Handicapping the SpaceX IPO - Cautious Optimism

Friday. The national argument about AI regulation is not over, though it has receded from headlines in recent weeks (more below on whether or not the Strait of Hormuz is open or not). The White House is leaning on Republican states to can efforts to regulate the burgeoning artificial intelligence industry, while OpenAI is supporting an Illinois law that would, per Wired, "shield AI labs from liability in cases where AI models are used to cause serious societal harms."

If the OpenAI move raises your hackles, relax. We want AI labs to release their new technology with reasonable alacrity, so limiting usage risk by third-parties is critical. Either we continue to accelerate, or the next Mythos-level breakthrough may not be ours to hold back.

Don't worry, we're not only talking about AI today. Coming up: Why the market is still discounting the value of software, corporate spend on tokens, the Strait of Hormuz (and CPI data!), and we'll close with notes on SpaceX's post-merger profitability. To work! -- Alex

Software pessimism: One cloud ETF is off 50.0% over the last five years (and 20.3% lower than its year-ago price) while a different, broader software ETF is at its lowest ebb since late 2023. Concerns about the future of many software companies in the AI era continue to weigh on the value of many public and still-private software concerns.

The new bad news for software stocks is another wave of analyst downgrades, this time with Citibank tossing demerits at Veeva, DocuSign, SimilarWeb, and others. From Yahoo Finance's Brian Sozzi, Citi sees rather poor vibes continuing in the SaaS space (emphasis original):

As we look ahead to Q1 earnings-and-beyond, amidst parabolic AI revenue inflections at foundational model leaders, we outline updated views on software. Put simply, we see risk that concerns around software application architecture, business model durability and terminal value intensify in the months ahead. Privately held AI companies are on track to add $100 billion+ of net-new revenue in the years ahead, materially eclipsing the $50 billion of traditional application software NNACV. While AI budgets may be mostly additive vs. replacement to software, checks indicate an uptick in software optimization costs/vendor consolidation

Related food for thought (concern) from venture capitalist Nichole Wischoff:

Tokenmaxxing: Ramp, the popular corporate spend and banking unicorn, has a new product out that helps companies track their AI spend. Given model subscriptions, API spend, tool testing, and more, companies may not know how much they are spending on AI. Ramp wants to help, naturally.

More salient for our purposes, the following chart shows the rate at which Ramp customers (lots of startups, sure, but the company is big enough now to sweep a portion of Main Street under its data-wings, too) spend on tokens:

Something fascinating in the above chart is that you can see the late-2025/early 2026 model improvement zone (GPT-5.2-Codex, Opus 4.5) clearly. Hell, even the end-of-year holiday period could only blunt token spend growth, not reduce it. Perhaps Nvidia's 'AI Factory' pitch was more correct, and ahead of its time, than many thought (though CO was generally bullish).