Latest / Tech IPO Conversations with Fexingo: Public Software Companies and Capital Markets / Why Post-IPO AI Spending Is Under Investor Scrutiny
Transcript
- Lucas: So there's a number floating around today that I think is going to become a major flashpoint for post ipo tech companies. A new report out this morning says 'ai pilled' firms are spending about $7,500 per employee every month on artificial intelligence. Luna: Wait — that's per employee. So a company with 10,000 knowledge workers is burning $75 million a month on AI tools, models, compute? Lucas: Exactly. And that's not even the high end. Some firms are reportedly spending north of $10,000 per employee per month. Now, for a pre ipo startup with venture funding, that's a growth story. But for a company that just went public and has to report earnings every quarter — that number is a lightning rod. Luna: And we're seeing the market react, right? Look at the big tech sell-off this week. META down almost 9 percent in five days. NVIDIA off over 8 percent. Microsoft down 6.7. Investors are getting twitchy about AI capex. Lucas: Right. But here's the nuance — the spending isn't uniform. Some of that $7,500 is recurring license fees for tools like Copilot or Claude Enterprise. Some is one-time fine-tuning costs. And a big chunk is compute — renting GPUs from cloud providers or building your own clusters. Luna: So how should a post ipo company frame this for investors? Because the old answer — 'we're investing for growth' — is wearing thin. Lucas: That's exactly the question. I think the smartest post ipo firms are now breaking out AI spend into two buckets: defensive and offensive. Defensive is the productivity layer — automating customer support, internal code generation, that sort of thing. That's easier to justify because you can tie it to headcount savings or faster feature delivery. Luna: And offensive is the bet on new products or revenue streams. Lucas: Right. And that's where investors are getting skeptical. Because offensive AI spend is harder to measure. A company like Palantir — they spend heavily on AI, but they can point to specific government contracts. But a consumer tech company? It's harder to draw a straight line from an AI chatbot to incremental revenue. Luna: Yeah, and Palantir's stock is down over 8 percent in the last five days too. So even companies with a clear AI story aren't immune. Lucas: That's the climate we're in. I think what's changed is the market's patience. In 2023 and 2024, investors rewarded any company that said 'AI' in an earnings call. Now they want to see ROI. And for post ipo companies that don't have the balance sheet of a Microsoft or a Meta, that scrutiny is brutal. Luna: What's the risk for a newly public company that overspends on AI and can't show results? Lucas: Two things. One, the stock gets punished — we're seeing that play out now. But two, and maybe more importantly, it constrains future options. If you've burned through cash on AI infrastructure that isn't paying off, you have less flexibility to acquire, to hire, to invest in other areas. And your next earnings call becomes a referendum on every dollar spent. Luna: So what's the benchmark? Is there a healthy percentage of revenue that a post ipo tech company should be spending on AI? Lucas: There's no universal number, but if you look at the companies that are navigating this well — they're spending around 5 to 8 percent of revenue on AI, and they're showing a clear path to that spend declining as a percentage over 18 to 24 months. The ones getting crushed are spending 15 percent or more with no timeline for efficiency. Luna: And that's where the $7,500 per employee figure becomes a red flag. Because if you're a 5,000-person post ipo company, that's $37.5 million a month. Over a year, that's $450 million. For a company with maybe $2 billion in revenue — that's over 20 percent. Lucas: Exactly. And that's why I think we'll see a wave of AI cost optimization in the second half of 2026. post ipo companies will start publicly announcing efficiency initiatives — consolidating model providers, moving from real-time inference to batch processing, cutting back on experimental projects. Luna: Interesting. So the narrative flips from 'how much we're spending on AI' to 'how efficiently we're spending on AI'. Lucas: That's the shift I'm watching. And I think the companies that communicate that transition early — before their next earnings — will be rewarded. Because right now the market is hungry for evidence that AI spend is disciplined, not just big. Luna: Speaking of disciplined spending — you know, it's funny, keeping this podcast ad-free and independent is kind of our own version of that. We're not burning cash on a big production machine. And a couple of dollars a month from listeners who find these conversations useful genuinely makes a difference. If you've gotten something out of the show, buy me a coffee dot com slash fexingo. Lucas: Yeah, and it's a small thing that keeps us from ever having to chase sponsorship dollars or tailor content for advertisers. So if you value that independence, it helps. Luna: Alright, back to it. So you mentioned companies will start consolidating model providers. Who do you think benefits from that shift? Lucas: I think the big cloud providers — Amazon, Microsoft, Google — they win because they offer the integrated stack. But also the companies that help manage AI costs, like Databricks or Snowflake. They position themselves as the control plane for efficient AI. Luna: And what about the model makers themselves? OpenAI, Anthropic. Does this pressure them to offer more cost-effective tiers? Lucas: Absolutely. We're already seeing OpenAI introduce cheaper, faster models. And I think Anthropic is going to have to follow. Because if post ipo companies start optimizing, they'll drop the most expensive API calls first. The model providers that can't show a path to lower per-token costs will lose market share. Luna: So the winners and losers in AI are being decided not just by technology, but by the capital discipline of the customers buying that technology. Lucas: Exactly. And for a post ipo company, that's the real skill — knowing when to go all in and when to pull back. The ones that get that timing right will have a durable advantage. Luna: Alright, great conversation. That's all for this episode of Tech IPO Conversations. We'll see you next time.