Just spent the last 3 days at SF Tech Week hitting every VC event I could find. The fundraising landscape in late 2025 is… complicated. Let me share what I learned from the trenches. 
Events attended:
- Andreessen Horowitz “State of AI Funding” breakfast
- YC Alumni “Surviving the AI Winter” panel
- Sequoia “What We’re Funding Now” fireside chat
- SPEEDRUN Demo Day (100+ AI startups pitching)
The Tale of Two Markets
The headline numbers look INSANE:
From Bloomberg report (Oct 2025):
- $192.7B invested in AI startups in 2025 YTD
- 70% of ALL VC dollars going to AI companies
- Q3 2025: Global VC funding up 38% YoY
- 498 AI unicorns valued at $2.7T combined
My reaction: “Great! Easy to raise, right?”
Reality check from a16z partner: “That money is going to 20 companies. Everyone else is fighting for scraps.”
The Concentration Problem
Data shared at Sequoia session:
Top 20 AI companies raised: $132B (69% of total)
Remaining 10,000+ AI startups raised: $60B (31% of total)
Mega-rounds dominate:
- 69% of AI funding goes to rounds $100M+
- These are growth-stage companies (Series C+)
- OpenAI alone raised $30B+ in 2025
For seed/Series A founders like me:
- Seed: $9B across 3,500 companies = $2.6M average
- Series A (AI): $16M median (vs $7M for non-AI)
- Series B+: Getting much harder unless you have monster traction
Translation: If you’re not OpenAI, Anthropic, or Perplexity, you’re competing with 10,000 other AI startups for 1/3 of the money.
What I Learned at SPEEDRUN Demo Day
100 AI startups pitched. Here’s what I observed:
Category breakdown:
- AI agents/automation: 35%
- Developer tools: 25%
- Enterprise AI: 20%
- Healthcare AI: 10%
- Everything else: 10%
The pattern: 80% of pitches sounded IDENTICAL.
Typical pitch:
“We use AI to [common task]. We’re different because we have [minor technical detail]. We have $50K MRR and growing 15% month over month.”
VC feedback (overheard at after-party):
“I saw 30 companies doing essentially the same thing. Why would I bet on any of them when I can wait and see who wins?”
The brutal truth: Differentiation is REALLY hard when everyone has access to the same foundational models.
The “AI Winter” Warnings
The term “AI winter” came up in EVERY panel I attended.
Apollo Global Management economist (quoted everywhere):
“The current AI bubble is stronger than the dot-com boom. We’re looking at order-of-magnitude overvaluation.”
Sam Altman himself said (widely reported):
“The AI market is in a bubble. Investors are overexcited.”
Data that scared me from Yale professor’s analysis:
OpenAI case study:
- Valuation: $340B (latest round)
- 2025 revenue: $11B
- Profitable by: 2029 (projected)
- Cumulative losses 2023-2028: $44B
That’s a 31x revenue multiple for an unprofitable company.
Comparison:
- Google at IPO: 10x revenue (profitable)
- Amazon at peak bubble: 15x revenue
- OpenAI today: 31x revenue (losing money)
Bain & Co. prediction (presented at a16z):
- By 2030, AI companies need $2T annual revenue
- Projected actual revenue: $1.2T
- Gap: $800B shortfall
Question that keeps me up at night: When does this unwind?
The Funding Bar Has RISEN Dramatically
From YC Alumni panel:
YC acceptance rate: 1% (10,000 applications, 100 accepted per batch)
What used to work (2023):
- Idea + founding team
- Maybe a prototype
- Raise $1-2M seed
What you need now (2025):
- Live product
- Paying customers
- $50K+ MRR
- Clear differentiation
- Path to $100M ARR
- Raise $2-4M seed
YC partner quote:
“In 2023, we funded ideas. In 2025, we fund traction. The bar has moved.”
Series A is even crazier:
2023: $5-10M ARR gets you a Series A
2025: $10-20M ARR gets you MAYBE a Series A
Sequoia partner:
“We’re seeing companies with $15M ARR struggling to raise Series A because investors are worried about market saturation.”
The Valuation Trap
Cautionary tale from a founder at after-party:
2024: Raised seed at $30M valuation (AI hype peak)
2025: Trying to raise Series A
Problem: Need to price at $100M+ to avoid down round
Reality: Investors offering $60M (based on actual metrics)
Result: Can’t raise, burning cash, might die
The trap:
- Raise at inflated valuation during hype
- Spend 18 months building
- Market corrects
- Can’t raise at reasonable valuation (down round scares investors)
- Run out of money
Advice from multiple VCs:
“Raise at reasonable valuations. Survive to the next round.”
What VCs Are Actually Funding
From 3 days of conversations, here’s what’s working:
1. Applied AI in boring industries
- Healthcare workflows
- Legal document processing
- Manufacturing optimization
- Logistics
Quote from investor:
“Everyone wants to build the next ChatGPT. We want companies applying ChatGPT to problems that make money.”
2. AI infrastructure/tools
- Model evaluation platforms
- AI observability
- Cost optimization
- Security for AI systems
Why: These have immediate, measurable ROI
3. Vertical AI agents
- Sales automation
- Customer support
- Recruiting
- Specific industry workflows
Must have: Clear ROI story, not just “it’s cool”
4. Companies with DEFENSE
- Proprietary data
- Unique distribution
- Network effects
- Switching costs
Not: “We fine-tuned GPT-4” (that’s not defensible)
Revenue Expectations Are Insane
Data shared at multiple events:
To raise in 2025, you need:
Seed ($2-4M):
- $30-50K MRR
- 10-20% month over month growth
- 3-6 month runway from previous funding
Series A ($10-20M):
- $1-2M ARR
- Net retention 120%+
- Path to $10M ARR in 18 months
- Unit economics that work
Series B ($30-50M):
- $10M ARR
- Growing 3x year over year
- Clear path to $100M ARR
- Proven go-to-market
These numbers were unthinkable 2 years ago.
YC data point:
“25% of our current batch has 95% of code written by AI. These founders don’t need 50 engineers.”
Implication: You can get to revenue faster with less capital, so investors EXPECT more traction earlier.
The Profitability Pressure
MAJOR shift I’m seeing:
2023: Growth at all costs
2025: Efficient growth
Metrics investors kept asking about:
1. CAC payback period
- Want: <12 months
- Acceptable: 18 months
- Pass: 24+ months
2. Burn multiple
- Excellent: <1.5x (burn $1.50 for every $1 revenue)
- Good: 2x
- Pass: 3x+
3. Gross margins
- SaaS: Need 70%+
- AI companies: 50% acceptable (model costs)
- Below 50%: Explain why
Quote from Sequoia:
“We learned from 2021. We’re not funding growth with terrible unit economics anymore.”
My Fundraising Experience (Real Talk)
My company:
- AI-powered developer tools
- $80K MRR
- Growing 12% month over month
- Team of 4
- Burning $60K/month
- 6 months runway
Goal: Raise $3M seed
Results after SF Tech Week:
- Talked to 25 VCs
- 3 interested in follow-up
- 0 term sheets
- Multiple “come back when you’re at $200K MRR”
The feedback:
- “Developer tools is crowded” (heard this 10 times)
- “Your growth rate is good but not exceptional”
- “Come back in 6 months”
- “We only do $5M+ rounds now” (this hurt)
- “How are you defensible vs GitHub Copilot?”
Reality check: I thought $80K MRR would be enough. It’s not.
The Alternative Funding Paths
What founders are doing when VC funding is hard:
1. Revenue-based financing
- Pipe, Capchase, Lighter Capital
- Get capital based on MRR
- Pros: Fast, no dilution
- Cons: Expensive (18-30% effective APR)
2. Angel syndicates
- Smaller rounds ($500K-1M)
- More friendly terms
- Pros: Easier to close
- Cons: Time-consuming (need 20-30 angels)
3. Strategic investors
- Corporate venture arms
- Industry players
- Pros: Smart money, distribution
- Cons: Slower, potential conflicts
4. Staying lean and profitable
- Skip fundraising entirely
- Bootstrap to profitability
- Pros: Control, no dilution
- Cons: Slower growth
Trend I’m seeing: More founders choosing option 4
The Bust Predictions
Timeline estimates from panels:
Optimistic (a16z):
“AI is real technology. Market will mature, not crash. Timeline: Gradual correction over 2-3 years.”
Pessimistic (hedge fund manager):
“When OpenAI or Anthropic misses revenue projections badly, dominoes fall. Timeline: 12-18 months.”
Realistic (YC partner):
“Some overvalued companies will fail. Real businesses will survive. The market will bifurcate.”
Comparison to dot-com:
- Dot-com: Infrastructure wasn’t ready, revenue wasn’t there
- AI now: Infrastructure is ready, but revenue isn’t scaling as fast as valuations
Key difference: AI is real technology (unlike pets.com)
But: 90% of AI startups will fail (like dot-com)
Advice I’m Taking to Heart
From founders who survived previous downturns:
1. Raise more than you think you need
- Markets can stay shut for 18-24 months
- Want 2 years runway minimum
2. Raise at sensible valuations
- Down rounds kill momentum
- Take the lower valuation if it means surviving
3. Focus on revenue, not hype
- Revenue compounds, hype fades
- Investors forget promises, remember numbers
4. Build something defensible
- AI model improvements are commoditized
- Need data, distribution, or network effects
5. Watch your burn
- Default to profitability mindset
- Growing 50% faster while burning 3x more isn’t worth it
6. Don’t assume the market stays hot
- Have Plan B if funding dries up
- Can you get to profitability on current runway?
The Uncomfortable Truth
My biggest takeaway from SF Tech Week:
The AI boom is real.
The AI funding bubble is also real.
Both things can be true.
Good AI companies will get funded.
Mediocre AI companies will die.
The difference is traction, differentiation, and timing.
For founders:
- If you can raise now (and get reasonable terms), do it
- If you can’t raise, focus on revenue
- If you have 18 months runway, use it to get profitable
For investors:
- The best deals are getting done quietly
- FOMO investing leads to pain
- Fundamentals matter again
Bottom line: We’re in the “trough of disillusionment” phase of AI. The survivors will build real businesses. The rest will be case studies in “what not to do.”
Anyone else fundraising right now? What are you seeing?
David 
SF Tech Week - Andreessen Horowitz, YC, Sequoia events
Sources:
- Bloomberg “AI Funding 2025” (Oct 2025)
- Bain & Co “AI Revenue Gap Analysis” (2025)
- Yale SOM “AI Bubble Analysis” (Sept 2025)
- Crunchbase Global VC Report Q3 2025
