Meekers AI Report Ten Essential B2B Learnings
So Mary Meeker has been doing extremely well researched, deep analyses of internet trends since the earliest days the web took off. First at Morgan Stanley, then at Kleiner Perkins, and since then, at her own growth VC fund, Bond Capital.
The latest one is all AI with a big enterprise / B2B slant and is very good — but dense. 300+ pages. So we’ve summarized it for B2B founders below!
1. AI User Adoption Is Literally Unprecedented
We know this, but still, the numbers do sort of blow your mind:
- ChatGPT: 0 to 800MM weekly users in 17 months (vs. Netflix’s 10+ years to 100MM)
- Time to 100MM users: ChatGPT (2 months), TikTok (9 months), Instagram (2.5 years)
- Global adoption: 90% of ChatGPT users are outside North America by Year 3 (vs. Internet’s 23 years to reach this level)
Why This Matters for B2B: Unlike previous tech waves that started in Silicon Valley and slowly diffused globally, AI hit the world simultaneously. This means your global TAM expanded overnight, but so did your competition. Every B2B and SaaS company now competes in a global, AI-enabled market from Day 1.
The Kicker: ChatGPT’s daily usage increased 202% over 21 months, with users spending more time per session (47% longer) and having more sessions per day (106% more). This isn’t just adoption – it’s addiction-level engagement.
2. The Infrastructure Math Is Unprecedented
The Capital Intensity Is Off The Charts:
- Big Six tech CapEx: $212B annually (63% YoY growth)
- Microsoft AI business: $13B run-rate (175% YoY growth)
- NVIDIA data center revenue: $39B quarterly (78% YoY growth)
- Amazon AWS CapEx as % of revenue: 49% (vs. 4% during initial cloud buildout)
What’s Really Happening: This isn’t just “cloud 2.0” – it’s the biggest infrastructure buildout in tech history. Companies are spending more on AI infrastructure than entire countries’ GDP. xAI built a 200,000 GPU data center in 122 days (faster than building a single house).
For B2B and SaaS Leaders: The infrastructure layer is being rebuilt from scratch. If you’re not thinking about how to leverage this massive compute capacity, you’re missing the biggest infrastructure opportunity since the cloud transition. The companies building on this new stack will have 10x advantages over those still running traditional architectures.
The Scary Part: Energy consumption is exploding. Data centers now consume 1.5% of global electricity, growing 12% annually (4x faster than total electricity consumption). This infrastructure boom has real physical limits.
3. China Is Playing a Different Game Entirely
The Models You’ve Barely Heard Of:
- DeepSeek R1: 93% performance of OpenAI’s o3-mini at fraction of training cost
- Alibaba Qwen 2.5-Max: Outperforms both DeepSeek and ChatGPT on key benchmarks
- Baidu Ernie 4.5: 80% cheaper than predecessor, costs 0.2% of GPT-4.5
Market Reality Check:
- China leads in open-source AI model releases (3 large-scale models in 2025 vs. US competition)
- Chinese AI apps dominate domestically: Top 10 AI apps by MAUs in China are all domestic
- DeepSeek rose from 0% to 21% global LLM user share in just months
- China has more industrial robots installed than rest of world combined
Geopolitical Stakes: This isn’t just about better chatbots. China views AI supremacy as essential to geopolitical leadership. As Andrew Bosworth (Meta CTO) noted: “This is our space race…there’s very few secrets. And you want to make sure that you’re never behind.”
For SaaS Companies: If you’re building AI-powered products, you now have formidable competition from China. Chinese models are achieving similar performance at dramatically lower costs. Your moat better be more than just “we use GPT-4.”
4. Token Costs Collapsed 99.7% in Two Years
The Cost Curve That Changes Everything:
- Inference costs fell 99.7% from Nov 2022 to Dec 2024
- NVIDIA GPUs: 105,000x less energy per token (2014 vs. 2024)
- What cost dollars now costs pennies; what cost pennies now costs fractions of cents
The Developer Revolution: This cost collapse democratized AI development. Suddenly, every indie developer and startup can afford to build AI-native products. The barrier to entry didn’t just lower – it disappeared entirely.
Business Model Implications:
- Training costs rising (now $100M+, heading to $10B+ per model)
- Serving costs plummeting
- Result: Commodity pricing pressure on model providers
- Opportunity: Massive new markets opened for AI-powered applications
For B2B and SaaS SaaS: Your customers can now afford AI features they couldn’t before. But this also means your competitors can too. The companies that move fastest to integrate AI capabilities will capture disproportionate value before the market commoditizes.
5. Enterprise AI Revenue Growth Defies Traditional SaaS Gravity
The Vertical SaaS AI Winners:
Cursor (AI Code Editor):
- $1M to $300M ARR in 25 months
- Used by millions of programmers
- Generates more code than almost any LLM globally
- Edits over 1 billion characters per day
Harvey (Legal AI):
- $10M to $70M ARR in 15 months
- 235 customers across 42 countries
- Majority of top 10 US law firms as customers
- 4x ARR growth in 2024
Abridge (Healthcare AI):
- $50M to $117M CARR in ~5 months
- 25,000+ doctors using it at Kaiser Permanente
- Over 10 million patient visit summaries completed
- Doctors saying “You’d have to take it away from my cold, dying hands”
Why This Is Different: These aren’t normal SaaS growth curves. Traditional vertical SaaS might hit $100M ARR in 5-7 years. These AI-native companies are doing it in 1-2 years. The productivity gains are so dramatic that customers have no choice but to adopt.
The Pattern: AI is solving workflow problems that humans literally cannot scale. It’s not just automation – it’s augmentation that makes impossible tasks possible.
6. Open Source Models Also Exploding
The Open Source Explosion:
- Meta Llama: 1.2B downloads in 10 weeks (3.4x growth in 8 months)
- Hugging Face: 1.16MM AI models (33x growth since March 2022)
- 100,000+ derivative models built off Meta Llama alone
Developer Preference Shift: Developers increasingly choose open-source models because they can:
- Customize for specific use cases
- Run locally without API dependencies
- Fine-tune on proprietary data
- Avoid vendor lock-in
- Control costs completely
Performance Convergence: The gap between open and closed models is closing rapidly. DeepSeek R1 scores 93% vs. OpenAI o3-mini’s 95% on math benchmarks. For many use cases, the performance difference is negligible.
Strategic Implications:
- Closed model providers losing pricing power
- Barrier to AI development continues falling
- Innovation moving to application layer, not model layer
- Platform lock-in strategies being disrupted
For B2B and SaaS: You can now build competitive AI features without paying premium API costs or being locked into specific providers. This levels the playing field but also intensifies competition.
7. Physical World AI Is Scaling Faster Than Software
Autonomous Vehicles at Scale:
- Tesla FSD: 4B+ fully self-driven miles (100x growth in 33 months)
- Waymo: 27% market share of SF rideshare bookings (from 0% in 20 months)
- Applied Intuition: Serving all top 18 global auto OEMs with AI-powered vehicle intelligence
Industrial Applications:
- Carbon Robotics: 230K+ acres weeded with AI-powered laser systems
- Tesla Dojo: 8.5x increase in AI training capacity for autonomous driving
- Industrial robots: China has more installed than rest of world combined
Defense and Critical Infrastructure:
- Anduril: 2x revenue growth for AI-enabled defense systems
- KoBold Metals: Using AI to reverse 40+ year decline in mining exploration efficiency
- Agriculture: AI systems preventing 100K+ gallons of herbicide use
Why This Matters: Physical world AI generates proprietary datasets that software-only AI cannot replicate. Every mile driven, acre farmed, or mission completed creates unique training data. These companies are building moats through real-world deployment.
The Acceleration Factor: Unlike software that can be copied instantly, physical AI requires real-world validation. But once proven, it scales with compounding advantages. Each deployment makes the system smarter for all deployments.
8. The Next 2.6B Internet Users Will Be AI-First
The Untapped Market:
- 2.6B people (32% of world population) still not online
- New users coming online via satellite internet (Starlink: 5M+ subscribers, 202% annual growth)
- These users will skip traditional internet experiences entirely
The Interface Revolution: New internet users won’t start with:
- Browsers and search bars
- App stores and downloads
- Typing queries in English
Instead, they’ll start with:
- Voice-first AI agents
- Native language interactions
- Agent-driven interfaces managing multiple platforms
Geographic Opportunity: ChatGPT usage by region shows massive adoption in:
- India: 14% of global users
- Indonesia: 6% of global users
- Pakistan, Mexico, Egypt, Brazil: 3-5% each
Strategic Implications: The companies that capture these AI-first users will own the next wave of internet growth. Traditional platform advantages (iOS App Store, Google Play, web browsers) become less relevant when users interact through AI agents.
For B2B and SaaS: Your next billion customers might never see your traditional interface. They’ll interact with your product through AI agents. Are you building for agent-mediated experiences?
9. AI Job Postings Are Up 448% While Traditional IT Jobs Fall
The Great Reshuffling:
- AI job postings: +448% over 7 years
- Non-AI IT job postings: -9% over same period
- AI-related job titles: +200% over 2 years (60,000+ new titles created)
- Apple alone: 600+ open generative AI positions
Corporate Mandates: Shopify CEO Tobias Lütke (internal memo): “Reflexive AI usage is now a baseline expectation at Shopify…I don’t think it’s feasible to opt out of learning the skill of applying AI in your craft.”
Duolingo CEO Luis von Ahn (all-hands): “AI use will be part of what we look for in hiring. AI use will be part of what we evaluate in performance reviews. Headcount will only be given if a team cannot automate more of their work.”
Productivity Evidence:
- Customer support agents using AI: +14% productivity improvement
- Kaiser Permanente doctors using AI scribes: Majority of 25,000+ doctors adopted
- GitHub Copilot: 77,000+ organizations using AI-powered coding
The Urgency: As Jensen Huang (NVIDIA CEO) warned: “You’re not going to lose your job to an AI, but you’re going to lose your job to somebody who uses AI.”
For B2B and SaaS Leaders: Your hiring, performance reviews, and team productivity all need to account for AI literacy. Companies that don’t upskill their workforce will lose talent to those that do.
10. The Monetization Models Are Still Completely Broken
The Venture-Scale Burn: OpenAI (estimated):
- 2024 Revenue: ~$3.7B
- 2024 Compute Expense: ~$5B
- Net: -$1.3B+ annual burn
Private Market Valuations:
- OpenAI: $300B valuation on $9.2B revenue (33x multiple)
- Anthropic: $61.5B valuation on $2B revenue (31x multiple)
- xAI: $80B valuation on <$1B revenue (>80x multiple)
- Combined: $95B+ raised against $11B annualized revenue
The Capital Intensity Problem:
- Model training costs: $100M+ today, heading to $10B+ per model
- Inference costs falling 99.7%, destroying pricing power
- Customer acquisition costs rising as competition intensifies
- Revenue concentration risk (enterprise customers demanding discounts)
Historical Precedent: This mirrors other transformative tech cycles:
- Amazon: Lost $3B over 27 quarters before becoming profitable
- Tesla: Burned $9.2B over 10 years before profitability
- Uber: Burned $17B over 7 years before positive cash flow
The Bull Case: These companies are building the foundational layer for the AI economy. Like Amazon Web Services, the winners could eventually capture enormous value once the infrastructure stabilizes.
The Bear Case: Training costs rising while inference costs fall creates a structural profitability problem. Competition from open-source models and Chinese providers could make it impossible to maintain premium pricing.
For B2B and SaaS: Don’t build your business model assuming AI API costs will stay high. Plan for continued cost deflation and commoditization. Focus on proprietary data and workflows where you can maintain pricing power.
The Meta-Learning: Speed Is Everything
Every one of these trends points to the same conclusion: This technology cycle is moving faster than anything in human history. The companies that adapt quickest will capture disproportionate value. Those that wait for “certainty” will find themselves competing against AI-native challengers with 10x productivity advantages.
The question isn’t whether AI will transform your industry – it’s whether you’ll lead that transformation or be disrupted by it.
And our SaaStr + AI Summit 2025 opener touched on many of these times as well here:
Watch SaaS + AI: It’s Time to Move Much Faster with SaaStr CEO and Founder Jason Lemkin
