Is it time for a SaaS comeback?
Welcome to Cautious Optimism, a newsletter on tech, business and power.
📈 Trending Up: French AI? … AI efficiency … layoffs … understanding how LLMs work … Sam Altman doubters … hearing … idiocracy … working the refs …
Post of the Day: The Times has a look at how local, state, and Federal regulations made it difficult to build chip fabs in Arizona. I think even the most e/acc folks will find something to like in the reporting.
Quote of the Day: “An absolute ding-dong,” via Jane Coaston on X, discussing Curtis Yarvin (image via this New Yorker story).
📉 Trending Down: Agentic startups … PC gaming … House control … taking it sitting down … EVs in the United States … Meta’s VR losses …
Things That Matter
Billionaire sprawl: When Amazon founder Jeff Bezos announced that he was helping build a new startup called Project Prometheus, CO wrote the following:
Forget private jets, sprawling yachts, or merely owning your own space company. Real billionaires these days fight to control their own nation-state-sized empire of companies. Double the points if the collection is synergistic, akin to Musk merging X and xAI, or Bezos investing in AI that could help his other efforts.
Enter today’s news that OpenAI’s Sam Altman considered getting into the space game, using his AI company’s wealth to buy a majority stake in Stoke Space, a launch startup building rockets with reusable first and second stages. (SpaceX pioneered the reuse of a rocket’s first stage, landing it back on the planet after liftoff. Blue Origin has managed the same feat, and some Chinese launch companies are nipping at American heels.)
Why would Sam Altman want to own Stoke Space? Presumably, the same reason his former OpenAI co-founder, Elon Musk, is building a competing AI lab to OpenAI: competitive pique.
I don’t mind the world’s super-wealthy spending their wealth on space companies; in fact, I encourage it. Still, you have to wonder about the power of focus. OpenAI probably wouldn’t do well as an umbrella corp for space launches because why the fuck would it. Better AI models, please, Sam.
Selling chips to China: Much of 2025 has been spent trying to sort out if American chip companies can sell to China, what type of chips they will be given permission to ship to the nation, whether or not Chinese companies will be allowed to use non-domestic chips, and what price the Trump administration would exact for the right of doing business.
The answers, today, are only maybe, the less powerful type, probably not, and 15%, respectively.
But another day means another twist in the journey, and the latest news comes from the FT, which writes that POTUS et al are “preparing to hold a high-level meeting to decide whether to provide licences to allow Nvidia to export the H200, an advanced chip, to China.”
The H200 is an upgrade to the now prior-gen H100 chip, expanding its high-bandwidth memory from 80 gigabytes to 141 gigabytes, while also boosting its memory bandwidth by about 1.5 terabytes per second.
Hot damn. Will China, which has cracked down on foreign chip purchases and begun subsidizing the use of domestically-produced data center chips, allow for its AI companies and cloud providers to buy H200s? “We don’t know,” says Nvidia CEO Jensen Huang, “we have no clue.”
If the US does allow for H200 sales to China and if the customer nation allows their purchase then Nvidia’s growth rate in calendar 2026 could cook for yet another trip ‘round the sun. Else, Nvidia will have to content itself to the US market and those of its allies.
Small business in shambles: Yesterday, ADP reported a loss of 32,000 private-sector jobs in the United States. The headline figure blends data from companies of all sizes, of course, which means that we might find something interesting if we look at disaggregated data:
Companies between 1 and 49 employees: -120,000 jobs in November
Companies between 50 and 499 employees: +51,000 jobs in November
Companies with more than 500 employees: +39,000 in November
While the aggregate picture of domestic job growth has dimmed this year, smaller companies are leading the charge when it comes to shedding staff. This is what headline reporting on massive cuts at Amazon, Verizon, and other companies misses: the most painful jobs carnage is happening out of sight.
Meanwhile, the share of adults aged 22-28 who live with their parents is spiking to all-time highs (removing a short pandemic-driven spike), while the job market for recent college graduates is consistently weakening. Not precisely what we want to see.
Is it time for a SaaS comeback?
Let’s take stock of SaaS earnings this week:
On Tuesday, cybersecurity company CrowdStrike beat both revenue and profit expectations; identity software company Okta posted a top-and-bottom-line beat, but declined to guide for its next fiscal year; data and AI company Box beat revenue expectations while meeting profit forecasts; team productivity software concern Asana surpassed both revenue and profit expectations while raising guidance; developer tool company GitLab posted better-than-expected revenue and adjusted profit, but missed on GAAP incomes.
On Wednesday, CRM giant Salesforce posted a profit beat and a tiny revenue miss, while guiding strongly for the current quarter; Databricks rival Snowflake beat on revenue and profit metrics, and guided higher than the market expected (it also announced a major Anthropic deal); workflow automation company UiPath beat both profit and revenue expectations, and guided above expectations.
Quite the mix, yeah? Despite the soup of results, I think it’s fair to say that we’re seeing SaaS companies best market expectations today, even if their resulting share-price movements aren't always as responsive as the pertinent CEOs might have hoped.
Looking ahead I’m even more bullish; we could be in the early innings — quarters? — of a SaaS comeback.
Let’s examine two companies, Box and Salesforce, to learn a bit more about their growth and what they see ahead:
Box’s revenue growth was 9% in its most recent quarter, with the company guiding for another 9% in the current period. At the same time, Box has seen its billings rise from $242.3 million in the first quarter of its fiscal 2026 (three months ending April 30, 2025), to $264.9 million in the following quarter and $296.0 million in its most recent report.
What’s driving those gains? Per Box’s Dylan Smith, the company’s “growth is being fueled by strong customer demand for Box AI, resulting in a pronounced upgrade cycle and longer contract durations.” Put another way: Box’s work to bake AI into its product from top to bottom is having real impact.
Even more, Box has a new product called Extract coming out, which “uses agentic reasoning […] to understand documents and extract information.” For Box, a company that cut its teeth offering cloud storage, the ability to turn that historical work into a new, competitive edge in the AI era is downright exciting.
Box was once written off as a slow-growth, post-IPO unicorn dinosaur. Instead, it’s worked its tail off to build out its enterprise product set, and is now levering every ounce of AI it can find to earn perch in the market for the next decade. Even more, it’s doing a good job of bringing AI into its world — evidence: its logo gains and better-than-expected NRR — allowing it to grow faster and remold the narrative concerning its business back to positive.
So much for dead.
Then there’s Salesforce. The CRM concern has bought so many companies over the years that to list all its appendages and their discrete results would bore you. That in mind, here’s the lean meat we need today:
Salesforce raised guidance for its current fiscal year from “$41.1 billion to $41.3 billion” good for 8.5% to 9% growth, to “$41.45 billion to $41.55 billion,” good for 9% growth to 10% growth.
While the growth gains are predicated on several factors, one stands out when digesting Salesforce’s earnings: AI.
Salesforce reported that its “Agentforce [an AI service] and Data 360 products are the momentum drivers, hitting nearly $1.4 billion in ARR—an explosive 114% year-over-year,” with Agentforce reaching 9,500 paid deals (up from “over 6,000” in the preceding quarter).
Even more, “Agentforce ARR surpassed half a billion in Q3, up 330% Y/Y.”
There’s a vibe in the Valley that AI-first companies will eat legacy software companies (SaaS) that are AI-second by definition of having built their businesses in the pre-LLM era. That may come to pass. But while we wait for those predictions to bear out or not, Box and Salesforce are showing that old-school SaaS concerns still have teeth.
Thanks to a history of holding customer data, both SaaS companies are working the night shift to apply AI to those datasets, turning at-rest data into in-motion sales. In response, we’re seeing green shoots in their earnings reports.
The AI incomes in question are not rewriting the software business landscape. Yet. But the signs are increasingly positive.
Box and Salesforce are not the only examples of pre-LLM, public SaaS companies finding their footing in the AI era. For example, Asana said that its AI Studio product “delivered another good quarter, with solid growth in sequential bookings, including early traction with self-serve users,” adding later in its prepared remarks that it is seeing strong positive feedback from our initial set of 30 beta customers” for its agentic AI Teammates product that will become “generally available early next year.”
Some SaaS companies will get collapsed by AI-first startups that build better products. But there’s enough momentum from legacy software companies in building AI into their services that it may very well turn out that the AI-first CRM giant of the future is Salesforce, that Box is the AI-first ‘Intelligent Content Management’ as it likes to self-brand, and Asana is the AI-first team productivity software that the market wanted all along.
If that’s the case, perhaps some of those SaaS unicorns from the 2020-2021 era have a better shot at life than we thought. Viva la SaaS.