Good morning 👋
Mark Meeker is a venture legend, and her annual reports are must-reads.
Her latest on AI trends came out Friday night, and I’ve spent several hours going through it trying to use it to think more clearly about the future.
I’ve shared some thoughts on what stood out to me below …
TOP
Takeaways from Bond’s latest AI report
If you have not read the latest Bond report, I’d recommend you do that.
If you’re short on time, here are some high-level takeaways I had after reading through the report:
AI infra is the new oil—and we’re in a CapEx supercycle.
The AI buildout is mimicking the early internet and cloud waves—but on steroids.
CapEx across hyperscalers (Big 6) is up 63% in 5 years, projected to top $212B this year, largely on AI infra (H100s, datacenters, energy sourcing) .
Microsoft and Amazon are entering utility-like territory and locking in long-term pricing power through control of compute. This is a distribution advantage, not just a tech one.
LLMs are shifting enterprise software from SaaS to agentic workflows.
Traditional SaaS was built for CRUD interfaces; the next wave will be “agent-initiated” software that executes, not just displays.
Over 65% of AI enterprise use cases are already moving beyond productivity toward revenue generation, with CMOs leading adoption .
Implication: software businesses not adapting to AI-native distribution and UX will see churn accelerate (think: Notion vs. ChatGPT-embedded workflows).
Most frontier models are not commercially viable (yet).
Training cost CAGR = 360%, inference cost CAGR = -55%.
The best open models will be “good enough” for 80%+ of tasks, raising existential questions for closed model providers who don’t control distribution or vertical integration.
Look for a bifurcation: one group monetizes via API volume at scale (OpenAI, Anthropic); the other leverages models as enablers (Palantir-style GTM, vertical SaaS wrappers).
Geopolitical AI control is not about language models. It’s about compute, energy, and labor arbitrage.
China now leads the world in robot installations (52% of global total).
The U.S. leads in LLM development, but physical AI (hardware, robotics, energy-hungry training clusters) is tilting toward Asia .
A compute-rich, labor-light economy may invert the traditional GDP formula - VCs and LPs should adjust mental models on capital efficiency.
AI’s edge in science isn’t hype - it’s a time-arbitrage weapon.
Drug discovery timelines cut from 5 years to 18 months in first AI-native compounds.
Atomwise used AI to screen 3 trillion molecules to identify 100+ novel therapeutics.
The wedge? New data flywheels (bio, chemistry, climate, materials) that were previously too sparse or noisy for traditional ML.
Capital markets haven’t priced the cost of energy in the AI wave.
One ChatGPT search = 3x the energy of a Google search.
Datacenter energy demand is expected to double by 2026.
Smart capital is watching the intersection of energy infrastructure and compute provisioning (read: Berkshire quietly becoming a compute landlord).
Together with PortfolioIQ
Automate your fund ops
PortfolioIQ organizes your portfolio data, automatically.
Their platform helps you:
Set up automated data requests ✅
Extract data from documents ✅
Standardize data for metric name & currency variations ✅
Track and source-map every data point ✅
Stop the quarterly reporting scramble.
Hyper-automate your ops with PortfolioIQ.
TWEET
LINKS
🦄 When Do Unicorn CEOs Exit: Ilya Strebulaev, professor at Stanford, analyzed over 1500 US-based unicorns and found 1/3 leave before their companies reach unicorn status, and 2/3s are gone by the latest funding round
📽️ Mastering Video Communication for Founders: This episode of features Alfred Poor, a communication expert who discusses how founders and visionary leaders can show up with confidence and clarity in every virtual interaction—-from investor pitches to remote team meetings
✅ Storm Ventures’ Decision-Making Process: Storm’s AI Product Executive lists the 3 step process it goes through before a decision to invest is made
🤨 “Does My New Startup Idea Need to have AI in it?”: Andrew Chen addresses this question that lots of people are asking
💁♂️ What’s ARR, Really?: Here are some thoughts on what can be a confusing term
HEADLINES
GitHub co-founder raising $100M fund to back early-stage tech startups (Pitchbook)
The US direct VC secondary market is swelling (Pitchbook)
Early AI investor Elad Gil finds his next big bet: AI-powered rollups (TechCrunch)
VC Funds Are for Sale—at a Discount (The Information)
UPGRADE
"This is like getting an analyst for 1/100 of the price ..."
That’s how one reader described his experience when he upgraded his experience with this newsletter.
Where else can you get all of this without hiring at least a part-time person?
Investment memos on companies backed by Sequoia, Accel, Founders Fund, and others ✅
Market maps on up-and-coming verticals (with company data) ✅
Behind-the-scenes looks at how we build and scale this business ✅
Database of 2,000+ venture capital firms ✅
Ultimate VC resource library (400+ curated resources used by the top funds in the world) ✅
Community builder playbook (everything we have learned from running a community-first business) ✅
$50k+ in software discounts (from vendors like beehiiv, Airtable, Loom, Harmonic, and others) ✅
If all of that sounds interesting, see why others are upgrading their experience by checking out our different plans below …
Thanks for reading this far and giving us a little bit of your attention this week.
Feel free to unsubscribe whenever this stops becoming valuable to you.
- Clay
(Founder @ Confluence.VC | GP @ Outlaw)
