StarZero All shows Try the podcast clipper →

Jul 1, 2026 · 19m · top-founders

Featherless: $3.6M Revenue Running 6,700 Open Source AI Models — Eugene Cheah

Eugene Cheah · 11m spoken Nathan Latka · 6m spoken
0:00 / 0:00

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Eugene Cheah, co-founder and CEO of Featherless AI, discusses how his serverless inference platform scaled to over $3.6 million in revenue by hosting 6,700+ open-source models to slash enterprise AI costs.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Nathan holds 35.5% of the talking time here. How this is scored →

Nathan as informed peer 3.7 Guest teaching 4.7 Guest disagreement 1.2 Nathan pushing back 2.8
05100:0010:000:00–2:15 · Nathan as informed peer 4/10 Interview Preview: Scaling Featherless AI and Revenue Milestones Latka immediately presses on exact monthly revenue figures and customer counts before asking Cheah to dumb down the technology for a kindergartner. Cheah politely obliges, explaining how domain-specific open-source models serve localized use cases like agriculture.2:15–5:37 · Nathan as informed peer 2/10 Platform Capabilities, Pricing Architecture, and User Profiles Latka explicitly admits he is not a technical person while probing how end users interact with the platform and who covers server credits. Cheah educates him using the analogy of Heroku and Vercel abstracting infrastructure for AI models.5:37–8:59 · Nathan as informed peer 4/10 Origins of Featherless: Accidental Growth and Server Scaling Latka steers the conversation into financial milestones and rapid revenue growth, noting that engineers often find finance questions uncomfortable. Cheah shares the origin story of how an accidental pricing experiment rapidly eclipsed their original company Recurso.8:59–11:12 · Nathan as informed peer 6/10 Slashing AI Costs and Go-to-Market Enterprise Strategy Latka showcases his own SaaS investment background, breaking down how startup P&Ls and COGS lines are inflated by OpenAI and Anthropic API spend. Cheah validates this framing, explaining how Featherless targets startups burning heavy compute.11:12–14:46 · Nathan as informed peer 3/10 Engineering Headcount and the Long-Tail Model Advantage Latka asks why Featherless is the only inference provider for certain models. Cheah educates him on market dynamics, explaining that major providers fight over the top 100 commodity models while Featherless captures high demand across the long tail of specialized 200B parameter fine-tunes.14:46–17:41 · Nathan as informed peer 3/10 Sovereign AI Trends, Efficiency Research, and Future Vision Latka asks why Cohere provides inference, prompting Cheah to deliver an educational overview of sovereign AI ecosystems across Canada, the US, and France. Cheah also articulates why biological human learning proves current LLM architectures are computationally inefficient.0:00–2:15 · Guest teaching 4/10 Interview Preview: Scaling Featherless AI and Revenue Milestones Latka immediately presses on exact monthly revenue figures and customer counts before asking Cheah to dumb down the technology for a kindergartner. Cheah politely obliges, explaining how domain-specific open-source models serve localized use cases like agriculture.2:15–5:37 · Guest teaching 6/10 Platform Capabilities, Pricing Architecture, and User Profiles Latka explicitly admits he is not a technical person while probing how end users interact with the platform and who covers server credits. Cheah educates him using the analogy of Heroku and Vercel abstracting infrastructure for AI models.5:37–8:59 · Guest teaching 3/10 Origins of Featherless: Accidental Growth and Server Scaling Latka steers the conversation into financial milestones and rapid revenue growth, noting that engineers often find finance questions uncomfortable. Cheah shares the origin story of how an accidental pricing experiment rapidly eclipsed their original company Recurso.8:59–11:12 · Guest teaching 2/10 Slashing AI Costs and Go-to-Market Enterprise Strategy Latka showcases his own SaaS investment background, breaking down how startup P&Ls and COGS lines are inflated by OpenAI and Anthropic API spend. Cheah validates this framing, explaining how Featherless targets startups burning heavy compute.11:12–14:46 · Guest teaching 6/10 Engineering Headcount and the Long-Tail Model Advantage Latka asks why Featherless is the only inference provider for certain models. Cheah educates him on market dynamics, explaining that major providers fight over the top 100 commodity models while Featherless captures high demand across the long tail of specialized 200B parameter fine-tunes.14:46–17:41 · Guest teaching 7/10 Sovereign AI Trends, Efficiency Research, and Future Vision Latka asks why Cohere provides inference, prompting Cheah to deliver an educational overview of sovereign AI ecosystems across Canada, the US, and France. Cheah also articulates why biological human learning proves current LLM architectures are computationally inefficient.0:00–2:15 · Guest disagreement 1/10 Interview Preview: Scaling Featherless AI and Revenue Milestones Latka immediately presses on exact monthly revenue figures and customer counts before asking Cheah to dumb down the technology for a kindergartner. Cheah politely obliges, explaining how domain-specific open-source models serve localized use cases like agriculture.2:15–5:37 · Guest disagreement 1/10 Platform Capabilities, Pricing Architecture, and User Profiles Latka explicitly admits he is not a technical person while probing how end users interact with the platform and who covers server credits. Cheah educates him using the analogy of Heroku and Vercel abstracting infrastructure for AI models.5:37–8:59 · Guest disagreement 1/10 Origins of Featherless: Accidental Growth and Server Scaling Latka steers the conversation into financial milestones and rapid revenue growth, noting that engineers often find finance questions uncomfortable. Cheah shares the origin story of how an accidental pricing experiment rapidly eclipsed their original company Recurso.8:59–11:12 · Guest disagreement 1/10 Slashing AI Costs and Go-to-Market Enterprise Strategy Latka showcases his own SaaS investment background, breaking down how startup P&Ls and COGS lines are inflated by OpenAI and Anthropic API spend. Cheah validates this framing, explaining how Featherless targets startups burning heavy compute.11:12–14:46 · Guest disagreement 2/10 Engineering Headcount and the Long-Tail Model Advantage Latka asks why Featherless is the only inference provider for certain models. Cheah educates him on market dynamics, explaining that major providers fight over the top 100 commodity models while Featherless captures high demand across the long tail of specialized 200B parameter fine-tunes.14:46–17:41 · Guest disagreement 1/10 Sovereign AI Trends, Efficiency Research, and Future Vision Latka asks why Cohere provides inference, prompting Cheah to deliver an educational overview of sovereign AI ecosystems across Canada, the US, and France. Cheah also articulates why biological human learning proves current LLM architectures are computationally inefficient.0:00–2:15 · Nathan pushing back 4/10 Interview Preview: Scaling Featherless AI and Revenue Milestones Latka immediately presses on exact monthly revenue figures and customer counts before asking Cheah to dumb down the technology for a kindergartner. Cheah politely obliges, explaining how domain-specific open-source models serve localized use cases like agriculture.2:15–5:37 · Nathan pushing back 3/10 Platform Capabilities, Pricing Architecture, and User Profiles Latka explicitly admits he is not a technical person while probing how end users interact with the platform and who covers server credits. Cheah educates him using the analogy of Heroku and Vercel abstracting infrastructure for AI models.5:37–8:59 · Nathan pushing back 3/10 Origins of Featherless: Accidental Growth and Server Scaling Latka steers the conversation into financial milestones and rapid revenue growth, noting that engineers often find finance questions uncomfortable. Cheah shares the origin story of how an accidental pricing experiment rapidly eclipsed their original company Recurso.8:59–11:12 · Nathan pushing back 2/10 Slashing AI Costs and Go-to-Market Enterprise Strategy Latka showcases his own SaaS investment background, breaking down how startup P&Ls and COGS lines are inflated by OpenAI and Anthropic API spend. Cheah validates this framing, explaining how Featherless targets startups burning heavy compute.11:12–14:46 · Nathan pushing back 3/10 Engineering Headcount and the Long-Tail Model Advantage Latka asks why Featherless is the only inference provider for certain models. Cheah educates him on market dynamics, explaining that major providers fight over the top 100 commodity models while Featherless captures high demand across the long tail of specialized 200B parameter fine-tunes.14:46–17:41 · Nathan pushing back 2/10 Sovereign AI Trends, Efficiency Research, and Future Vision Latka asks why Cohere provides inference, prompting Cheah to deliver an educational overview of sovereign AI ecosystems across Canada, the US, and France. Cheah also articulates why biological human learning proves current LLM architectures are computationally inefficient.

speaking balance: gold is Nathan, purple is the guest (3 minute bins)

0:00 · Nathan 41.6% · guest 58.4%0:00 · Nathan 41.6% · guest 58.4%3:00 · Nathan 26.8% · guest 73.2%3:00 · Nathan 26.8% · guest 73.2%6:00 · Nathan 26.7% · guest 73.3%6:00 · Nathan 26.7% · guest 73.3%9:00 · Nathan 41% · guest 59%9:00 · Nathan 41% · guest 59%12:00 · Nathan 26.9% · guest 73.1%12:00 · Nathan 26.9% · guest 73.1%15:00 · Nathan 24.9% · guest 75.1%15:00 · Nathan 24.9% · guest 75.1%18:00 · Nathan 98.5% · guest 1.5%18:00 · Nathan 98.5% · guest 1.5%
Sharpest disagreement ▶ 13:00 Cheah rejects competing for top commodity models

Cheah dismisses the conventional strategy of fighting giant compute battles over the top 10 models, arguing that true market leverage lies in serving the unaddressed long tail.

 Hardest push from Nathan ▶ 0:14 Latka pins down exact monthly recurring revenue

When Cheah answers vaguely about scaling towards multi-million dollar annual contracts, Latka refuses the deflection and explicitly boxes in the monthly revenue bracket.

 Biggest teaching moment ▶ 14:58 Cheah breaks down sovereign AI geopolitical landscape

Cheah thoroughly educates Latka on how Cohere, Mistral, and US frontier labs operate within domestic sovereignty strategies, explaining where open-source fine-tuning fills the global gap.

 Nathan holds their own ▶ 8:59 Latka demonstrates deep insight into SaaS COGS and AI spend

Latka draws upon data across 550 portfolio companies to detail exactly how API expenses from Anthropic and OpenAI heavily degrade software gross margins.
the scores for every segment, with the reasoning behind each
ChapterTopicNathan as informed peerGuest teachingGuest disagreementNathan pushing backWhy
Interview Preview: Scaling Featherless AI and Revenue Milestones 4414 Latka immediately presses on exact monthly revenue figures and customer counts before asking Cheah to dumb down the technology for a kindergartner. Cheah politely obliges, explaining how domain-specific open-source models serve localized use cases like agriculture.
Platform Capabilities, Pricing Architecture, and User Profiles 2613 Latka explicitly admits he is not a technical person while probing how end users interact with the platform and who covers server credits. Cheah educates him using the analogy of Heroku and Vercel abstracting infrastructure for AI models.
Origins of Featherless: Accidental Growth and Server Scaling 4313 Latka steers the conversation into financial milestones and rapid revenue growth, noting that engineers often find finance questions uncomfortable. Cheah shares the origin story of how an accidental pricing experiment rapidly eclipsed their original company Recurso.
Slashing AI Costs and Go-to-Market Enterprise Strategy 6212 Latka showcases his own SaaS investment background, breaking down how startup P&Ls and COGS lines are inflated by OpenAI and Anthropic API spend. Cheah validates this framing, explaining how Featherless targets startups burning heavy compute.
Engineering Headcount and the Long-Tail Model Advantage 3623 Latka asks why Featherless is the only inference provider for certain models. Cheah educates him on market dynamics, explaining that major providers fight over the top 100 commodity models while Featherless captures high demand across the long tail of specialized 200B parameter fine-tunes.
Sovereign AI Trends, Efficiency Research, and Future Vision 3712 Latka asks why Cohere provides inference, prompting Cheah to deliver an educational overview of sovereign AI ecosystems across Canada, the US, and France. Cheah also articulates why biological human learning proves current LLM architectures are computationally inefficient.

Statements from this episode (9)

Opinion
Cheah: Open Models Now Match Claude Sonnet and GPT-4o Mini
“So, and the, this growing collection of open models includes some of the best models that are already on par or surpass, let's say, Plot Sonnet or even GPT-A for Mini.”
Eugene Cheah Jul 1, 2026 ▶ 5:19
Assertion Not checkable as stated
Cheah: RWKV Architecture Could Reduce Inference Costs by 1,000x
“Like this new AI architecture has the potential of reducing inference costs by over a thousand X.”
Eugene Cheah Jul 1, 2026 ▶ 5:58
Assertion Not checkable as stated
Cheah: Long-Tail Fine-Tuned Models Drive 50% of Featherless Workload
“You see, most providers, they only provide, let's say, less than a hundred models. That covers 50% of our inference work. It's the bottom 50% where they run all these interesting fine-tuned models that people came on board for.”
Eugene Cheah Jul 1, 2026 ▶ 6:34
Assertion Not checkable as stated
Featherless Began as Pricing Experiment, Outperforming Main Platform in Days
“And Federalist was meant to be a name pricing experiment. So we gave it a different name, but within the first few days, it became more profitable and more revenue than the original company platform that we were like, I guess we are Federalist now.”
Eugene Cheah Jul 1, 2026 ▶ 7:47
Disclosure
Featherless Targets Startups Burning $100K Monthly on OpenAI and Anthropic
“We also realized that there is a lot of money on the table right now where you can go after the startups that, hey, I just built my entire startup or SMB on OpenAI or Entropic, and I'm burning a 100,000 dollars a month. And I do not know what I was doing. And …”
Eugene Cheah Jul 1, 2026 ▶ 10:40
Assertion Not checkable as stated
Featherless Won Multiple Contracts by Exclusively Hosting the StepFun Model
“The step one model is a particularly popular model for us that easily ship several contracts for us on this model alone. And no one else is.”
Eugene Cheah Jul 1, 2026 ▶ 13:05
Assertion Not checkable as stated
Cheah: StepFun AI Model Processes Billions of Tokens Daily
“Like, Step Fun, for example, is not an unpopular model. It's shipping billions of tokens per day.”
Eugene Cheah Jul 1, 2026 ▶ 14:12
Prediction Not checkable as stated
Cheah: Global AI Market Will Segment into Domestic Sovereign Models
“So they are going to the direction of highly tailored sovereign AI models for the domestic market. And we actually see this happening more and more. So for Cohear, they will service the Canadian market. For the US market is going to be served by OpenAI Entropi…”
Eugene Cheah Jul 1, 2026 ▶ 15:08
Assertion Not checkable as stated
Cheah: Featherless AI's Largest Customer Pays $1M to $2M Annually
“So currently the biggest will be around one to two million dollars a year, which may sound extremely large, but when you actually peel behind the layers, it only comes to around like five, six of the largest servers you see in the market.”
Eugene Cheah Jul 1, 2026 ▶ 16:12
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 2,600 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more