AI Markets: Deep Dive with a16z's David George
Summary
AI-native demand is separating sharply from the rest of software. George says AI companies are growing more than 2.5x faster, with top performers at 693% year over year and the fastest reaching $100 million in revenue much sooner than SaaS predecessors. “AI demand is crazy,” yet the fastest-growing companies spend less—not more—on sales and marketing.
The strongest AI companies pair extraordinary growth with $500,000-$1 million of ARR per employee. That compares with a roughly $400,000 SaaS-era rule of thumb, while lower gross margins can be a “badge of honor” when high inference costs indicate customers are actually using AI features. George cautions that strong demand, lean staffing and general efficiency gains after the bloated 2021 era explain much of today’s efficiency; wholesale AI-driven organizational redesign remains early.
Pre-AI companies must “adapt to the AI era or die” across both products and internal operations. One founder gave two AI-fluent engineers unlimited access to Claude Code, Codex and Cursor; they rebuilt a product he was frustrated with at an estimated 10-20x faster pace, prompting him to rethink the product-and-engineering organization within 12 months. The extreme operating question is now: “Can I do it with electricity or do I need to do it with blood?”
Engagement data makes the best application revenue look durable rather than experimental. George says Harvey users spend roughly twice as much time in the product; Abridge maintained or increased engagement while rapidly adding clinicians; and Navan now handles 50% of travel interactions with AI, helping gross margins expand 20 percentage points over three years. George cites Flock as solving 700,000 crimes annually, with officers clearing almost 10% more where it operates.
Enterprise intent is running well ahead of implementation, creating a widening execution gap. Fortune 500 leaders say they must become AI companies, but George calls change management—not model readiness—the central constraint. Early results show the stakes: Chime cut support costs 60%, while Rocket Mortgage saved 1.1 million underwriting hours and reached $40 million of annual run-rate savings.
AI winners have produced almost 80% of the S&P 500’s return, but George sees earnings rather than speculative multiple expansion underneath the rally. Multiples are above average yet far below dot-com levels, and investors favor profitable growth over the loss-making growth rewarded in 2021. His durable factor remains growth: “Ultimately, growth is the biggest thing that drives returns over five to 10 years.”
The infrastructure buildout has bubbly features, but utilization and financing still differ materially from prior bubbles. Hyperscalers are supported largely by historically profitable companies and cash flows, seven- to eight-year-old Google TPUs remain fully utilized, and rental pricing for A100s and H100s has held up—hence the relayed line, “There are no dark GPUs.” The watchpoint is debt: Oracle is making a large, cash-flow-negative cloud bet, while its credit-default-swap cost has risen to roughly 2%.
The payback hurdle is enormous and may extend well beyond 2030, while private markets are now a major asset class. Against roughly $4.8 trillion of cumulative hyperscaler capex, annual AI revenue must approach $1 trillion by 2030—about 1% of global GDP—to clear a 10% hurdle rate; George’s rough current estimate is only $50 billion, albeit growing well above 100%. Meanwhile, about 86% of companies above $100 million in revenue remain private, and the ten largest North American and European unicorns hold almost 40% of a $5.5 trillion valuation pool.
Deep dive
1. AI-native demand is accelerating without a sales-spend crutch
George’s opening conclusion is that 2025 reversed the revenue slowdown that followed the 2022-24 rate hikes. Growth accelerated across cohorts, particularly among outliers, inside what he views as a 10- to 15-year product cycle that has barely started.
Kha’s definition check matters: the AI cohort is mostly post-ChatGPT, with some grace for companies founded around then, and comprises companies whose first market product was AI-native—not older businesses that merely used machine learning.
AI companies in a16z’s internal dataset grow more than 2.5x faster than non-AI peers; the top AI performers reached 693% year-over-year growth, a result the team “had to triple-check.” The fastest also reach $100 million of revenue considerably sooner than leading SaaS-era companies.
The mechanism is demand, not purchased growth: the fastest AI companies spend less on sales and marketing than SaaS counterparts. Lower gross margins can even be a “badge of honor” when high inference expense indicates that customers are actively using the AI.
2. Incumbents must rebuild both product and operating model
George’s prescription is categorical: “You need to adapt to the AI era or die.” On the front end, incumbents must redesign workflows rather than attach chatbots; internally, they must deploy the latest coding models for developers and the latest tools across every function.
His sharpest example is a pre-AI founder who assigned two AI-fluent engineers to rebuild a product he was frustrated with using Claude Code, Codex and Cursor, under an unlimited tools budget. He estimated progress was 10-20x faster, and the unusually high tool bill made him reconsider the structure of the entire product-and-engineering organization.
George calls December a turning point for coding. Over the next 12 months, he says this will either take hold in companies or they will move much more slowly than their peers. Kha adds that even post-AI companies revisit six-month-old systems because current tools can improve them vastly, raising the catch-up burden for incumbents.
Business-model disruption is less advanced: enterprise software moved from licenses to seat-based SaaS, then consumption pricing, with outcome pricing next. Customer support may be the only area where it is feasible today because resolution is measurable; broader adoption depends on models enabling objectively measurable outcomes.
3. Engagement, not headline ARR, is the durability test
George says the team goes beneath rapid revenue growth into retention, renewals and observed product activity. At Harvey, improved products and reasoning models have roughly doubled time spent because “lawyering and reasoning go hand in hand.”
Clinicians describe Abridge as a “trusted deputy.” As its user count expanded sharply, engagement held steady and even rose slightly—the opposite of the dilution that would suggest weaker incremental users.
ElevenLabs combines staggering voice-usage growth with unusually efficient operations. Navan supplies the operating proof: AI now handles 50% of complex travel-booking and change interactions, contributing to a 20-percentage-point gross-margin expansion over three years.
Flock’s customer proposition is unusually concrete: solving crime. George cites 700,000 crimes solved annually and almost 10% more clearances per officer where Flock is present.
4. Enterprise ambition is running ahead of implementation
Asked to calibrate Fortune 500 adoption, George emphasizes the gap between stated urgency and deployed change. CEOs say, “We’re going to become AI companies,” but replacing business processes and overcoming organizational resistance remain much harder than approving an assistant.
Coding is comparatively easy to understand, while customer support offers an obvious “better, faster, cheaper” case. General management and process redesign are harder; Kha notes that companies may still be rebuilding data systems and back ends before benefits become visible.
The early pockets are material: Chime reports 60% lower support costs, and Rocket Mortgage reports 1.1 million underwriting hours saved—6x the prior year—and $40 million in annual run-rate savings. George expects a five-year productivity reckoning between adopters and laggards.
5. Public-market gains rest on earnings, but concentration is extreme
AI winners account for almost 80% of the S&P 500’s return. George sees “minimal” evidence of froth because recent gains primarily reflect EPS growth while multiples have contracted, especially for SaaS; valuations exceed historical averages but remain nowhere near dot-com levels.
The market rewards the high-growth, high-margin quadrant most, while low-growth, low-margin companies trade poorly. Even high-margin companies struggle without growth, which George calls the largest driver of five- to 10-year returns.
Goldman Sachs estimates the AI buildout could generate $9 trillion of revenue. At 20% margins and 22x earnings, that implies $35 trillion of market capitalization versus roughly $24 trillion already pulled forward—though George notes that not all of that increase is necessarily attributable to AI.
His “model buster” analogy is Apple: four years after the iPhone, consensus estimates had understated performance by 3x. He expects pockets of AI to exceed spreadsheet forecasts similarly, creating value far beyond the capital required.
6. Capex looks productive now, but debt and payback are the watchpoints
The buildout is massive and concentrated, making it inherently risky, but it is financed primarily by historically profitable companies. Azure took seven years to reach one year of AI revenue and ten years for its revenue to surpass its capex; George expects AI’s ratio to improve faster.
Cash flow cannot finance every forecast data-center project, so debt and private credit are entering. George is comfortable with Meta, Microsoft, AWS and Nvidia as counterparties, but stresses that “not all counterparties are the same”; Oracle’s large cloud commitment will leave it cash-flow-negative for years, while its CDS cost has risen to about 2%.
Depreciation fears have not yet appeared in utilization: Google disclosed 100% utilization for seven- to eight-year-old TPUs, and rental pricing for A100s and H100s remains resilient. Gavin Baker’s comparison, relayed by George: fiber could remain dark, but “there are no dark GPUs.”
Public software companies added $46 billion of revenue in 2025; OpenAI and Anthropic alone added almost half of that on a run-rate basis. Yet roughly $4.8 trillion of capex requires about $1 trillion of annual AI revenue by 2030 for a 10% return, versus George’s rough current estimate of $50 billion, growing far faster than 100% year over year. He thinks payback may continue between 2030 and 2040.
7. Private-market power laws are intensifying around AI
The number of public companies has halved over 20 years, while roughly 86% of businesses above $100 million in revenue remain private. For George, private growth investing is now a substantive asset class.
North American and European unicorns carry about $5.5 trillion of aggregate value; the ten largest account for almost 40%, double their 2020 concentration. George estimated, while counting in real time, that seven of those ten are a16z portfolio companies.
Disruption is accelerating in public markets too: the average S&P 500 company’s tenure in the index has declined about 40% over 50 years. George acknowledges that an orderly private-market price path can help with employee retention, hiring and morale, but expects some large, long-private companies to go public over the next 18 months.
Databricks illustrates successful pre-AI adaptation. George credits Ali’s combination of commercial instinct and technical depth, the data platform’s suitability for AI workloads, aggressive iteration through Agent Bricks, and validation from cutting-edge customers; modern AI companies choosing the platform lets Databricks grow alongside them.