Over the past 12 months, OpenAI has restructured its organization five times. The Preparedness Team—responsible for catastrophic risk assessment—is now dissolved. Annualized revenue hit $40B, yet the $1T valuation implies a 25x P/S multiple. These numbers don't reconcile unless you understand the real signal: OpenAI is trading safety for speed, and the market is pricing in a future that may not materialize.
Speed is the only currency that doesn’t inflate. But when you sacrifice the integrity of your security architecture, you erode the trust that underpins that speed. Let me walk through the data.
Context: The Transition from Research Lab to Commercial Machine
OpenAI started as a nonprofit research lab. By 2023, it had pivoted to a capped-profit model, then to a full-scale commercial entity. The 2024 leadership crisis—Sam Altman's brief ouster—was a signal that the governance structure was brittle. Fast forward to August 2025: the company has gone through five major reorganizations in 12 months. The Preparedness Team, established in 2023 to assess frontier risks (bioweapons, cyberattacks, autonomous replication), has been disbanded and its functions distributed across product teams. Chief Revenue Officer Denise Dresser left. Ethics lead Chloé Bakalar left. The CTO role is in flux.
These are not isolated departures. They are a pattern. And they coincide with an IPO preparation that could value the company at $1 trillion.
Core: The Numbers Behind the Narrative
Let’s start with the financials. $40B annualized revenue, up from $24B at the end of 2024—a 67% growth rate. That’s impressive by any metric. But the $1T valuation expectation implies a 25x price-to-sales multiple. Compare that to Microsoft at ~12x, Google at ~7x. For a company still burning cash on training costs and inference infrastructure, that multiple is justified only if growth continues at 50%+ for the next three years.
The structural problem is that the very engine of that growth—product velocity—is being fueled by a reduction in safety oversight. My analysis of the Preparedness Team dissolution shows a clear organizational shift: safety is no longer a separate gatekeeper. It’s now embedded within product teams whose primary KPI is shipping speed. This is a classic principal-agent problem. When a product manager’s bonus depends on launching GPT-5 on schedule, how much weight will they give to a risk assessment that delays the launch?
From a quantitative perspective, the trade-off can be modeled. Let’s assume the safety team reduced the probability of a catastrophic failure by 0.5% per quarter. If the cost of such a failure (including regulatory fines, customer churn, and reputational damage) is $50B, then the expected value of the safety team is $250M per quarter. That’s a fraction of the revenue growth. But the risk is asymmetric: one major incident could wipe out 50% of the valuation. The market is not pricing that tail risk correctly.
Contrarian: The Market Overestimates the Impact of Departures
Here’s where my experience from the 2021 Sushiswap governance war comes in. I spent 72 hours straight tracking on-chain wallets during that governance dispute. I learned that high-frequency organizational changes often signal a pivot toward operational efficiency rather than a collapse. The same pattern applies here.
The departures of Dresser and Bakalar are not necessarily fatal. Dresser’s exit, for example, likely reflects OpenAI’s shift from a pure API sales model to a ChatGPT-centric subscription model. That’s a strategic repositioning, not a leadership vacuum. Similarly, the Preparedness Team disbanding could be interpreted as a sign that OpenAI believes its models are now safe enough to embed in product workflows—essentially, they are moving from a “safety theater” approach to a “safety by design” approach.
But the contrarian angle goes deeper: the market is fixated on the chaos, but the real risk is not the departures themselves. It’s the loss of institutional knowledge. The Preparedness Team had a unique understanding of the model’s failure modes. That knowledge doesn’t transfer smoothly to product teams. The probability of a latent safety bug being missed increases by an order of magnitude.
And let’s not forget the competitive dynamics. Anthropic is growing faster. Its revenue base is smaller, but its enterprise traction is accelerating. The 2025 corporate AI market is a zero-sum game for the first $100B in revenue. OpenAI’s organizational turbulence gives Anthropic a window to lock in high-value contracts with financial and healthcare clients who prioritize stability and safety.
Takeaway: The Next Six Months Will Define the Valuation
Three signals to watch:
- Revenue growth sustainability: If OpenAI’s quarterly growth rate drops below 30% (annualized 120%), the 25x P/S multiple will compress. The IPO will be postponed or priced lower.
- Anthropic’s enterprise wins: If they announce a major customer (e.g., JPMorgan or UnitedHealth), the market will reprice both companies. The gap between $1T and $250B valuations will narrow.
- Former Preparedness Team members’ destinations: If they join Anthropic, that’s a direct transfer of safety expertise. If they go to academic institutions, it’s a loss for the entire industry.
Governance is theater. Power is the script. OpenAI is rewriting its script at the worst possible time—right before the biggest IPO in tech history. The market will forgive a lot for growth. But it won’t forgive a safety catastrophe that could have been prevented.
Arbitrage closes the gap. You open the wallet. The question is whether you’re buying the hype or the fundamentals.
Speed is the only currency that doesn’t inflate. But it’s also the asset that devalues fastest when trust breaks.