Pinned post
ashu garg
Verified account This account is verified. Learn more@ashugarg
Enterprise AI VC @FoundationCap | Early investor in @databricks @tubi & 6 other unicorns- @cohesity @eightfoldai @turingcom @amperity @alation @anyscalecompute
2.7K Following 13K Followers 11K Posts
Latest posts 5
Founders -
Painful sales cycles are your moat.
Whatever is easy to sell is also easy to substitute.
Customers who make you go through a long and painful sales process typically have very specific needs - security reviews, certifications, complex deployment requirements, or multiple layers of approvals.
They’re not being painful for the sake of being painful.
If you can meet those needs and still deliver something of value, you end up being deeply integrated into their workflows.
Any competitor who wants to replace you must also navigate the same process. Painful is good.
Painful sales cycles are your moat.
Whatever is easy to sell is also easy to substitute.
Customers who make you go through a long and painful sales process typically have very specific needs - security reviews, certifications, complex deployment requirements, or multiple layers of approvals.
They’re not being painful for the sake of being painful.
If you can meet those needs and still deliver something of value, you end up being deeply integrated into their workflows.
Any competitor who wants to replace you must also navigate the same process. Painful is good.
AI Founders -
Anthropic’s IPO means 2 things for you:
1. The IPO is huge validation of the AI market.
Anthropic is planning to raise $100 billion or more, which would make it the largest IPO in history. Validation like this in the public markets bodes well for AI adoption in enterprises.
But like any stock, there will be troughs. And no matter what we’d like to believe, private markets follow public markets. When public tech stocks fall, private valuations fall much harder.
So shore up your cash base. Tougher times will come, and you want the runway to get through them.
2. Post IPO, Anthropic will get much more aggressive about expansion.
Training new models is very expensive, but building applications is relatively cheap. So post IPO, I anticipate Anthropic will become more and more of an application provider. Which verticals and use cases is still unclear.
If you’re building an application company, expect Anthropic to become a direct competitor.
Your best defense is being deeply integrated with your customers. Find customers that have specific needs - security reviews, certifications, approvals from internal stakeholders. Getting through all of that takes months.
If you’re able to navigate those unique needs, and sell them something of value, you’re much harder to replace. Any competitor who wants to swap you out has to start the whole process again.
Anthropic’s IPO means 2 things for you:
1. The IPO is huge validation of the AI market.
Anthropic is planning to raise $100 billion or more, which would make it the largest IPO in history. Validation like this in the public markets bodes well for AI adoption in enterprises.
But like any stock, there will be troughs. And no matter what we’d like to believe, private markets follow public markets. When public tech stocks fall, private valuations fall much harder.
So shore up your cash base. Tougher times will come, and you want the runway to get through them.
2. Post IPO, Anthropic will get much more aggressive about expansion.
Training new models is very expensive, but building applications is relatively cheap. So post IPO, I anticipate Anthropic will become more and more of an application provider. Which verticals and use cases is still unclear.
If you’re building an application company, expect Anthropic to become a direct competitor.
Your best defense is being deeply integrated with your customers. Find customers that have specific needs - security reviews, certifications, approvals from internal stakeholders. Getting through all of that takes months.
If you’re able to navigate those unique needs, and sell them something of value, you’re much harder to replace. Any competitor who wants to swap you out has to start the whole process again.
@ramkris and @ajkrish95 are building AI agents that do the work of a full finance team. We’ve backed what they’re building at @HyphenateAI since their seed round, and last week they announced their $45M Series A.
In this month’s B2BaCEO, I share their story and how it aligns with our thinking on the opportunity in AI apps. In verticals like finance, where mistakes are costly and the work is heavily regulated, a smarter model alone won’t get the job done. We believe the winners will be engineers like Ram and Ajay, who know their customers’ work deeply and are building their apps with the technical depth of an infra company.
More on how they got here: foundationcapital.com/ideas/always-a…
In this month’s B2BaCEO, I share their story and how it aligns with our thinking on the opportunity in AI apps. In verticals like finance, where mistakes are costly and the work is heavily regulated, a smarter model alone won’t get the job done. We believe the winners will be engineers like Ram and Ajay, who know their customers’ work deeply and are building their apps with the technical depth of an infra company.
More on how they got here: foundationcapital.com/ideas/always-a…
We were scared of plenty of technologies before AI.
Take PlayStations.
In 2000, Japan put export controls on the PlayStation 2 amid concerns that its chips could be used for military purposes.
Sensationalist stories spread that Saddam Hussein was stockpiling 1,000s of consoles, wiring together chips into a supercomputer to simulate missile guidance.
Consoles kept getting more powerful, and those fears faded.
I hear a version of the same argument about AI today: the models are too powerful to release, and the safe thing is to keep them closed.
Some of the concern about AI is fair.
But a scary story is a bad reason to keep a technology closed. We need to be specific about what could go wrong and how to prevent it.
That starts with separating what a model can do from what we allow it to do: what systems it can access, which actions need human approval, and where deterministic rules take over.
The risks are real, but they’re product and engineering problems we can solve. We shouldn’t underestimate how much agency we have to get this right.
Take PlayStations.
In 2000, Japan put export controls on the PlayStation 2 amid concerns that its chips could be used for military purposes.
Sensationalist stories spread that Saddam Hussein was stockpiling 1,000s of consoles, wiring together chips into a supercomputer to simulate missile guidance.
Consoles kept getting more powerful, and those fears faded.
I hear a version of the same argument about AI today: the models are too powerful to release, and the safe thing is to keep them closed.
Some of the concern about AI is fair.
But a scary story is a bad reason to keep a technology closed. We need to be specific about what could go wrong and how to prevent it.
That starts with separating what a model can do from what we allow it to do: what systems it can access, which actions need human approval, and where deterministic rules take over.
The risks are real, but they’re product and engineering problems we can solve. We shouldn’t underestimate how much agency we have to get this right.