AI Isn’t a Trend

BOND and Mary Meeker just dropped their first AI Trends Report, and it’s a monster—over 300 pages deep. But let’s be clear: AI isn’t just the next wave. It is the platform shift.

This isn’t your typical disruption cycle. It’s the internet + cloud + mobile—all at once. Here’s what actually matters:

The Shift Is Real (and Already Happening)

• AI IT jobs: +448% from 2018 to 2025
• Non-AI IT jobs: –9%
• Median time to $5M ARR (AI startups): 24 months (Compare that to SaaS in 2018: 37 months)‍
• Inference cost per token: Down ~99.7% in just two years

AI-native companies are moving faster, scaling bigger, and attracting massive capital.

Infrastructure Is Booming — Just Not Where You Think

• AWS CapEx as % of revenue: 49% in 2024
• Spending focus: Chips, cooling, hyperscale compute — not office space
• Training costs: Up ~2,400x since 2016 (Frontier models now cost $100M–$1B+)‍
• Inference: Dirt cheap, massively scalable

We’re in the second wave of cloud. But this time, it’s built for AI.

AI Dev Ecosystem Is Exploding

• NVIDIA devs: 2.4x growth → 6M by 2025
• AI startups: 3.9x growth → 27K
• GitHub AI repos: +175% in 16 months
• Open-source AI:
  • 100K+ derivatives of Llama models
  • Llama 3, DeepSeek, etc. rivaling closed models on reasoning, coding, multilingual tasks

Open source is eroding proprietary moats faster than anyone expected.

Global AI Scale — Unprecedented

• ChatGPT: 800M users in 17 months (fastest adoption curve ever)
• LLM adoption in US: Projected to hit 90% in 3 years
  • (Internet took 23 years to get there)
• Token scale:
  • Google: 480T tokens/month (50x YoY)
  • Azure: 100T tokens/quarter (5x YoY)
  • Gemini devs: 7M (5x YoY)

We’re not building apps anymore—we’re building at token scale.

It’s Not Automation — It’s Reinvention

AI isn’t about efficiency. It’s about reinventing how we work:

• How we build
• How we think
• How we compete

Legacy playbooks are done. Hiring, pricing, roadmaps—all being rewritten in real time.

Who’s Winning?

The AI-first companies. Not the AI-curious. They:

• Ship faster
• Make better decisions
• Automate the boring stuff
• Invent what’s never been built

Examples:

• Anysphere Cursor: $1M → $300M ARR in just 25 months
• Harvey, Decagon, Abridge: Scaling like it’s 1999

‍

What You Should Be Doing (Yesterday)

You don’t need a billion-dollar GPU cluster. You need urgency—and courage.

Use AI to:

• Accelerate R&D
• Rewrite workflows
• Personalize everything
• Build 10x faster
• Actually scale your content

Because the companies waiting on the sidelines? They’re just slowly failing.

‍

Read the full BOND AI Trends Report here

‍