The price of data has 5x’ed

👇🏼 Join me at the annual DECODE conference on Sat 11/14 in Berkeley!

Join the conference, and let’s catch up in-person!

• DECODE is the largest founder community co-hosted across UC Berkeley and Stanford. I’ve had the honor of serving on the DECODE board since 2016.

• The DECODE annual conference brings together experienced founders, tech executives, and venture investors to help early-stage founders. I just posted about the upcoming conference this year here.

• What makes this gathering unique? Speakers don’t just speak on a big stage and disappear. They stick around for off-the-record roundtable discussions with a dozen of handpicked founders, giving attendees a chance to not just listen but to actually meet them.

• You can use SHUO_VIP for 30% off tickets!

From past conferences, founders have …

• Successfully raised pre-seed, seed and series A rounds from top-tier investors (including a16z, Sequoia and Y Combinator), met through DECODE roundtables and dinners.

• ​Identified advisors and mentors (including experts from Apple, Google and Microsoft) who’ve helped them 10x their startup’s growth.

• ​Found co-founders and early hires for their startups through DECODE community members, like the founding teams of Databricks ($62bn valuation), Agora (IPO) and FlowGPT (Series A).

For a complete list of speakers …

3 trends in startups/tech/venture

📈 The price of data has 5x’ed

• Mercor, a ~$20bn valuation startup that sells data to train AI models, is famous partly for offering $100k to anyone who refers a company that ends up licensing its enterprise data to Mercor to use to train AI models.

• As I shared at the end of August with the story of Google spending $10M to acquire Spirit Airlines’ data, data is a scarce and valuable enough resource in the AI race that companies are willing to pay top dollar for.

• And, as I’ve learned from recent conversations with founders, the average licensing agreement for AI training data has jumped 5x since the beginning of this year, from $250k–$400k to ~$1M–$1.5M for 2-year access (for datasets much smaller than that of Spirit Airlines, of course).

• The takeaway? It’s been 20 years since the phrase “data is the new oil” was coined by British mathematician Clive Humby — and it’s as true as ever today.

🤖 Robots can do ~80% of physical tasks, but are cheaper than humans for only 0.3% of them

• The latest data from Anthropic shows that robots can already perform 74% of physical tasks in the U.S., with the jobs at the highest risk of automation being those that are held by male, less-educated, and lower-paid workers such as those in driving and warehousing roles.

• Just because robots can perform a task doesn’t mean it makes economic sense, though. The same study showed that robots are cheaper than humans for a mere 0.3% of job tasks. At historical price decline rates, it would take 40 years for robots to be cost-competitive with even 10% of job tasks — and this forecast doesn’t even take into account potential regulatory barriers that could slow things down.

• The takeaway? Sure, we may see robots replacing drivers and warehouse packers in the short-run, but, over the long-run, it’s still anybody’s guess as to whether and how much the labor market for physical tasks will change.

💰 Emerging funds back 2x more outperforming startups than megafunds

• Per the latest data, the average startup VC (called an “emerging manager”) has 2x as many winning startups in its portfolio as the average “megafund” (think a16z and Sequoia).

• This phenomenon makes sense. After all, VCs try to do one of two things: discover new innovations or dominate existing categories. And while VCs may claim to do both, the revealed preference of megafunds is to dominate, leaving small and emerging managers to discover (and then for the megafunds to swoop in and do “mega-deals” for the ones that show the most initial traction in hopes that they will go on to dominate).

• The takeaway? Limited Partners can achieve far greater diversification by investing in 5 emerging managers (which are more likely to be invested in completely different companies) than 5 megafunds (which are likely to hold the same homogeneous portfolios).

2 theses on what’s next

🤖 The winning AI model = the one that remembers the most about you + has access to the most compute

• Imagine asking an LLM for dinner date ideas. One gives you a generic list. Another gives you a list — but remembers that you prefer options with free parking and your significant other likes tiramisu. The model that we will all go to will likely be the one that knows and remembers the most about us, so it can tailor its responses to our unique circumstances.

• But, personalization takes “compute” (processing power), which is increasingly difficult to come by. As an example, Amazon Web Services’ cloud computing capacity is largely sold out through 2027, ​with significant commitments extending into 2028.

• The result? AI startups are no longer merely in the business of building the best models. They also need to be the most cost-efficient at securing compute.

🤔 The best startup founders follow the Ten Year Rule

• The Ten Year Rule finds that it takes on average 10 years to get meaningful traction in any career, which is true for founders of startups.

• Similar to the Gartner hype cycle (which maps out people’s expectations on the Y-axis over a X-axis of time), you can plot out a founder’s expectations on the Y-axis over a X-axis of time — and see a visual depiction of the emotional rollercoaster founders experience over the course of building their startup:

  • 1. Startup Trigger (where founders are optimistic + uninformed): Every founder starts here, often unaware of the pain ahead (even Jensen Huang noted he wouldn't start Nvidia today if he knew how tough it’d be).

  • 2. Peak of Inflated Expecations (where founders are still optimistic + uninformed): Many founders raise early funding and receive external validation (e.g. win a startup competition or get into an accelerator) early on, but haven’t started to experience hardships yet.

  • 3. Trough of Disillusionment (where founders are pessimistic + uninformed): The stark reality hits, causing the largest percentage of founders to quit in this phase as they learn why starting and scaling is so difficult.

  • 4. Slope of Enlightenment (where founders are pessimistic + informed): Founders recognize their gaps and realistically assess whether the grueling path is worth their time. The second largest percentage of founders quit in this phase.

  • 5. Plateau of Productivity (where founders are optimistic + informed): Survivors master scaling and predictable business growth.

• It takes on average 10 years for founders to hit the fifth stage. Any founder that hopes to succeed should plan to commit at least 10 years to tackling a given problem.

• The biggest mistake I see founders making? Jumping from industry to industry and problem space to problem space. While many call it “pivoting,” I don’t — at least not in cases where founders are jumping from AR/VR to crypto to AI (and each time chasing a tool rather than investigating a specific problem). What they are doing is stopping short of becoming experts in any given area and re-starting their 10-year clock each time.

• My top recommendation to founders is to “pivot” in the form of running experiments quickly and cheaply, but otherwise committing to a problem area for the long-haul.

1 tool I love

💻 Personalized software

• Most software asks you to adapt yourself to tools designed for someone else. 

• Imbue Studio — by Imbue, a portfolio company — turns the tables by allowing you to:

  • Right-click on any tool and tell Studio what you want to be different (e.g. add a feature, change the interface, build an automation). Studio can even modify its own source code.

  • Turn your calendar, email, Slack, files, etc into a unified inbox, a personal CRM, a custom workflow, or something entirely your own.

  • Share a tool as easily as sharing a Google Doc, or share an entire Studio and build and edit software together in real time.

  • Make your tool accessible on desktop or on mobile (or even make Studio your phone home screen).

• You can learn more here.

More of my work 👇🏼

🎤 Decode videos | For top Berkeley and Stanford founders

• DECODE is the largest founder community co-hosted across UC Berkeley and Stanford. The DECODE annual conference focuses on helping founders in their earliest stages of starting a startup.

• I’ve had the honor of serving on the DECODE board since 2016. You can find videos from the latest conference I helped host here 👇🏼

🤖 Founders' prompt library

• My team and I have crowdsourced the largest AI prompt library optimized for founders – especially fractional founders — to help them build, sell and operate 10x faster and better.

👥 Pop up board

• I regularly host pop-up boards, a unique 45-minute session where founders ask their toughest strategic questions and get tailored advice from top builders and operators from Google, Microsoft, Meta and more.