20VC: Thrive & OpenAI Partnership | Eventbrite Acquired for $500M | Databricks Raising $5BN at $134BN Valuation: Cheap or Not? | Why SaaS is Like Japan and The TAM Trap in Software

Analysis of OpenAI's partnership with Thrive and the subsequent 'code red' announcement focusing on core products. The discussion reveals how VCs go deeper with their winners and why OpenAI is pulling back from peripheral initiatives to compete with Google's resurgence.

• •OpenAI announced a code red to focus on core products, pushing back on agents and ads to compete with Google
• •The Thrive partnership demonstrates the power law of VC investing - going as deep as possible with your biggest winners
• •VCs should turn their best investments into holding companies and double down on winners
• •The partnership is more valuable for Thrive's brand halo than for OpenAI's strategic advantage

Deep analysis of Databricks raising $5B at $134B valuation (32x revenue) while growing 55% YoY. Comparison with Snowflake's 28% growth at 20x revenue reveals the premium paid for reacceleration at scale and why this might be justified.

• •Databricks at 55% growth vs Snowflake at 28% growth - the key question is how much extra multiple you pay for extra growth
• •Reacceleration at $4B scale is extremely rare and potentially infinitely valuable if it continues
• •Only one public company (Palantir) is growing over 30%, making it hard to benchmark these valuations
• •Late-stage deals in proven companies may offer better risk-adjusted returns than early-stage bets
• •The TAM question becomes critical - growth only stops when you've sold to everyone

Examination of struggling SaaS companies - PagerDuty at 2x revenue with 4% growth, Eventbrite acquired for 1.5x revenue. The discussion introduces the concept of the 'TAM Trap' where even great founders couldn't expand beyond their initial markets.

• •PagerDuty valued at only $1B despite 500M ARR - a sobering reality for finite TAM businesses
• •Eventbrite acquired for $500M at 1.5x revenue with 50% premium shows desperation in low-growth SaaS
• •The 'TAM Trap' - most public SaaS companies couldn't figure out how to expand beyond their core market
• •Overpayment only works when TAM is huge; in finite TAMs you need pricing discipline
• •Smart money (PE firms) sees opportunity to add AI capabilities to these assets and reignite growth

Discussion on why companies need to think about second products much earlier and the value of solving hard technical problems. Uses Zoom as a cautionary tale of a brilliant founder who couldn't find a second act.

• •Companies must start building second products before hitting the TAM trap, not after
• •Zoom's failure to expand despite Eric Yuan being one of the best technical founders shows how hard this is
• •Database companies have inherent defensibility - you can't easily churn from your database
• •Hard technical problems provide more defensibility than easily cloneable products
• •Successful multi-product companies layer in new products early that compound over time

Analysis of why seat-based pricing is under existential threat as companies become radically more efficient with AI. Discussion of how ARR per employee is skyrocketing across all tech companies.

• •Workday correctly identifies seat reductions as an existential threat to their business model
• •HubSpot is 2.8x more efficient than 2021, Salesforce 2x - all companies getting more ARR per employee
• •Microsoft has declared they're past peak employee permanently
• •Usage-based pricing (like Twilio) is more insulated than seat-based models
• •The expectation is now 100% revenue growth with only 50% headcount growth
• •SaaS has become 'like Japan' - a great economy but with shrinking population (seats)

Critical discussion of how security breaches at Gainsight and Drift led to permanent platform bans from Salesforce, and what this means for the future of platform ecosystems and AI security.

• •Gainsight has been locked out of Salesforce for two weeks with no resolution - app is completely down
• •Drift was permanently banned after 700 customers' data was downloaded and held for ransom
• •Security will benefit incumbents as enterprises become more conservative about data access
• •AI agents accessing sensitive data across multiple systems increases security risk exponentially
• •Most startups have inadequate SecOps teams (5-10 people) for the new threat landscape
• •Incumbents may use security as an excuse to lock out competitors while launching their own AI products

Exploration of how AI is creating unprecedented capital efficiency at the apps layer, while simultaneously creating an employment crisis. Discussion of why all three categories of companies (mature, model providers, AI apps) don't need people.

• •AI app companies like Gamma are growing so fast they can't hire quickly enough - traction ahead of hiring ability
• •Model companies need geniuses and GPUs, not large headcount
• •Mature companies must optimize for efficiency, leading to lower employment
• •The fastest growing AI companies have the lowest burn multiples despite high inference costs
• •You literally cannot brute force to $100M ARR in 10 months with humans - you need AI and massive inbound demand
• •This creates a tough employment market where labor loses to capital in the near term

Analysis of Google launching a Replit/Lovable competitor and what it means for startup defensibility. Discussion of which markets model providers will enter versus which are safe from competition.

• •Google launched a coding tool competitor in less than 10 months - you don't get years anymore
• •The product launched without database or OAuth (coming soon) showing big company prioritization challenges
• •Datadog launched a PagerDuty competitor 16 years after PagerDuty was founded - competition is accelerating
• •Model providers will compete in obvious adjacent spaces like coding but not in specialized verticals
• •OpenAI's code red suggests they won't expand into all verticals - focus on winning the core chat wars
• •Vertical-specific applications (wealth management, compliance) are safer from model provider competition

Debate on whether slow-compounding businesses like Wealthfront are good investments in the age of rapid AI growth. Discussion of how wealth management could be disrupted by AI and whether VCs should prioritize momentum over compounding.

• •Wealthfront took 10-15 years to build but is now a compounding machine that will last 30+ years
• •Charles Schwab went public in 1982 and is worth $60-80B today - few tech companies from that era survived
• •AI could disrupt wealth management by automating estate planning, taxes, and investment advice for mass market
• •The average American retiring has $1.8M in assets - huge TAM for AI-enabled wealth management
• •VCs are playing a 'relevance game' prioritizing quick markups over long-term compounding
• •When you have high certainty a big company can be built, prioritize that over hype and FOMO

Final debate on whether to invest in Supabase ($5B) or Lovable ($6B), revealing different investment philosophies around defensibility, TAM, and the value of solving hard technical problems versus riding momentum.

• •Lovable is the better bet if vibe coding becomes a category - it captures more value as the front-end
• •Supabase represents a harder technical problem (databases) with more inherent defensibility
• •Hard problems are reassuring because they're difficult to clone and customers can't easily churn
• •If vibe coding isn't a category, both investments fail - so bet on the one with more upside
• •The year of tolerating churn is over - investors want defensibility and hard problems
• •40% of Q1 unicorns already had follow-on rounds - unprecedented velocity of validation in AI