The $1 Trillion AI Buildout | State of Markets

1. Key Themes

Tech Is the "Everything Cycle," and It Is Earnings-Driven, Not Multiple-Driven

The discussion opens with the scale of tech's share of the economy and then rebuts the bubble thesis with valuation data. David Pawlan: "high-tech equipment, software, and R&D now account for roughly 55% of U.S. capital spending, which is just a staggering number." 00:04:01. On valuation: "stocks are up about 20% while multiples are down about 20%. So, what that means is the performance is driven by fundamental earnings, right? Not increased multiples." 00:05:47. Alex Immerman adds the growth context: "the market's up 90%, which is 17% annualized... when you compare that to... the market trading for below 20 times earnings, growing 15%, this definitely feels a little different than maybe the 2021 period that was recent or the 2000 period when multiples and growth were not really going together." 00:00:05. Pawlan contrasts this with the dot-com era ("would trade, in many cases, for 100 times PE") and notes memory companies trade at "six times, seven times forward earnings." 00:06:16

Hyperscaler CapEx Keeps Getting Underestimated, and Demand Exceeds Supply Everywhere

Pawlan: "their CapEx in 2026 is about $780 billion. That's up from $416 billion in 2025. And all expectations point to them spending over a trillion dollars annually from 2027." 00:07:13. The recurring error is forecasters assuming a plateau. Immerman: "for the last four years, like, as an economy, we keep underestimate just the strength of the trend." 00:09:01. Santiago Rodriguez points to OpenAI pausing new subscriptions: "two weeks ago, we all saw that they had to pause new subscriptions on their pro plans, like showing up with a $2,000 service, no thanks. Pretty amazing, insatiable demand." 00:09:25. Pawlan: "There are certain elements in the data center supply chain where you can't get access to materials or products until 2028." 00:10:03

The Buildout Has Demand-Side Backing and a J-Curve Financial Profile

Pawlan cites evidence beyond capex: "Microsoft, Google, and Amazon have about 1.7 trillion of combined cloud backlog together," with free cash flow depressed now and "consensus forecasts... showing recovery from 2028." 00:10:32. He points to Amazon's explanation that "the useful life of GPUs is actually pretty... long... it's been longer than I think any of us expected." 00:11:02. Sarah Wang adds that aggressiveness has been rewarded: "the ones who have blinked and been less aggressive have regretted it." 00:11:24. Immerman notes spot-market GPU pricing as a signal: "any GPU that you can bring online is being priced at an attractive rate." 00:11:43

The Tech Boom Becomes an Industrial Boom: A New Age of Atoms

Rodriguez reframes CapEx: "we should think of it as someone else's order book... it's been a boon for chip orders, for power, for construction. And that's why this broader technology boom has become an industrial boom." 00:12:38. "Global infrastructure investment needs are estimated at $90 trillion through 2040... it includes power, water, roads, transit." 00:00:13. Immerman flags how different the business is: "traditionally, our world was more about monetizing existing IP, you know, build once and then sell infinitely. But with these businesses, there's many more complexities... like financing, managing vendor relationships, predicting capacity and predicting demand." 00:13:06

Enterprise Adoption Is Broad but Shallow: Huge Room to Deepen

Wang: "live deployments at S&P 500 companies, that's at 69%... quantifiable impact... that's 30%. Now, if you go to the ultimate barometer, which is a metric tracked over time, that's actually only at 2%." 00:00:29. She adds: "most of the enterprises that we speak with, anecdotally, their exposure to AI is still mostly with Microsoft Copilot." 00:19:17. Pawlan offers a spending benchmark: "relatively forward-leaning, large Fortune 500 type companies are probably today at like 1%... the most forward-leaning, you know, AI pilled companies in our portfolio can be as high as like 10%" of headcount cost spent on AI tools. 00:22:22

Power Users Are Pulling Away

Wang: "median AI vendor spending in the top 1%, that's roughly eight times the level of the top 10%." 00:20:54. Inside AI-native portfolio companies, "the top users spend anywhere from, call it, $7,500 to $9,000 a month. And the median users... $200 to $400. So over 20 times the spend." 00:21:39

Falling Inference Costs Unlock Jevons-Style Demand and Better App Margins

Wang: "You're starting to see things like two orders of magnitude cost differences really impact the number of use cases. It's sort of classic Jevons paradox." 00:16:29. Rodriguez: "the 14x growth in agent token usage on OpenRouter," and "Hebbia... financial chat workloads become 10x cheaper to run." 00:26:46. Routing and fine-tuning compound the gains: Databricks' "smart router... solves more problems at 35% lower cost than the strongest individual model," and Eliza AI's fine-tuned model was "60% cheaper" with lower latency, making live audio use cases viable. 00:27:44 Wang: "the relevant unit for them is really the cost of getting the customer's job done... we're going to see more margin improvements in the best app companies." 00:28:43

Revenue Growth Is the Real Test for Incumbent Software

Only "30% of the public software companies are growing at 20% plus." 00:37:04. Pawlan says the fear is dispelled only when companies keep "98% gross dollar retention" while driving "revenue growth acceleration," and the firm's target is "10% plus acceleration." 00:38:14 Differentiation matters: "cybersecurity observability have really stood out. And vertical software as well has generally held up much better than the horizontal applications." 00:39:46

Private Markets Are Now Where the Scale Is

Immerman: "the top six companies today... Anthropic, OpenAI, Databricks, Stripe, Waymo, Revolut by last round valuation, they add up to about 2.4 trillion. This is more than the combined market cap of IPOs we've seen in the last 10 years, excluding SpaceX." 00:43:16. "AI related companies account for 86% of US VC deal activity in this 2026 snapshot up from 65% in 2025." 00:48:27

Agentic Commerce Threatens Ad-Funded Profit Pools

Rodriguez frames the questions incumbents must ask: "how much incremental demand can I get from an AI agent?... how much of my profit pool comes from owning the customer relationship and discovery?" 00:32:32. Pawlan: "Amazon has an over 70 billion advertising business that's extremely high margin... totally predicated on the fact that consumers go to the website and click on the ads." 00:34:03

2. Contrarian Perspectives

Data Centers Lower Electricity Bills

Against the popular narrative that AI infrastructure raises household costs, Rodriguez cites a study: "for every 10% increase in data center capacity, residential rates went down by 40 bps." His reasoning: "a power grid is a shared fixed cost base, poles, wires, substations. And so a large, stable customer like a data center can help spread those costs across more units of electricity." 00:15:33

The Profit Pool Disruption of Agents May Be Net Positive for Everyone, Including Market Caps

While the narrative was "negative sum," Immerman argues: "maybe that $70 billion that's spent on advertising on Amazon just finds a different channel... with the lower friction, we might just see more consumption." 00:35:31. Rodriguez adds that the market reaction has been "net market cap positive across the ecosystem. The meta gains have far exceeded any of the losses from the marketplaces." 00:36:32

The "SaaS Apocalypse" Is Bifurcating, Not Universal

Immerman: "the software index is actually back to, you know, where it started the year, but it's really bifurcated into like some companies that are deemed AI losers and some companies that are deemed winners." 00:42:04. And Stripe's data contradicts the doom narrative: "a little bit of a narrative violation in Stripe's SaaS customer data, which actually shows growth accelerating into 2026 across both young and mature businesses." 00:42:49

Employees at Private Companies Are Declining to Sell

Against the "cashing out" narrative, Immerman: "participation in tenders struck by Carta has only been 58%. So this is employees choosing not to take liquidity because they have so much conviction." 00:46:11. Secondary discounts have also vanished: "the discount to the last round price is basically zero." 00:47:51

Low Consumer Paid Penetration Is Not (Only) a Ceiling

Only "just over 2% of U.S. households actually have a paying subscription for AI," far below Netflix or Prime. 00:29:21 Yet Rodriguez cites Josh Elman's view that "consumer AI is what I use for my daily life and not necessarily what I expense," and that the best assistants may not command the most screen time since they will be "persistent and always on and super proactive." 00:30:54

3. Companies Identified

OpenAI

Frontier AI lab. Combined with Anthropic, its revenue ramp is called a favorite chart. Wang: "the combined scale of revenue for OpenAI and Anthropic... the estimates for NetNew for the leading labs has well surpassed" the best software companies ever built. 00:17:28 Also praised for foresight: "they got a lot of flack a year or so ago for their massive compute commitments... at this point like everyone would say, they're incredibly prescient." 00:09:25

Anthropic

Frontier AI lab, counted among the top six private companies by valuation and in the revenue comparison with OpenAI. 00:43:16

Databricks

Data and AI platform. Its "smart router performed better. It solves more problems at 35% lower cost than the strongest individual model." 00:27:45 Also cited for big new-product bets that drive revenue acceleration. 00:44:15

Stripe

Payments infrastructure. Its SaaS customer data shows "growth accelerating into 2026 across both young and mature businesses," which Stripe calls "the renaissance." 00:42:49 Also cited as a forward-leaning company discussing where to put AI dollars. 00:25:02

Waymo

Autonomous ride-hailing. "Aggressively expanding their depots." 00:14:10 Also counted among the top six private companies. 00:43:16

Revolut

Fintech. "Had to partner with Eleven Labs and take Eleven's best-in-class voice model and harness it to connect securely with customer accounts and banking workflows." 00:20:21 Also among the top six private companies.

ElevenLabs

Voice AI. Its "best-in-class voice model" was the layer Revolut harnessed for customer service. 00:20:21

Chime

Public neobank. "Reduced their cost to serve by over 10% a year for the last four years. Compounded, we're talking about almost a 50% reduction." 00:23:11

Decagon

AI customer support, a portfolio company that helped support Chime's cost-to-serve reductions. 00:23:12

Shopify

Commerce platform. Its AI Sidekick raised "the percentage of customers that reach five orders within 15 days after onboarding" by 8%. 00:23:40

ServiceNow

Enterprise workflow software. "Reported more than a billion in AI ACV and actually a 9x increase in agentic deployments." 00:24:42

Hebbia

AI for finance and knowledge work. "Financial chat workloads become 10x cheaper to run." 00:26:46

Harvey

Legal AI. Uses fine-tuning, and is named among "the fastest growing companies we're seeing in the private markets" that grow "faster than any of the precedents ever have in their industry." 00:41:47

Eliza AI

Vertical AI company (transcribed "Elise"). Fine-tuned a smaller model that was "60% cheaper, but also had way lower latency so live use cases from an audio perspective became tenable." 00:28:14

Abridge

Healthcare AI, among the fastest-growing vertical AI companies. 00:41:17

Cursor and Cognition

AI coding tools. "Many enterprises are really excited when they buy Cursor or Cognition for the first time. But the reality is that's just like the beginning of the journey." 00:22:49

CrowdStrike

Cybersecurity. Highlighted as an example of security and observability holding up as AI creates "greater demand, honestly, for the incumbents in these markets." 00:40:15

Instacart

Grocery delivery, which agreed to integrate with an AI agent while Amazon declined. Rodriguez: "online penetration of groceries, still relatively early, a lot more orders to go get." 00:33:01

Amazon

Hyperscaler and marketplace. Its earnings call explained the GPU J-curve 00:11:02, while its "over 70 billion advertising business" is exposed to agentic commerce. 00:34:03

Microsoft, Google, Alphabet, Meta, Oracle

Hyperscalers whose 2026 capex is about $780 billion. 00:07:13 Microsoft, Google and Amazon hold about $1.7 trillion of cloud backlog. 00:10:32 Google search "has been really resilient." 00:35:03 Meta gains were noted as outsized relative to marketplace losses. 00:36:32

Anduril

Defense tech. "Massive manufacturing facility... 87 football fields." 00:14:10 Also named with newer defense vendors that are under 5% of military spend today. 00:52:06

SpaceX

"$100 billion investment in Louisiana" 00:14:10; also excluded from IPO totals at $1.7 trillion. 00:43:16 Tesla and SpaceX are cited as talent factories for physical-world founders. 00:15:03

Sierra/Saronic/Castelion (defense vendors)

The transcript names "Sironic or Castellian" (likely Saronic and Castelion) as newer defense vendors expected to grow dramatically. 00:52:06

Navan

Travel and expense platform that "went public and we've seen a reacceleration just on the basis of some of the larger enterprises actually trusting them more because they're a public business." 00:45:27

Uber and Lyft

Together about "1% of miles traveled in the U.S." with expected expansion "by at least an order of magnitude" as autonomy scales. 00:50:08

Netflix and Amazon Prime

Benchmarks for subscription scale ("over 200 million households," "70 million households") versus AI's roughly 2% paid penetration. 00:30:17

Meta, TikTok, Snap, Facebook, Instagram

Benchmarks for time spent: "30 to 60 minutes per day from their active users." 00:30:54

OpenRouter

Model routing platform whose data shows "14x growth in agent token usage." 00:26:17

Carta

Cap table platform whose data shows "participation in tenders... only 58%." 00:46:41

YipitData

Data provider used to measure median AI vendor spending by decile. 00:21:08

Microsoft Copilot

Named as the main AI exposure for most enterprises, showing "how far they have to go." 00:19:17

Tesla

Cited for "the factory is the product" and as a source of great founders. 00:15:03

4. People Identified

Sam Altman

OpenAI CEO. Credited, with Sarah Friar, for making "massive compute commitments" that looked reckless but proved "incredibly prescient." 00:09:25

Sarah Friar

OpenAI CFO (transcribed "Sarah Fryer"). Same context as Altman. 00:09:25

Elon Musk

Coined "the factory is the product." Immerman: "a lot of the great founders that we've backed have come out of Tesla or SpaceX because they just learned that... executing on the factory ends up being a massive competitive advantage as they scale." 00:14:37 Also called an exception to public-market short-termism with Zuckerberg. 00:44:15

Mark Zuckerberg

Named with Musk as exceptions who take long-duration bets in public markets; Meta's stock "got below 100 bucks a share" when people doubted AR/VR spending. 00:44:15

Ali Ghodsi

Databricks CEO. Cited for the idea that "the AI can know a lot about the world but know very little about your company" 00:19:52, and for the "chopping block of AI" framework. 00:40:45

Josh Elman

a16z partner. His view: "consumer AI is what I use for my daily life and not necessarily what I expense." 00:30:25

Marc Andreessen

Referenced via "software is eating the world," which Rodriguez says this cycle extends. 00:05:17

Dina Powell

Cited as having discussed lowering electricity costs via data centers. 00:15:59

Gavin

Podcast guest whose episode touched on Microsoft and hyperscaler debt. 00:11:24

Martin Casado

Co-hosted the conversation with Ghodsi and Wang. 00:19:52

5. Operating Insights

Use Where You Put AI Dollars as a Litmus Test

Pawlan describes how the firm reads founders: "where are they actually putting their incremental AI investment dollars?... the opportunity to drive revenue growth is unbounded to the upside. Whereas the opportunity for cost improvement... is sort of a latent opportunity... if you think that the best and highest use of your dollars today is to optimize your cost structure, what does that say about the revenue opportunity for you?" 00:25:02

Don't Spend Engineers Rebuilding Systems of Record

Immerman: "six months ago, our industry was really focused on like rebuilding systems of record internally. And if your engineers are focused on... rebuilding a system of record to save a couple hundred thousand dollars, I expect there to be better uses of like those resources." 00:25:54

Price on the Customer's Job, Not the Model Call

Wang: "the relevant unit for them is really the cost of getting the customer's job done. And if you can decouple that, you get the customer's job done, you charge for that, and you can use better routing to make the product more economical, that's pretty magical because now you're not using the most expensive model for each step." 00:28:43 Tactics cited: caching, routing, fine-tuning.

Keep Private Valuations Fresh via Tenders

Pawlan: "we're advocates of sort of more frequent resetting of your valuation... it's easier to talk about that with employees and prospective employees. But two... in the event that you want to do M&A... you have fresh currency that you can use." He also frames tenders as "a weapon of competition" against public-company RSU liquidity. 00:47:25

Set a Quantified Acceleration Target

For incumbents, the firm's rule of thumb is "target 10% plus acceleration" in revenue growth, evidenced by retention metrics like 98% gross dollar retention. 00:38:44

Pick the North-Star Metric That Predicts Retention

Shopify tracks customers reaching "five orders within 15 days after onboarding" because "once you reach five orders, you know, you're going to stick around." 00:23:40

6. Overlooked Insights

Agents Break Outside-In Engagement Data, Making Investing Harder

Immerman mentions almost in passing: "it's going to make our job harder when we can't look at engagement via outside-in data because the agents are working on the background instead of us just looking at, like, screen time." 00:31:31. This is significant because the standard toolkit for diligencing consumer products (screen time, DAU, time spent) stops working as usage moves to background agents, which favors those with direct access to usage or token data and may mean the best consumer AI winners look invisible in conventional metrics.

The Top 1% of AI Spenders Are Nearly Equal to the Next 9% Combined

Immerman's aside on the YipitData chart: "it's almost as much as the 2% to 10% combined is basically the 1%." 00:21:29. Combined with Wang's 20x spend gap between median and top users, this implies AI vendor revenue is extremely concentrated in a small set of intense users and organizations. That makes vendor revenue more exposed to a few heavy customers, and it suggests that enterprise AI growth is driven by depth of usage rather than seat count, so the real market is a function of how many organizations migrate into the power-user tier.