Discover
BILLIONS
BILLIONS
Author: Guillaume Moubeche
Subscribed: 4Played: 47Subscribe
Share
© Guillaume Moubeche
Description
After building my company to a $150M valuation in 4 years, I had one question left: How do you build a billion-dollar company? I’m Guillaume Moubeche, and on the BILLIONS Podcast, I’m taking you inside the room with the world’s most iconic builders, founders, and investors to find the answer. This is more than just another startup podcast; it’s a masterclass in high-growth SaaS, AI implementation, and wealth creation.
From SaaS growth strategies and AI Agent pivots to the raw truth behind venture capital and exit strategies, we go where others don't.
What you’ll learn on BILLIONS:
SaaS Scal
From SaaS growth strategies and AI Agent pivots to the raw truth behind venture capital and exit strategies, we go where others don't.
What you’ll learn on BILLIONS:
SaaS Scal
38 Episodes
Reverse
Today on BILLIONS, I’m sitting down with Peter Wagner, founding partner at Wing Venture Capital.Peter has backed more than 30 companies that went on to IPO or be acquired, including leading Wing’s seed investment in Snowflake.Today, he’s trying to answer the same question again:How do you spot billion-dollar companies before everyone else?At the earliest stages, there are no obvious growth metrics.No dominant market share.Sometimes barely even a product.So Peter looks at three things: market, technology, and founders, and whether they have the ambition to build something enduring.In this conversation, we break down:→ How Peter identifies billion-dollar opportunities before they become obvious → Why Wing often starts with the market before evaluating the founder → Which layers of the AI stack are actually going to capture value → How proprietary data creates compounding advantages → Why Gong’s data layer is becoming more valuable as its agents improve → Whether everyone will eventually have a personal AI assistant → Why 2027 could be the year enterprise AI moves from experimentation to real production → The “model, harness, context” framework for building agentic systems → Why enterprise AI needs reliability that frontier labs aren’t necessarily optimizing for → Why open-weight models became one of the biggest AI shifts of 2026 → Where world models and world-action models could take robotics next → Why the next generation of AI companies will need entirely new training data → Where the moat really sits: models, applications, or proprietary context → Why AI-native software may never have traditional SaaS gross margins → Why compute infrastructure increasingly looks more like an industrial business than software → Why demand for compute could exceed supply for years → Whether today’s AI market is actually a bubble → Why Peter thinks this AI supercycle could dwarf the internet era → How seed investing changes when rounds and valuations keep getting biggerOne of the most interesting points: Peter believes we’re dealing with two separate cycles.The technology cycle keeps progressing.The capital market cycle overshoots and undershoots around it.So yes, some AI valuations may be excessive today.But that doesn’t mean the underlying transformation is.If you want to understand how one of Silicon Valley’s most experienced early-stage investors thinks about the next generation of billion-dollar companies, this is the episode:00:31 – How do you spot billion-dollar companies before everyone else?01:47 – Founder vs. market: what matters most?03:11 – Where will value be captured in the AI stack?04:38 – AI-native companies transforming traditional industries07:38 – How to evaluate proprietary data as a moat09:05 – Gong and the compounding data advantage10:58 – Can AI replace traditional intermediaries?11:48 – Will everyone have a personal AI assistant?14:12 – Pinecone and the enterprise context layer15:30 – Why reliability is still a major AI problem16:02 – LLMs, world models and open-weight models19:33 – AI, biology and the next generation of training data21:45 – What are production traces?24:53 – Where is the real AI moat: models or applications?25:46 – Why AI software won’t look like traditional SaaS27:42 – The infrastructure and data centers behind AI30:02 – Building a marketplace for compute33:18 – Is AI actually a bubble?35:55 – Why AI is different from the telecom bubble36:51 – Are venture funds becoming too big?39:33 – Where to follow Peter and Wing
Today on BILLIONS, I’m sitting down with Aytekin Tank, founder and CEO of Jotform.Over the last 20 years, he built Jotform from a one-person bootstrapped startup into a global software platform used by more than 40 million people.He never raised outside capital.He still owns 100% of the company.And after reaching nine figures in revenue, Jotform is now going through one of the biggest transformations in its history: rebuilding its products around AI.Aytekin’s story is a masterclass in patience.For the first two years, he worked alone.For the first five years, the company basically grew one employee at a time.Then came Google Forms, Typeform, multiple rewrites of the core product, SEO shifts, new products, and now AI.Through all of it, one principle stayed the same:A founder’s job isn’t just to start the company.It’s to keep founding its next chapter.In this conversation, we break down:→ How Aytekin bootstrapped Jotform to 40M+ users without raising a dollar → Why he believes founders should keep “founding” their company every day → Why stepping away from the product almost hurt Jotform twice → How Google Forms forced him to rethink his role as founder → Why Jotform has rewritten its core forms product four times — and is doing it again with AI → How customer interviews helped Jotform become a multi-product company → The SEO strategy that helped Jotform grow through thousands of highly specific use cases → Why losing Google traffic barely affected Jotform’s real business metrics → How Jotform maintained ~50% growth for years → Why Aytekin still owns 100% of the business → His two financial rules: one year of payroll in the bank and ~25% margins → Why your best employees should feel “painfully expensive” → Why AI could make great employees 10x — and eventually 100x — more valuable → Why Aytekin doesn’t believe AI will replace people → Why he thinks SaaS companies could actually grow MORE because of AI → Why vibe-coding your own SaaS may cost far more than simply buying it → His advice for anyone trying to bootstrap a company todayOne of my favorite ideas from the conversation:AI can automate the work.But it cannot take responsibility for the outcome.And for Aytekin, that means great people become more valuable — not less — as AI gets better.If you want to understand what 20 years of patient, profitable, founder-led compounding looks like, this is the episode.00:32 – Aytekin Tank: 20 years bootstrapping Jotform to 40M+ users 01:07 – The different stages of building Jotform 03:14 – Why founders must keep “founding” their company 06:08 – Delegating everything to focus on one thing 09:06 – Rewriting the product again and again 11:26 – How Jotform became a multi-product company 12:57 – The real growth engine behind Jotform 16:08 – What happened when Google traffic started falling 19:19 – Why Jotform never shares its revenue publicly 21:58 – Growing 100%, then 50%, without outside capital 23:05 – Owning 100% of the company and taking dividends 24:08 – Aytekin’s 2 financial rules for bootstrapping 26:18 – How to retain your best employees without equity 29:41 – Will AI destroy SaaS? 31:08 – Why AI can’t replace people 32:04 – From 1x to 10x to 100x employees with AI factories 33:39 – Why vibe-coding won’t kill SaaS 37:30 – Does Aytekin ever want to sell Jotform? 38:49 – His advice to bootstrap founders today 40:46 – Where to follow Aytekin and Jotform
Today on BILLIONS, I’m sitting down with David Steinberg, a seven-time founder who has spent more than 30 years building companies.He has sold four, taken two public, and today leads Zeta Global, an intelligent AI infrastructure company worth around $8 billion.But one of the most important decisions of his career came in 2017.Zeta was on track to generate around $50 million in EBITDA.David went to his board and told them he wanted to effectively give up profitability for the next two to three years, put 400 engineers on a new platform, and rebuild the company around data and artificial intelligence.The board meeting became one of the most contentious of his career.Then, after one investor pushed back hard against two acquisitions David wanted to make, he decided to buy both.That decision helped create the Zeta that exists today.In this conversation, we break down:→ What David learned building companies through the internet, mobile and AI revolutions→ Why proprietary data and distribution become MORE valuable during technology shifts→ Why he believes giving company data to frontier models can destroy your competitive advantage→ The decision to sacrifice ~$50M of EBITDA and rebuild Zeta from the ground up→ The board meeting that ended with David deciding to buy two companies instead of one→ How Zeta turned data and AI into the foundation of its platform→ Why its customers generate roughly 600–700% return on marketing spend→ How 89.6% of Zeta’s new code is now generated automatically→ Why a platform that required 400 engineers in 2017 could potentially be built with 40 today→ Why enterprises pay for fighter jets but often only know how to fly prop planes→ How Athena is changing the way customers interact with enterprise software→ Why Zeta’s net retention rate is above 120%→ Why David believes small inference models can outperform frontier models on business tasks→ What the agentic internet could look like→ The 5 rules Zeta uses before acquiring a company→ How Zeta completed 19 major acquisitions with 18 considered successful→ Why so many of Zeta’s senior executives came through M&A→ David’s pitch for turning entrepreneurs into “intrapreneurs”We also discuss OpenAI, Anthropic, Palantir, data sovereignty, AI agents, the future of enterprise software and what 30+ years of building companies has taught David about surviving technology revolutions.If you want to understand how a serial founder thinks about rebuilding companies before technology forces them to, this is the episode.00:00 – David Steinberg: 7 startups, 4 exits and 2 IPOs00:32 – Meet the founder behind Zeta Global01:01 – What the internet, mobile and AI revolutions have in common03:24 – Why proprietary data is your biggest competitive advantage04:53 – Why David sacrificed $50M in EBITDA to rebuild Zeta08:03 – The two acquisitions that transformed Zeta09:58 – Inside Zeta's most contentious board meeting12:23 – From the 2017 pivot to AthenaOS13:20 – How Zeta delivers 600–700% return on marketing spend15:33 – Why Zeta is expanding beyond marketing17:45 – How AI automates 89.6% of Zeta's new code18:54 – Why enterprises buy fighter jets but fly prop planes21:13 – Athena's adoption: 40% of major clients and 83% voice usage21:45 – How proprietary context drives 120%+ net retention24:35 – Why inference models could outperform frontier models27:39 – The future of AI agents and agentic commerce29:27 – What David learned from 19 acquisitions30:53 – Zeta's 5 non-negotiable rules for M&A32:35 – How to turn acquired founders into intrapreneurs33:12 – David's most important job as a founder34:11 – Final thoughts and where to follow David Steinberg
Today on BILLIONS, I’m sitting down with George Bonaci, VP of Growth & Demand Gen at Ramp.Before Ramp, George led growth at Gong and helped scale Samsara from under $100M ARR to more than $650M ARR through its IPO.Today, he’s helping build the growth engine behind Ramp — the finance platform valued at $44 billion and used by more than 70,000 companies.And his approach to growth is pretty unconventional.Ramp intentionally sets aside budget with zero expectations of return.Why?Because if every experiment works, you’re probably not taking big enough swings.One of those swings involved spending six figures advertising Ramp inside fortune cookies.George thought it sounded insane.It worked.In this conversation, we break down:→ How Ramp keeps operating like a startup at a $44B valuation → Why George believes most growth teams don’t fail enough → The six-figure fortune cookie experiment that actually worked → Why learning in one week can be worth spending 5x more → How Ramp structures small autonomous “pods” to move faster → How George allocates talent and budget across growth bets → What Ramp’s spending data reveals about the AI model wars → Why companies investing heavily in AI are actually hiring MORE people → Why AI could make great operators 10x — or even 100x — better → Why George refuses to automate away the jobs that train future experts → Why the “SaaS apocalypse” narrative may be completely wrong → Why Ramp prefers using AI to create new tools instead of replacing existing SaaS → What happens when AI agents start buying from other AI agents → Why AEO could become much bigger than “SEO for ChatGPT” → How products will need to become discoverable, understandable and usable by agents → The one bet Ramp absolutely cannot afford to get wrongWe also discuss why AI-generated 40-page documents are usually a sign of poor thinking, why the best candidates simply say “let me show you,” how work becomes invisible inside AI-native products, and why speed may still be the greatest competitive advantage a startup has.If you want to understand how one of the fastest-growing fintech companies in the world thinks about growth, AI and experimentation, this is the episode.00:00 – George Bonaci: Samsara, Gong and $44B Ramp00:46 – Why your growth bets should fail01:58 – Ramp’s six-figure fortune cookie experiment03:38 – How much should you spend on an experiment?07:11 – How Ramp still operates like a startup at $44B08:46 – Building autonomous growth pods11:38 – Taking a new Ramp product from zero to growth14:00 – How to allocate your best people across bets15:45 – What Ramp’s data reveals about the AI model wars17:13 – Why AI adopters may actually hire MORE people18:23 – Is AI making us work harder instead of less?21:12 – Hiring great people in the age of AI slop22:16 – Why great thinking should fit on one page25:01 – What happens to jobs when AI does the execution?28:06 – Will AI kill entry-level jobs?30:38 – Is the “SaaS apocalypse” actually happening?32:28 – Using AI to build vs. replacing existing software34:03 – When AI agents start buying from AI agents35:28 – How to make your product discoverable by AI37:24 – Giving AI agents the ability to spend money41:03 – Ramp vs. incumbents like American Express42:24 – Ramp’s long-term vision43:15 – What George would do differently44:08 – The one bet Ramp cannot get wrong
Today on BILLIONS, I’m sitting down with Arvind Jain, founder and CEO of Glean.Before Glean, Arvind spent years at Google working on search, then co-founded Rubrik, which became a multi-billion-dollar public company.Instead of retiring, he started again from scratch.Today, Glean has crossed $300M ARR and was last valued at $7.2B.And Arvind has some very strong views on where AI is going next.He believes open-source models will ultimately win for most technologies, and says they can already handle 80–90% of what enterprises want to do with AI.He also believes frontier models will increasingly become indistinguishable from one another.Which raises a much bigger question:If the model itself becomes a commodity, where does the real value live?For Arvind, the answer is context.In this conversation, we break down:→ Why Arvind started again after already building a multi-billion-dollar company→ How a problem inside Rubrik led to the creation of Glean→ Why enterprise context may matter more than the underlying AI model→ How Glean reduces hallucinations by validating AI outputs against company data→ Why enterprises need more than “zero data retention” to safely deploy AI→ Why Arvind believes open source will ultimately win→ Why open-source models can already handle 80–90% of enterprise AI tasks→ Why frontier models could become increasingly commoditized→ How Glean routes every task to the right model based on quality and cost→ Why specialized smaller models can outperform frontier models on specific tasks→ Why AI may actually make us work MORE, not less→ Whether intelligence still matters when everyone has access to AI→ How AI agents could fundamentally change the way the internet is designed→ Why autonomous agents could potentially spend hundreds of thousands of dollars per day without the right guardrails→ What Arvind thinks Google and Gemini need to do next→ The biggest mistake he made while building GleanOne of the most interesting parts of the conversation:Glean was already training transformer-based language models for individual enterprise customers in 2019, years before ChatGPT launched.If you want to understand where enterprise AI is really going, beyond the hype around the latest models, this is the episode.TIMELINE:00:26 – Arvind Jain: from Rubrik to $7.2B Glean00:52 – Why start again after already building a huge company02:24 – How Glean works and why enterprise context matters04:57 – Solving AI hallucinations with validation06:28 – Keeping enterprise data safe with AI09:43 – Why Arvind believes open source will win11:57 – If AI models become commodities, where is the moat14:47 – Why specialized AI models will outperform general models16:50 – How Glean chooses the best model for every task19:00 – How to actually measure whether AI did a good job21:13 – Will AI make us work less… or more?23:42 – What happens when intelligence becomes abundant26:41 – How AI agents will change the internet29:24 – Building at AI speed inside slow-moving enterprises34:09 – Arvind’s take on Google, Gemini and frontier models36:16 – The one thing he would do differently building Glean37:28 – Where to follow Arvind and Glean


![What do 30+ IPOs & exits teach you about spotting $1B companies? - Peter Wagner [Wing VC] What do 30+ IPOs & exits teach you about spotting $1B companies? - Peter Wagner [Wing VC]](https://s3.castbox.fm/01/58/4b/33d8c03662ff36f2763e313757ebaad373_scaled_v1_400.jpg)

