20VC Newsletter - May 26th 2024
Here are the transcripts and top takeaways from 20VC episodes released this last week!
Monday’s episode with Nikesh Arora, CEO of Palo Alto Networks:
Download the full transcript:
My 6 key takeaways:
What Palo Alto Networks Looks for in Acquisitions
We buy innovation & we buy early.
I don’t believe in buying later & paying multiples for revenue.
We buy great teams & great products with proof of concept.
Why Competition is Inevitable
When you find a market with no competition, you’re either a genius or totally stupid.
Genius because you’re early & competition will follow.
Stupid because there’s absolutely NO business & you’re burning cash.
How to Sustain a Competitive Advantage
Every enterprise SaaS business has a 2-3 year competitive advantage.
You need to build a moat in that time frame.
The top $BN companies are always a few years ahead of their competition.
How Great Leaders Communicate
They make sure everyone understands the principle.
You can’t let people do their work not knowing how they contribute to the overall whole.
They should understand how their work adds value.
How a Cold Letter and Luck Started this Billionaire's Career
I wrote over 400 cold letters to the Northeastern MBA program alumni.
One of them sent my CV to their direct reports at Fidelity.
I got an interview & landed the job.
What Makes Masa So Special?
His risk appetite hasn’t changed through age.
He has extremely high marks for conviction.
One bad deal wouldn’t deter him from going all-in on the next.
Wednesday’s episode with Box CEO & Co-Founder Aaron Levie:
Download the full transcript:
My 6 key takeaways:
What You Need to Know in This New AI Wave
We are in a window of opportunities for AI.
It will be as much an incumbents' game as a startups' game
If you’re in one of these startups, you shouldn’t be focusing on anything but survival & execution.
Agents: What is the 10 Year Vision
We will have category winners in job functions that exist today.
We will have AI security engineers, AI customer support agents, AI marketers...
We will need management systems for AI labor.
How AI Will Change the Business Model of SaaS
Everyone is experimenting on different versions right now.
Pricing could be on consumption, value delivered, customer support tickets…
Eventually we’ll figure out a unit that we can all agree with.
What Founders Should Build in the World of OpenAI
They need to avoid businesses that could be subsumed by a horizontal chat interface.
They should be focusing on workflows, aka the boring business
This should be used by a human who wants to run a full business process.
Foundational Model & Application Layer: Can Startups Win Both?
Pure play LLMs will largely be subsumed by the big players.
There may be room for 1 or 2 independent companies but there won’t be room for 50.
Startups could focus on specific domains that incumbents are too afraid to go into.
Why I'm Bullish on Apple
Apple isn’t under any immediate threat of AI.
They could turn their hardware into a command center for AI.
Friday episode with Sam Altman, Arthur Mensch and more.
Download the full transcript:
My 9 key takeaways:
Sam Altman on the Future of Foundation Models
There will be a small number of providers doing large & complex models.
The long term differentiation will be the model that’s most personalized to you.
Brad Lightcap: Will OpenAI Steamroll Your Startup?
“Ask the company whether a 100x improvement in the model is something they’re excited about.”
Most companies have a clear story if better models accelerates their product.
Des Traynor on the Value in the Application Layer.
Founders need to find an area where OpenAI won’t go deep.
OpenAI will never put five engineers going hard on wealth management.
Find that area & go all-in.
Arthur Mensch on Two Directions for AI
The application layer will grow thin
The model part will grow thin.
“The model part is going to be big, we need to build the platform on top of that & enable all of the vertical applications that will be interesting for humanity.”
Tom Hulme on Why Foundation Models Are Like Power Stations
The training time is the station & the inference is the power.
Everyone is building similar power stations with little edge.
“...You basically gotta depreciate that asset in these models over a few months. I just can't see it happening.”
Sarah Tavel on Why the Value Will Accrue in the Application Layer
“Each successive model will be more and more expensive to train, that suggests a world where you're going to have an oligopoly.”
The real value will come from the application layer.
Whoever owns the user will provide more value.
Emad Mostaque on Who Will Be the Winners in AI
“There's only gonna be 5 or 6 foundation model companies in 5 years. I think it's gonna be us (Stability AI), Nvidia, Google, Microsoft, OpenAI, Meta and Apple.”
Google spends $20BN/Year on AI, DeepMind’s salary budget is $1.2BN/Year.
They have $150BN to win this.
Miles Grimshaw on Why Copilot is an Incumbent’s Strategy
“Incumbents own distribution, they own data, the UX, and a business model that all aligns to a copilot.”
The opportunity for a startup copilot probably won't be that amazing.
The real opportunity to disrupt is to be orthogonal to the incumbents.
Tom Blomfield on the Sustaining Value in the Application Layer
Most application layer startups are 90% traditional software with 10% AI.
The sustaining value lies in identifying an industry & deeply understanding the regulation, tooling, language, etc.
Tailor your software to fit into that industry in a way that's extremely deeply embedded.
Let us know what your big takeaways from this week’s shows were in the comments below!
Thank you for reading, and don’t miss the great guests we have next week:
Mike Schroepfer @ Gigascale
Matt Lerner @Systm
Thank you for reading 20VC.
This post is public so feel free to share it.
Thanks for reading 20VC! Subscribe for free to receive new posts and support my work.