Keith Rabois on Ramp's moat, humanoid robot risks from China, and why he's skeptical of the crypto reserve

Summary

Keith Rabois on Ramp, humanoid robots, and the crypto reserve

Ramp's pitch has always been simple: save companies time and money. What's changed, according to Rabois, is that there's now empirical evidence across company types to prove it — 5% savings here, 10% there — in a CFO function he describes as one of the most archaic in any organization, even at cutting-edge firms. The broader vision is digitizing the entire finance suite, and Rabois argues Ramp will eventually be a monopoly in that space. What makes the current moment notable is that Ramp is defying the typical growth-stage slowdown: product velocity is accelerating, talent quality is rising, and Rabois describes the intern classes of the past two years as the best anywhere globally.

Founder formula

Rabois credits co-CEOs Eric Glyman and Kareem Atiyeh as the core engine. Glyman has rare marketing instincts — the ability to cut through a crowded category and reinvent the value proposition every year or two. Atiyeh sets the engineering tempo and quality bar. The parallel Rabois draws is Stripe: two founders in tight sync, effectively giving investors two CEOs for the price of one. He says that combination is extraordinarily hard to find and may be the clearest single signal to invest when you do find it.

AI revenue durability

Not all nine-figure ARR is created equal. Rabois draws three tiers. OpenAI's ChatGPT subscription base is durable — individual consumer and business habits are sticky. Pilot-stage enterprise contracts are not durable at all. The middle tier — where an AI product clears procurement, gets rolled out across a 14,000-person organization, and changes business processes — is sticky for five-plus years. The category he's most skeptical of is training and post-training revenue: advances are moving fast enough that today's fine-tuning methods could be obsolete in one to two years, and companies built on that revenue may not survive the transition.

Humanoid robots and the China risk

Rabois is direct about the competitive threat: China is ahead in humanoid robotics, and he expects the Trump administration will need to restrict Chinese robots before the window closes. The DJI playbook is the template — Unitree is already selling humanoids below component cost, replicating the strategy that flooded the drone market. A robot that can run 17 miles per hour and collect data inside a facility has obvious military applications, which Rabois argues makes this a national security issue well beyond TikTok.

For US startups, the sector is currently the hottest category in venture — robotics founders are walking in and being offered hundreds of millions of dollars. Rabois is cautious about deployment economics, though. Depreciation and wear curves are genuinely unknown: a robot doing light warehouse work might last years; the same unit under heavy industrial use could degrade in months. Until operators have actually run robots through real tasks and tracked what breaks, how fast, and at what replacement cost, the unit economics are guesswork.

Traba's longer-term positioning

In light industrial warehouses, most tasks are not yet ready for robotic substitution. But Rabois sees a credible 10-year path where Traba — having already earned customer trust by managing the most sensitive part of a warehouse operation, the human labor — becomes the natural channel to introduce mixed human-robot systems. A company parachuting in from the outside with no understanding of how warehouses actually work is unlikely to succeed. The advantage goes to whoever already has the operational relationship.

Bitcoin reserve

Rabois supports the Trump administration broadly but is not a fan of the Strategic Bitcoin Reserve. If it had to be done, he would have done Bitcoin only — including other cryptocurrencies created messaging problems he sees as self-inflicted. His deeper objection is philosophical: he doesn't think government should perform business functions, whether that's a crypto reserve or a sovereign wealth fund. The only internally consistent argument for a reserve, he concedes, is the gold analogy — if you believe Bitcoin is the future store of value, you can talk yourself into it. Funding it with seized criminal assets rather than taxpayer dollars is the version he finds least objectionable.

Google vs. Apple

Asked about Sundar Pichai's job security, Rabois argues Pichai is at least outperforming Tim Cook — a pointed observation given Apple's announcement that its AI features are delayed by roughly two years. Two years in the current AI environment, Rabois says, is effectively equivalent to 20. He thinks Pichai made miscalculations, and that Microsoft and Satya Nadella, despite getting significant credit for the OpenAI bet, are starting to show the limits of their strategy. Microsoft's AI lock may not be as durable as it appeared, and the people who control AI's future may not be in Redmond.