0.64

Realistic audio synthesis from 11 Labs is incredible and already being used in production settings at scale, as evidenced by Spotify accepting AI-translated audiobooks in 29 languages powered by AI voice technology.

factualhigh valueestablishednovelty 1/4durability 2/4· Nate (Host/Analyst)

“Realistic audio. If you've ever tried one of the audios from 11 Labs, it's incredible. Um, and they absolutely are being used in production settings. And you can see that in sort of the scale of uh sort of how 11 Labs is getting used. Um and Spotify accepting audiobooks AI translated into 29 languages.”

0.64

There are government sovereign AI partners in various countries (specifically mentioned: Middle East partnerships with NVIDIA and ChatGPT), which is one reason we will see a proliferated AI future with many different data centers in different locations.

factualhigh valueestablishednovelty 1/4durability 2/4· Nate Mastello (inferred)

“There's definitely work between NVIDIA and Chad JPGT in the Middle East as of the last uh couple weeks. But this is one of the reasons by the way why we are going to have a proliferated AI future. Look at how many different data centers we have already. Um, we're going to have more chat GPT will be all over the place and it is to be expected that we will have multiple AIs that are very highly capable.”

0.64

Cloud revenue has grown linearly as companies continue to invest heavily in cloud infrastructure, with new entrants like IBM Cloud, Alibaba Cloud, and Oracle Cloud becoming more relevant alongside AWS and Microsoft, creating a multi-cloud competitive landscape.

factualhigh valueestablishednovelty 1/4durability 2/4· Nate (Host/Analyst)

“Uh but you have some new players coming in too. Uh IBM cloud is nothing to sneeze at now. It was quite small in 2016, but it's starting to come in. Alibaba cloud was becoming well it was becoming relevant here but it's had trouble scaling up past that and we have newer players like Oracle cloud which is becoming more serious since 2020 and so it's really becoming a multi cloud world”

0.64

FDA approval timelines for AI medical devices are accelerating, with the approval pipeline moving faster and getting faster, and further acceleration is expected in 2025 as regulatory frameworks mature.

forecasthigh valueestablishednovelty 1/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“FDA approved AI medical devices are scaling very rapidly. I would expect this to scale even faster in 2025. Um, and we will have to see how that sort of nets out. But I think that the approval pipeline has gone very very fast and it's getting faster at least by FDA standards.”

0.63

A geopolitical AI leadership race and winner-take-all AI scenario is unlikely because the proliferation of AI technology is too easy and fast, meaning we will see multiple capable AIs globally rather than one country or company achieving super-intelligence dominance.

forecasthigh valuecontestednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“I think a geopolitical leadership race for AI or an AI leadership race that beggets geopolitical risk only transpires if we buy the idea that just scaling more intelligence, which is what we see evidence of today, is enough to get us to a sort of super intelligent scenario where one company in one country has found super intelligence and quickly evolves to the point where nobody else can catch up. Even Sam Alman no longer thinks that's the most likely outcome. We live in a proliferating world. We're going to get all the AIs. And so I in a sense I think it's not so much global AI leadership as it is the global crowd of AIs that we're getting.”

0.63

Performance convergence is occurring across different AI models as server costs decline and data center techniques proliferate, indicating that cost-based differentiation rather than capability differentiation will define future market competition.

forecasthigh valuecontestednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Performance is converging across models as all of these servers are cheaper, as all of the the data centers proliferate, as the techniques to make these models proliferate, everything is converged. This is what I mean about a multimodel world.”

0.63

Training compute costs are scaling 2400x over eight years and have already exceeded $100 billion to train a single model, raising questions about sustainable scaling and whether companies will continue investing at 10x cost multiples.

factualhigh valuecontestednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“AI model compute costs, uh, AI model training. We are scaling up 2400x growth over eight years, which is just insane. Um, so it just gets more and more expensive, like we're past hundred billion dollars to train a model. Are we going to be willing to keep spending more and more and more at factors of 10? That's really the question.”

0.63

IT jobs category decline of 9% is misleading framing because it represents a relative shift rather than absolute job loss, with the baseline essentially flat versus 2018 after accounting for the 2019 XP-era tech boom, now replaced by AI job growth.

causalhigh valuecontestednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“I do not think IT jobs is the right frame for tech. And that's a categorization issue from the Department of Labor that I just don't think is super helpful. If you look at the trends more broadly, and this kind of shows it, down 9% is a little bit disingenuous. And the reason why is because it's basically flat versus 2018. We had the massive Xerpier run up in all tech jobs and now you have a runup in AI.”

0.63

Some capabilities listed as 2035 targets by Meeker (alpha folding conducting scientific research and generating hypotheses, running simulations) already have working examples today, suggesting that the deck's timeline is somewhat mixed up or that some of the claimed future capabilities are nearer-term than suggested.

causalhigh valueestablishednovelty 2/4durability 2/4· Nate (Host/Analyst)

“Ironically, some of the stuff that's listed here by 2035, we already have examples of doing. So, alpha evolve conducts scientific research. It generates hypotheses. Um, it's on the verge. Uh, I think some of them run simulations. So, this is not really new.”

0.63

Microsoft's investment in OpenAI was a significant moment in the AI revolution, helping establish Microsoft as a major player and making it difficult for competitors to catch up

factualhigh valueestablishednovelty 0/4durability 3/4· Unknown Speaker (Nate - Analyst)

“Microsoft's investment in open AI was a significant moment in this overall trend in this overall AI revolution.”

0.63

AI inference costs have dropped 99.7% relative to the lightbulb's 75-year decline to similar levels, compressing that efficiency improvement into just 2 years, demonstrating extraordinary acceleration in cost reduction

factualhigh valueestablishednovelty 2/4durability 2/4· Unknown Speaker (Nate - Analyst)

“AI inference costs are 99.7% lower. Cost efficiencies. This is the light bulb which took 75 years. This is chat GPT that took two years uh to go down 99%. It's insane.”

0.62

It is uncertain whether ChatGPT will actually displace Google Search as the primary search interface because the monetization model (high-intent search advertising) and user intent patterns are unclear, despite ChatGPT's rapid user acquisition

forecasthigh valuecontestednovelty 1/4durability 3/4· Unknown Speaker (Nate - Analyst)

“Does that mean they're going to actually eat Google? Sam Alman said he didn't think so. He thinks that it's going to be different. He doesn't see a case where Google really disappears. Uh, I think it's going to be really interesting to see are these high intent searches. How does Fiji build ads at OpenAI? Those are really open questions.”

0.61

Inference costs are falling through the floor with declining cost and improving performance converging, representing what the speaker calls 'getting a Porsche for like a 100x cheaper every year', and a token is effectively four characters in English.

factualhigh valueestablishednovelty 1/4durability 3/4· Nate Mastello (inferred)

“Inference costs are falling through the floor. This is down into the right. Um, I love this. If you don't know what a token is, I think this is a great definition. Four characters in English. But fundamentally the energy required to generate a token is just falling through the floor.”

0.61

Ilya Sutskever gave a talk at NeurIPS claiming data is the new oil and we are running out of pre-training data, after which he founded Safe Super Intelligence and has done nothing public since then.

factualhigh valueestablishednovelty 1/4durability 3/4· Nate Mastello (inferred)

“Uliagiver gave a talk at Nurips last year basically saying data is the new oil. we don't have an infinite sort of amount of data left. We're going to run out of data for pre-training. We'll need to find other scaling laws. Uh and then he went off and founded I think safe super intelligence and has done nothing public since then.”

0.61

AI models have achieved and exceeded human baseline performance on benchmark tests, though this includes model overfitting and saturation effects that Meeker's deck does not address, limiting the meaningfulness of the superiority claim.

factualhigh valueestablishednovelty 1/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“System performance on bench test. Honestly, like she doesn't get into this, but this is like it's not really to me about surpassing the human baseline. I mean, it is. It's also about the models overfitting and saturating, and she doesn't get into that.”

0.60

AI systems would need to demonstrate adaptive improvement after being released into the wild (like how filmmaker Christopher Nolan or Steven Spielberg improve with each film), but it is not clear how an AI in the wild would adaptively get better over time or what alignment implications such adaptive learning would have.

factualhigh valuespeaker onlynovelty 3/4durability 4/4· Nate (Host/Analyst)

“Christopher Nolan gets better every time he makes a full-length film. Steven Spielberg arguably gets better and has been getting better for decades every time he makes a full-length film. But it is not clear how an AI in the wild would adaptively get better after being released and if it did what that would mean uh in terms of how quickly it would improve. Those are all very unanswered questions and they do raise alignment questions like would the AI stay aligned if that were the case.”

0.59

ChatGPT has reached 800 million users and is on track to hit 1 billion by the end of the year, achieving in two months what took Google 11 years to accomplish for search.

factualhigh valueestablishednovelty 1/4durability 1/4· Nate (Host/Analyst)

“they are hitting absolute vertical on 800 million users. I would expect them to hit a billion by the end of the year. They are thanks to the power of an internet that's already available, they are very very rapidly hitting uh the same search mark that Google took 11 years to hit”

0.57

Bank of America's Erica virtual assistant and JP Morgan's end-to-end AI modernization initiatives demonstrate large-scale enterprise AI deployment and capability, with case studies showing rapid scaling at major financial institutions.

factualhigh valueestablishednovelty 0/4durability 2/4· Nate (Host/Analyst)

“Bank of America, Erica virtual assistant, JP Morgan, endtoend AI modernization. Uh you can just see them sort of scaling that up.”

0.57

AI-generated audio from services like 11 Labs has reached production-quality realism and is being used in production settings at scale, as evidenced by Spotify's adoption of AI-translated audiobooks in 29 languages.

factualhigh valueestablishednovelty 0/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Realistic audio. If you've ever tried one of the audios from 11 Labs, it's incredible. Um, and they absolutely are being used in production settings. And you can see that in sort of the scale of uh sort of how 11 Labs is getting used. Um and Spotify accepting audiobooks AI translated into 29 languages.”

0.57

ChatGPT user retention is improving as the service offers more utility, and ChatGPT retention metrics are approaching Google Search retention levels, with continued growth expected in the next 1-2 years if scaling continues.

forecasthigh valuecontestednovelty 1/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Retention, uh, it's definitely improving. I think that they're offering more utility, so it's getting better. Uh, Google search retention is the gold standard here, right? Like, and that's stable as a table. But the fact that they're scaling up and you can expect them to start to catch Google in the next year or two if they keep scaling is interesting.”

0.57

Major tech incumbents like Google, Facebook, and Microsoft are so aggressively pursuing AI that they have eliminated most opportunities for disruptive startups to unseats them, because the incumbents are already integrating AI into their core business models.

causalhigh valuecontestednovelty 1/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“I think there are arguably less opportunities for disruption because these companies like Google, like Facebook, like Microsoft have been so aggressive about keeping up with the trends. I think Mary calls this out later, but Microsoft's investment in open AI was a significant moment in this overall trend in this overall AI revolution.”

0.57

There is a massive gulf between ChatGPT's current capabilities (assembling new tokens) and the claimed 2030/2035 capabilities (creating full-length films, adaptive learning) that Meeker does not adequately tease out, because several unresolved breakthroughs are required: adaptive learning in the wild, context handling, memory, and intent tracking over time—capabilities native to educated humans but not intuitive to AI systems.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Nate (Host/Analyst)

“There's a massive gulf between these two slides. It's a gulf not just in terms of assembling new tokens, but in terms of several breakthroughs we haven't seen yet that I don't think Mary does a good enough job teasing out. And maybe that is just not in the scope for this deck because it's a phenomenal deck overall. One of them is how these systems adaptively learn in the wild, how these systems handle context, how these systems handle memory, how these systems handle intent over time.”

0.57

Developer growth in the Nvidia ecosystem increased 6x, driven by long-term trends in machine learning and GPU accessibility rather than primarily by ChatGPT, as the growth curve shows acceleration around the time of cryptocurrency and late Zerp-era AI development but before ChatGPT's release.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Nate (Host/Analyst)

“Developer growth in the Nvidia ecosystem 6x like just an absolutely stunning gauge. And you know what's interesting? It gapped up around the time when we started to talk about crypto around the time when we started to talk about late Zerp era some AI but not Chad GPT release. And so it's a little disingenuous to see AI as driving this trend. when Chad GPT came out here. It's more accurate to say that developer ecosystem has been booming for Nvidia for a long time and arguably set up AI to be a super boom, which I think is an interesting insight that doesn't get reported a ton.”

0.56

The geopolitical risk framing in Meeker's deck is overstated because achieving superintelligence through scaling alone is unlikely, and even if it were possible, the technology is proliferating globally so rapidly that no single country or company can establish uncontested dominance before others catch up.

causalhigh valuecontestednovelty 2/4durability 2/4· Nate (Host/Analyst)

“I think a geopolitical leadership race for AI or an AI leadership race that beggets geopolitical risk only transpires if we buy the idea that just scaling more intelligence, which is what we see evidence of today, is enough to get us to a sort of super intelligent scenario where one company in one country has found super intelligence and quickly evolves to the point where nobody else can catch up. Even Sam Alman no longer thinks that's the most likely outcome. We live in a proliferating world. We're going to get all the AIs.”

0.56

Training dataset size may be hitting scaling limits around 10^13 words, with Ilya Sutskever (who founded Safe Super Intelligence) claiming data is becoming the limiting factor, suggesting future AI scaling will need to rely on algorithmic improvements and better chip utilization rather than raw data growth.

factualhigh valuecontestednovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“We are one e to the 13 on uh number of words in the data set. Does it make sense to keep going up? Is there a limit? We are one e to the 13 on uh number of words in the data set. Does it make sense to keep going up? Is there a limit? Uliagiver gave a talk at Nurips last year basically saying data is the new oil. we don't have an infinite sort of amount of data left. We're going to run out of data for pre-training. We'll need to find other scaling laws.”

0.56

Global developer growth is absolutely driven by AI 100%, as evidenced by a sharp inflection in computing-related patents from 1995-2003 and again recently, indicating we are at the beginning of a new revolution with people doing new things and innovation popping

factualhigh valueestablishednovelty 1/4durability 2/4· Unknown Speaker (Nate - Analyst)

“This is one of the eye openeners for me. This is absolutely driven by AI 100%. Um, and I think that when you see that pop 1995 to 2003, and this is like an even sharper line, at the end of the day, this is indicating we're at the beginning of a new revolution. Like, people are doing new things and the innovation is really popping.”

0.56

The story of AI can be understood through two trend lines: one showing metrics that are up and to the right (user growth, capex, compute) and one showing metrics that are down and to the right (inference costs, model training efficiency), representing diverging trajectories for adoption vs cost.

definitionhigh valuespeaker onlynovelty 2/4durability 4/4· Nate (Host/Analyst)

“basically the story of AI is the story of charts that are up and to the right and charts that are down and to the left or down and to the right my brain. So the up and to the right ones we know about, right? The user gain is the popular one, the capex growth, etc. The down and to the right is how cheap AI is becoming. like that's just a straight vertical cliff.”

0.56

AI capability improvements on benchmark tests (like passing the Turing test) are less important analytically than understanding where models are overfitting or saturating on those benchmarks, because saturation indicates diminishing returns and the need to shift to different evaluation methods.

normativehigh valuespeaker onlynovelty 2/4durability 4/4· Nate Mastersonn (channel host)

“System performance on bench test. Honestly, like she doesn't get into this, but this is like it's not really to me about surpassing the human baseline. I mean, it is. It's also about the models overfitting and saturating, and she doesn't get into that.”

0.54

The Volta GPU architecture iteration (circa 2018) dramatically reduced energy required per token generated, making AI economically viable, and this hardware innovation took 7-8 years to unfold globally and drive machine learning investment waves, with most users still unaware that legacy Volta drives current AI deployments.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“If you look for the true roots of AI, Volta may be it because so much of the gain in energy required per token came at the Volulta iteration for Nvidia. When it became that cheap in 2018, well, a year later, you suddenly see big machine learning investments. In a sense, hardware GPU innovation can take years to unfold. Here we are 7 years later, 8 years later, we're starting to see the impact of Volulta across the globe and no one uses Volta anymore. It's just that this innovation was enough to change the unit economics for AI. One of the untold stories of AI.”

0.52

Performance is converging across models as servers become cheaper, data centers proliferate, and techniques to make models proliferate across the industry, indicating a multi-model world where differentiation becomes harder as costs commoditize.

factualhigh valuespeaker onlynovelty 2/4durability 3/4· Nate Mastello (inferred)

“performance is converging across models as all of these servers are cheaper, as all of the the data centers proliferate, as the techniques to make these models proliferate, everything is converged. This is what I mean about a multimodel world.”

0.52

ChatGPT users show pessimism about AI in survey responses (Pew Research), yet simultaneously show high quantitative engagement with AI tools (20 minutes/day average across millions of users), indicating a significant disconnect between stated attitudes and revealed behavior that undercuts pessimistic survey narratives.

factualhigh valuespeaker onlynovelty 2/4durability 3/4· Nate Mastersonn (channel host)

“What she doesn't get at, by the way, yes, usage is rising, but the difference in estimates versus what Pew surveys say is notable. And Pew is also the company that has been or the survey company that has been reporting the most pessimism about AI usage among American adults. And so I'm a little bit shocked that we see this big a disparity...people lie on surveys all the time. Um, and I think she doesn't do us a service by not calling out how pessimistic Pew survey respondents are about AI. And I think that's definitely a factor. Again, quantitatively, people are using AI a ton, whatever they say. Like, you can actually see this. Um, and it is really hard to get minutes out of someone's day. So, being able to carve out 20 minutes from someone's day on average across millions of users is huge.”

0.52

The question of whether ChatGPT can build effective advertising models and monetize high-intent search users (similar to Google's model) remains genuinely open and is a critical unanswered question for OpenAI's long-term financial viability.

factualhigh valuespeaker onlynovelty 2/4durability 3/4· Nate (Host/Analyst)

“Uh, I think it's going to be really interesting to see are these high intent searches. How does Fiji build ads at OpenAI? Those are really open questions.”

0.52

AI's energy consumption has catalyzed a downstream revolution in nuclear power policy in the US, with policymakers greenlighting significant new nuclear power projects to meet AI data center electricity demand

factualhigh valuespeaker onlynovelty 2/4durability 3/4· Unknown Speaker (Nate - Analyst)

“One of the things that I think is interesting is that this has catalyzed a downstream revolution in nuclear power in the US where we're now greenlighting a lot of nuclear power.”

0.52

Film directors like Christopher Nolan and Steven Spielberg improve adaptively over their career with each new film they make, but it is unclear how an AI system released into production would achieve similar adaptive improvement over time or whether such improvement would maintain alignment with original design intent.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Christopher Nolan gets better every time he makes a full-length film. Steven Spielberg arguably gets better and has been getting better for decades every time he makes a film. But it is not clear how an AI in the wild would adaptively get better after being released and if it did what that would mean uh in terms of how quickly it would improve. Those are all very unanswered questions and they do raise alignment questions like would the AI stay aligned if that were the case.”

0.52

AI general intelligence (AGI) represents a phase shift in capability that, if achieved even partially, will reshape how humans think about work and decision-making in institutions, representing a high-stakes outcome that justifies serious discussion even if full AGI probability is uncertain.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“AGI. What is AGI? Um, you know, Sam says that we can build AGI. We'll see if he's right. And the the thing that she calls out is like this is a phase shift in capability and how does it reshape things? And I think that's actually a fairly mature way to look at it. We will see how far we get. But if it's even a part of the way toward AGI, we are still going to see some of that phase shift in capabilities and it will reshape how, you know, we think about work and decision-m in institutions.”

0.50

Sam Altman has stated he does not believe ChatGPT will eat Google, suggesting he views the two services as addressing different use cases and that Google will maintain meaningful utility and market position despite ChatGPT's growth.

factualhigh valueestablishednovelty 0/4durability 2/4· Nate (Host/Analyst)

“Does that mean they're going to actually eat Google? Sam Alman said he didn't think so. He thinks that it's going to be different. He doesn't see a case where Google really disappears.”

0.49

IT jobs as tracked by Department of Labor show a 9% decline but this is misleading because the category is basically flat versus 2018 when accounting for a massive tech job increase from the cryptocurrency era followed by an AI-driven uptick, making it a reallocation rather than a net loss.

causalhigh valuespeaker onlynovelty 2/4durability 2/4· Nate (Host/Analyst)

“this is one where I really like we'll get into this, but I think this is just poor framing. I do not think IT jobs is the right frame for tech. And that's a categorization issue from the Department of Labor that I just don't think is super helpful. If you look at the trends more broadly, and this kind of shows it, down 9% is a little bit disingenious. And the reason why is because it's basically flat versus 2018. We had the massive Xerpier run up in all tech jobs and now you have a runup in AI.”

0.49

Enterprise AI adoption targeting is focused on production output, customer success, and sales as the top three areas, with marketing surprisingly low in adoption priority, suggesting enterprise focus is on cost centers and operational efficiency rather than revenue generation, contrary to what leaders anecdotally report.

factualhigh valuespeaker onlynovelty 2/4durability 2/4· Nate (Host/Analyst)

“what's interesting here is this is where generative AI is targeted at large companies. And I think it's interesting that uh production output, customer success, and sales are the top three. And I'm a little bit surprised that marketing is so low. I I would have expected more interest in marketing. And I think that this is one of the slides that surprised me the most because anecdotally the the line I have heard from a lot of leaders is not revenue focused. It is cost focused.”

0.49

AI agents are currently generating high interest among investors and businesses, but actual successful deployments and execution capabilities lag significantly behind interest levels, suggesting the hype-to-reality ratio for agents is higher than for other AI applications.

factualhigh valuespeaker onlynovelty 2/4durability 2/4· Nate Mastersonn (channel host)

“She calls out agent growth. This is interest. I think it's really interesting that it's interest because personally, interest way outweighs the ability of someone to execute on this agent side of things. It's really fascinating.”

0.49

CEO adoption of cloud technologies was previously constrained by uncertainty, but now CEOs are willing to jump to cloud because of the power of AI, suggesting a structural shift in cloud adoption willingness driven by AI capabilities.

causalhigh valuespeaker onlynovelty 2/4durability 2/4· Nate Mastello (inferred)

“and one of the things that I think is a good anecdote is that CEOs anecdotally are saying that they are willing to jump to cloud when they weren't before because of the power of AI and so that's something to keep in mind when you look at the cloud AI relationship”

0.49

GPT-4.5 (or equivalent model version) is being rolled back from user-facing deployment despite having superior capabilities compared to GPT-4, likely because the model is extremely expensive to serve and the company is optimizing for cost-effectiveness rather than pure capability.

causalhigh valuespeaker onlynovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“AI performance, uh, GPT4.5. The irony is they're rolling this back, even though it is as good as she's describing here, I think implicitly because the model is so expensive to serve, they're trying to figure out how to serve it uh, in the right way with GPT5.”

0.49

Self-reported AI usage estimates in surveys exceed behavior-based measurement estimates significantly, and Pew survey respondents are notably pessimistic about AI, suggesting a discrepancy between stated skepticism and actual usage behavior.

factualhigh valuespeaker onlynovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“What she doesn't get at, by the way, yes, usage is rising, but the difference in estimates versus what Pew surveys say is notable. And Pew is also the company that has been or the survey company that has been reporting the most pessimism about AI usage among American adults. And so I'm a little bit shocked that we see this big a disparity. I guess people lie on surveys all the time.”

0.49

Enterprise AI adoption is focused primarily on production output, customer success, and sales, with surprisingly low marketing focus, suggesting that cost-focused and efficiency-focused deployments dominate over revenue-focused deployments, though there are measurement concerns with the categorization.

factualhigh valuespeaker onlynovelty 2/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“traditional enterprise AI adoption increasing priority. This is the whole S&P 500 and what they value enterprise AI focus. Um and what's interesting here is this is where generative AI is targeted at large companies. And I think it's interesting that uh production output, customer success, and sales are the top three. And I'm a little bit surprised that marketing is so low. I I would have expected more interest in marketing. And I think that this is one of the slides that surprised me the most because anecdotally the the line I have heard from a lot of leaders is not revenue focused. It is cost focused. And so if she is correct and she actually is highlighting a trend that they're not talking about, but they're acting out, I think that's interesting.”

0.48

AGI is characterized as a phase shift in capability and reshaping of how people think about work and decision-making in institutions, and the question is how far we get toward it and what that phase shift enables.

definitionhigh valuespeaker onlynovelty 1/4durability 3/4· Nate Mastello (inferred)

“AGI. What is AGI? Um, you know, Sam says that we can build AGI. We'll see if he's right. And the the thing that she calls out is like this is a phase shift in capability and how does it reshape things? And I think that's actually a fairly mature way to look at it. We will see how far we get. But if it's even a part of the way toward AGI, we are still going to see some of that phase shift in capabilities and it will reshape how, you know, we think about work and decision-m in institutions.”

0.48

Interest in AI agents vastly outweighs the current ability to execute agent implementations, suggesting there is significant hype ahead of substance in the agent space

factualhigh valuespeaker onlynovelty 1/4durability 3/4· Unknown Speaker (Nate - Analyst)

“She calls out agent growth. This is interest. I think it's really interesting that it's interest because personally, interest way outweighs the ability of someone to execute on this agent side of things. It's really fascinating.”

0.47

AI functions as a meta-technology that enables organizations to accomplish their existing missions more effectively (as seen with Google, Facebook, Microsoft optimizing against AI to improve search, social networks, productivity), rather than fundamentally disrupting those missions.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“This is a meta technology that enables us to do a lot of other things more effectively including use the internet, including uh make it easy to do business everywhere, including organize the world's information. Like you can see the efforts of these major companies as essentially efforts to optimize against a meta technology that makes it easier to fulfill their mission.”

0.46

Meeker's 340-page deck is attempting to connect disparate data points and narrative threads about AI into a coherent framework, and the deck's structure (from momentum/capex through usage/cost/growth to monetization to robotics to global impact) exemplifies good practice in organizing complex information.

factualhigh valuespeaker onlynovelty 1/4durability 4/4· Nate (Host/Analyst)

“I think the key thing is that Mary is looking at this as a collective effort. So 340 pages, she's trying to connect, as she says, several disperate data points and it just she didn't expect it to get this big either, right? It turned into a beast and she wants to find a way to make sense of it.”

0.45

AI inference costs have declined 99.7% in two years, representing an unprecedented speed of cost reduction compared to the light bulb which took 75 years to achieve a 99% cost decline, demonstrating the accelerating pace of technological cost reduction.

factualhigh valuespeaker onlynovelty 2/4durability 1/4· Nate (Host/Analyst)

“AI inference costs are 99.7% lower. Cost efficiencies. This is the light bulb which took 75 years. This is chat GPT that took two years uh to go down 99%. It's insane.”

0.45

GPT-4.5 is being rolled back from deployment despite excellent performance, likely because the model is so expensive to serve that OpenAI is trying to figure out how to serve it efficiently with GPT-5 or other strategies, indicating that deployment decisions are cost-driven rather than capability-driven.

causalhigh valuespeaker onlynovelty 2/4durability 1/4· Nate (Host/Analyst)

“The irony is they're rolling this back, even though it is as good as she's describing here, I think implicitly because the model is so expensive to serve, they're trying to figure out how to serve it uh, in the right way with GPT5.”

0.45

AI adoption agents (like Operator, etc.) are currently marked by 'sketchy implementations' that don't work as well as advertised, with the exception of internal AI agents built by large AI companies using proprietary tools like 11 Labs or custom LLaMA implementations.

factualhigh valuespeaker onlynovelty 2/4durability 1/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“AI agent deployments. uh you do see some right like uh these are agent deployments in op software I would say operator got much better in the last week and a half when 03 was released but by and large these are very much sketchy implementations um and they do not work as well as advertised and I think it's a little bit disingenuous to call these the flagship agents”

0.45

The cloud revenue model has scaled linearly for AWS, Microsoft, and cloud providers broadly, with new entrants like IBM Cloud, Oracle Cloud, and regional players becoming more serious competitors, creating a multi-cloud world where cloud AI adoption is now a CEO-level priority for legacy enterprises not previously cloud-committed.

factualhigh valuespeaker onlynovelty 1/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“you have AWS, you have Microsoft, uh but you have some new players coming in too. Uh IBM cloud is nothing to sneeze at now. It was quite small in 2016, but it's starting to come in. Alibaba cloud was becoming well it was becoming relevant here but it's had trouble scaling up past that and we have newer players like Oracle cloud which is becoming more serious since 2020 and so it's really becoming a multi cloud world and one of the things that I think is a good anecdote is that CEOs anecdotally are saying that they are willing to jump to cloud when they weren't before because of the power of AI”

0.44

Tens of billions of GPU units are being deployed globally as part of the AI infrastructure buildout, though speaker notes he is looking at Johnny Ive and Apple's directions and expects this to change soon as OpenAI releases new device form factors.

factualhigh valuespeaker onlynovelty 2/4durability 2/4· Nate Mastello (inferred)

“She's talking about GPUs here, but I actually looking at Johnny Iive and I'm looking at like where they're going with devices. Feels like that's coming pretty fast. So I would not be surprised if this is an entirely looking different looking slide a year from now when the devices uh that OpenAI are is working on are released.”

0.41

Global CMOs (Chief Marketing Officers) show AI implementation heavily skewed toward initial testing rather than full implementation, but this distribution has likely changed substantially since end of 2024, suggesting Meeker's data may already be outdated.

factualhigh valuespeaker onlynovelty 1/4durability 1/4· Nate (Host/Analyst)

“enterprise AI focus global CMOs here we are jumping to marketing right what are they doing they're fully implemented not so much running initial tests yes this is changing so fast like I I think that you actually have changes in this even since the end of 2024 that are quite substantial like I would have expected this graph to be like up here”

0.39

There may be slight erosion in Google search retention even if the overall metric looks stable as a table, suggesting potential early warning signs of competitive pressure from ChatGPT.

factualhigh valuespeaker onlynovelty 1/4durability 2/4· Nate Mastello (inferred)

“And I just want to call out it's stable as a table, but it might be a little bit bigger here than it is here. And that may be that erosion that Google gets so stressed about due to AI potentially.”

0.36

The estimated timeline to 50% household technology adoption is three years for the AI era, which is much faster than desktop internet adoption (12 years) and PC adoption (20 years), though the actual timeline is still being validated against current trajectories.

forecasthigh valuespeaker onlynovelty 1/4durability 1/4· Nate (Host/Analyst)

“years to 50% adoption of household tech AI era. They're guessing three years or we're not there yet, right? So it is a guess but it's it's on track for something very fast which is for the record uh much faster than desktop internet which was at 12 12 years and PC which is at 20 years which is just crazy”

0.32

Usage of AI at work reportedly improves quality and productivity, according to survey respondents, suggesting that users perceive real value from AI tools in professional contexts, though survey results may reflect social desirability bias.

factualestablishednovelty 0/4durability 2/4· Nate Mastersonn (channel host)

“AI chat bots at work. Uh what are people saying? It improves their quality and allows them to do more things more quickly. I don't think this is super informative.”

0.24

Demis Hassabis (CEO of Google DeepMind and Nobel Prize winner) presents a bull thesis for AI: first solve AI, then AI solves everything, and he remains very bullish on this vision despite it being a simplified narrative of how AI benefits materialize.

factualestablishednovelty 0/4durability 2/4· Nate (Host/Analyst)

“she has uh Deiss Hapabis Hassabis uh CEO of Google DeepMind Nobel Prize winner. Um this is basically the bull thesis for AI, right? First we solve AI then AI solves everything. Okay. He's still very bullish on that.”

0.24

Deep research and deep search tools represent automation of specialized knowledge work where users need extensive reasoning across the internet, and these tools now exist at scale

factualestablishednovelty 0/4durability 2/4· Unknown Speaker (Nate - Analyst)

“They're going back to like blue books and pencils usage expansion. Um so deep research deep research deep search blah blah blah. She calls it automating specialized knowledge work. I think it's more narrower than that. I think it's basically I need to do a lot of reasoning across the internet. Can you help me? And there are a lot of tools for that now.”

0.24

Nvidia's computing power improvements follow one of the clearest exponential curves ever measured, with sustained 150% annual performance gains over six years, making it one of the most beautiful examples of Moore's Law continuation in modern hardware.

factualestablishednovelty 0/4durability 2/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Nvidia computing power is this like one of the most gorgeous exponential curves I've ever seen. It's just pretty. Uh Nvidia global data center capex also scaling really really fast.”

0.20

Machine learning models show clear academic trends tracked through Stanford, and Mary Meeker has a specific interest in how AI circulates through academic institutions and how findings from academia can be presented at scale.

factualspeaker onlynovelty 0/4durability 3/4· Nate Mastello (inferred)

“I think that the idea like she has an interest in in academia anyway. She wrote one of her rare decks that was not as widely circulated on AI in educational institutions. And so when she you know references Stanford and talks about the academia era um I think that's something that has been an interest for her for a long time.”

0.20

The analyt does not fully agree with Mary Meeker's framing of geopolitical risk factors as the most important risks to AI development, viewing other risk categories as more compelling and pressing

factualspeaker onlynovelty 0/4durability 3/4· Unknown Speaker (Nate - Analyst)

“I don't know that I agree with Mary on the global competition frame, and we'll get into why.”

0.19

ChatGPT passes the Turing test, but Meeker's deck may not emphasize this sufficiently because it's not surprising to investors who track AI, though it may be novel to traditional investors unfamiliar with recent AI progress.

factualestablishednovelty 0/4durability 1/4· Nate (Host/Analyst)

“AI performance, realistic conversation, touring tests. We know it passes the touring test. Maybe the investors don't.”

0.19

AI-generated images from models like DALL-E are very good, and the quality is not surprising to audiences familiar with AI progress, as demonstrated by Meeker's slide showing a realistic image and AI-generated image pair that catches people by surprise.

factualestablishednovelty 0/4durability 1/4· Nate (Host/Analyst)

“The images are very good. Now, this should not be to anybody's surprise. The images are very, very good. I love this real image, AI generated image. That's a lovely little reverse. Catches people by surprise.”

0.19

Mary Meeker released a 340-page AI trends and artificial intelligence deck four days ago, her first AI-focused deck since 2019 when she stopped doing internet trends decks, driven by her shift to investor activities at her VC firm.

factualestablishednovelty 0/4durability 1/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“Mary Miker is a general partner um and she is an investor now and that is really the core reason why she has not been doing these trends in artificial intelligence uh decks earlier. She did not do one in 2023. She did not do one in 2024 and before it was internet trends and that last one was back in 2019.”

0.19

ChatGPT was initially not very capable in its early form and was dismissed by some observers as only suitable for marketing copy, but has evolved dramatically through successive versions to become genuinely useful for complex tasks.

factualestablishednovelty 0/4durability 1/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“chat GPT 2022 and it was not that smart guys. I don't know if you've forgotten but it kind of dumb to start with. I remember when I was like oh this is just going to be used for marketing copy.”

0.19

AI-generated images have reached very high quality and realism, with AI-generated images now indistinguishable from real photographs, as evidenced by Meeker's side-by-side comparison of real and AI images.

factualestablishednovelty 0/4durability 1/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“The images are very good. Now, this should not be to anybody's surprise. The images are very, very good. I love this real image, AI generated image. That's a lovely little reverse. Catches people by surprise.”

0.19

AI passes the Turing test reliably and consistently, meaning it can convince humans in conversation that it is human, though investors may not be aware of this achievement despite it being a realized milestone

factualestablishednovelty 0/4durability 1/4· Unknown Speaker (Nate - Analyst)

“AI performance, realistic conversation, touring tests. We know it passes the touring test. Maybe the investors don't.”

0.17

Mary Meeker did not release a trends in artificial intelligence deck in 2023 or 2024 because her investment activities as a general partner have been front and center, and she is only now taking time aside from that to brief the industry.

factualspeaker onlynovelty 0/4durability 2/4· Nate (Host/Analyst)

“she has not been doing these trends in artificial intelligence uh decks earlier. She did not do one in 2023. She did not do one in 2024 and before it was internet trends and that last one was back in 2019. And so her investment activities have been front and center for her.”

0.17

Deep research and deep search tools are automating specialized knowledge work by enabling users to search and synthesize information across the internet with reasoning capabilities, representing an emerging product category that combines search with language model reasoning.

definitionspeaker onlynovelty 0/4durability 2/4· Nate Mastersonn (channel host)

“Deep research deep research deep search blah blah blah. She calls it automating specialized knowledge work. I think it's more narrower than that. I think it's basically I need to do a lot of reasoning across the internet. Can you help me? And there are a lot of tools for that now.”

0.17

The overall organization of Meeker's deck is strong: it structures coverage through development/developers, AI user usage and capex growth, monetization, physical and robotics, and global and work evolution, providing a logical framework for understanding AI's impact

factualspeaker onlynovelty 0/4durability 2/4· Unknown Speaker (Nate - Analyst)

“I actually really like the way she structured this. If anyone is wondering, how do you structure a gigantic deck so it feels understandable? Well, this is one way to do it, right? The overall takeaway seems like change is happening faster than AI user and capex growth. The headlines we tend to see what is driving that compute. How does this start to translate into the market? We get into usage, cost, and growth. We get into monetization. we get into physical and robotics. Um, and then finally, we start to get into global and work evolution at the end. That's a pretty good organization.”

0.17

Demis Hassabis, CEO of Google DeepMind and Nobel Prize winner, articulates the bull thesis for AI as 'first we solve AI, then AI solves everything,' representing an optimistic view that remains actively held

factualspeaker onlynovelty 0/4durability 2/4· Unknown Speaker (Nate - Analyst)

“She has uh Deiss Hapabis Hassabis uh CEO of Google DeepMind Nobel Prize winner. Um this is basically the bull thesis for AI, right? First we solve AI then AI solves everything. Okay. He's still very bullish on that.”

0.13

Meeker does not discuss Spotify accepting AI-powered music and AI-generated music, which is a different conversation from audiobook translation, suggesting her deck focuses on certain applications while omitting others.

factualspeaker onlynovelty 0/4durability 1/4· Nate (Host/Analyst)

“What this she doesn't talk about is uh Spotify accepting AI powered music and AI music which is a different conversation.”

0.13

The speaker estimates that within one hour of presentation, Meeker covered approximately 143 slides from her 340-page deck, establishing a pace that would require the full deck to take approximately 2.4 hours to present in full detail.

factualspeaker onlynovelty 0/4durability 1/4· Nate Mastersonn (channel host)

“Uh, so we'll call this part one. We got through 143 slides in about an hour.”

0.13

Meeker's deck notably omits discussion of AI-powered music generation and AI music composition, despite covering AI voice, representing a significant gap in discussing emerging AI applications that are already in production use.

factualspeaker onlynovelty 0/4durability 1/4· Unidentified Speaker — AI Trends 2025: Mary Meeker Deck Deep Dive Part 1 [_g1LFxC31EY]

“What this she doesn't talk about is uh Spotify accepting AI powered music and AI music which is a different conversation.”