Most companies using AI are 'lighting money on fire,' says Matthew Prince
Claims that most companies are “lighting money on fire” with AI spending have triggered debate over whether current investments are wasteful hype or a rational bet on a potentially massive upside. Commenters note that many firms are bolting AI onto products for marketing reasons or vague mandates to “use AI,” echoing past bubbles like blockchain and the metaverse, while others point to concrete productivity gains from tools like code assistants and customer-service automation. Several argue that even if most projects fail, treating AI as high-risk R&D can still be justified, provided the few successes deliver outsized returns.
Article and Headline Quality
- Many commenters say the article is shallow or clickbait: bold claim, almost no context, reasoning, or examples.
- Some note that it mostly strings together a couple of quotes without real analysis or conclusions.
AI as Hype, Marketing, and Buzzword
- Strong sense that many companies start from “we must use AI” and then hunt for a problem.
- AI is often seen as a marketing label on trivial features or existing techniques, similar to past waves (blockchain, metaverse, B2B buzzwords).
- Some users consciously downgrade products that slap “AI” on everything; others argue it’s rational trend-chasing because customers and investors demand it.
R&D, Expected Value, and Gamble Framing
- Several comments frame AI spending as R&D or high-risk bets: many failures are acceptable if just a few hits have huge payoff.
- Debate over whether high-upside, low-probability bets make sense for individual firms vs. only at portfolio or industry level.
- Some argue large corporations can treat AI projects as diversified “bets” among many initiatives.
Disruptive vs. Sustaining Innovation
- One line of argument: current AI is mostly a sustaining technology that improves existing use cases, favoring incumbents with data and capital.
- Others counter that future agentic and robotics applications could be genuinely disruptive, but this is still speculative.
Real Use Cases vs. “Stochastic Parrot” Skepticism
- Reported productive uses: code assistance (e.g., Copilot), marketing copy, text summarization, Q&A, creative prompts, internal automation, and RAG.
- Some say generative models already meaningfully boost productivity and replace certain roles.
- Others insist current models “solve nothing,” likening the moment to the crypto bubble; defenders rebut with concrete personal productivity gains.
Labor, Jobs, and Trade Work
- Discussion that AI is hitting creative/knowledge work faster than manual trades, contrary to earlier automation narratives.
- Some expect widespread displacement in bookkeeping, marketing, and other white-collar fields; others note legal/liability and licensing constraints.
Cost, Infrastructure, and Commoditization
- Training and running competitive models is seen as extremely expensive and quickly commoditized, with features rapidly copied.
- Open-source models are getting better and cheaper, pressuring commercial providers.
Startups, VC, and “Lighting Money on Fire”
- Multiple comments say “lighting money on fire” describes most VC-backed startups, not just AI.
- Stories of lavish spending, offsites, and perks at unprofitable companies support the view that hype-driven waste is common across cycles.