Key Takeaways

  • Glean founder Arind revealed his company spent $1 million per month on a single internal AI triage agent, deeming current AI model prices "absurdly expensive."
  • He challenges direct comparisons between technology costs and human labor, believing AI inferencing prices will drop “by orders of magnitude” as technology matures and commoditizes.
  • Arind contends that an "overabundance of capital" in the startup ecosystem is fostering unsustainable practices, such as seed-stage companies paying half-million-dollar engineer salaries.
  • These inflated costs, both for AI services and talent, are not a marker of efficiency or enduring business but rather a "failure path" enabled by cheap money.
  • Founders should brace for a future where AI becomes a cheap commodity, and current spending patterns based on excessive capital are simply not sustainable for the long run.

The $1 Million Dollar Wake-Up Call

Founders are pouring capital into AI, but at what cost? Arind, founder of Glean, just pulled back the curtain on one startling example. He revealed his company was dropping a cool $1 million per month on an internal AI agent designed for engineering triage. His reaction? "Absurdly expensive." The kind of number that makes you question if the technology is truly more efficient than a human, even as he acknowledged the agent was "really cool." This isn't just about big numbers; it's about a distorted economic reality where cutting-edge tech demands a king's ransom, yet its long-term value against traditional labor is still unproven at these prices.

Arind's core argument isn't that AI isn't powerful. It's that its current pricing is out of whack. He quickly pushes back on the idea that today's AI costs should be benchmarked directly against human salaries. Why? Because “good technologies figure out how to make technology really, really cheap and it's going to happen here too.” He predicts that inferencing costs—the actual compute power needed to run AI models—will “come down by orders of magnitude.” The writing is already on the wall, Arind notes, with open-source alternatives already capable of doing the “same work for a tenth of the cost.”

The Capital Glut's Distorted Reality

The real culprit behind these "absurdly expensive" prices, according to Arind, isn't just the technology itself. It's the "overabundance of capital available today in the for startups." This flood of money distorts market prices, making both AI services and talent appear far more valuable than they sustainably are. He points to seed-stage companies paying half a million dollars to an engineer as a glaring example. "It's happening today," he says, and while founders and investors might be "okay with it," he's clear: “it's just surely not a sustainable path to actually win.”

This isn't merely an observation; it's a warning. When capital is too cheap and plentiful, it encourages practices that look like success on a spreadsheet but erode the foundation of a real business. Startups become accustomed to inflated costs—whether for AI models or top-tier talent—that can't possibly hold up once funding rounds get tighter or the market corrects. Arind sees these as "failure paths," setting companies up for a fall by valuing short-term burn over long-term endurance.

What to Do With This

If you're building, stop modeling your AI spend based on current vendor prices. Instead, build your next 18-month financial projections assuming a 70-90% drop in AI inference costs. This forces you to think about sustainable unit economics now, before the market corrects. When hiring, benchmark your offers against what an enduring, profitable business can afford, not against the highest seed-round salaries your competitors are burning through. Resist the artificial market pressures; your long-term viability depends on it.