The return of the press release
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Wednesday, friends. Last night, low-power AI chip concern Ambiq priced its IPO at $24 per share, towards the upper end of its $22 to $25 range. While the world waits for Figma to price its IPO after the bell today and commence trading tomorrow, Ambiq’s IPO is a foretaste of the feast to come. While it failed to price above its range, the company sold 600,000 more shares in its debut than initially anticipated. That’s a win.
In economic news, US GDP growth in the second quarter was 3.0%, better than expected (2.3%). Consumer spending also picked up, gaining 1.4% in Q2 2025. Recall that GDP contracted by 0.5% in the first quarter.
Not to be pedantic, but when first-quarter GDP growth came in low, some boosters of the present administration pointed to Final Sales to Private Domestic Purchasers (FSPDS) as a better metric to track the private (non-governmental) market. As that figure was +1.9% in Q1 2025 instead of -0.5%, you can see why.
So, what happened in the second quarter? While GDP growth (headline) was 3.0%, FSPDS was a smaller +1.2%. Expect the people who ditched GDP in the wake of Q1 data to about-face today. Just keep them honest. — Alex
📈 Trending Up: GPU markets … earnings results … AI usage in media? … Teramount … global trade barriers … private payrolls in July … in-kind ETFs … the value of the written word? … encryption … tsunami warnings … EU AI rules … Ramp’s worth …
📉 Trending Down: Push notifications … US-India relations … lmao … Tea, again … Russia-US relations … the value of IP? … investor contentment after Palo Alto Networks’s CyberArk deal came real …
Nibbles
Everyone wants to help* the kids
Smartphone bans in schools. Worries that the global fertility crunch is being driven by too much screen time and too little time spent chatting IRL. Youth who aren’t dating as much as they once did. Chatter that the kids these days don’t want to learn to drive.
As we’ve made interacting with the digital world ubiquitous, older people are worried that younger people might not be engaging with meatspace as much as they should.
Enter a host of tweaks to social media services and local rules regarding who can access online services, and at what age they can begin. Sure, a social media platform that makes its bread by serving ads claiming that it wants to help kids use its service less is probably lying, but the trend is now more than material:
TikTok has new tools for parents, including blocking their children from interacting with specific accounts.
YouTube intends to estimate user age in the United States, “in order to provide a more age-appropriate experience,” TechCrunch reports.
The UK internet has become ring-fenced by age verification rules, which are naturally creating a growth market for VPNs.
Australia is adding YouTube to its list of services banned for teens under its new social media rules.
After reportedly disrupting the entire educational system, OpenAI is working on a ‘study mode’ for ChatGPT.
The problem with all the above is that we don’t let kids roam. So, they are at home. If we decide as a species that phone and social apps are a problem to be solved, simply blocking their use is only part of the puzzle. We’ll also have to do something about this.
Look, Ma, revenue
Cohere: After reaching a $70 million run rate in February, Cohere expects to reach a roughly $200 million run rate by the end of the year. It hopes that’s enough to raise more capital.
Groq: Capacity delays could cut inference provider Groq’s expected revenue in 2025 from $2 billion to $500 million. Still hefty, but far smaller.
What cost my data?
If you buy a game, do you own it? If you digitally purchase a song, are the rights yours forever? What about your saved game data? Your user profile? The questions get even stickier in a business setting: Do you own the data that Salesforce holds? What about your bank?
These are not idle questions for consumers buying digital products, nor are they small queries for companies that store much of their proprietary data inside cloud services worldwide.
That data is more valuable than ever in the AI era is a fact, but a more traditional business model is under threat thanks to attempts by holders of customer data to charge third parties for access to that data at the behest of the customer in question.
For example:
JPMorganChase wants to charge fintech companies like Plaid that access customer data for the customer so that they can connect products and services to their banking information.
Salesforce is blocking the use of Slack data in certain AI contexts.
Celonis sued SAP earlier this year over claims that it placed “restrictions on [customer] data access to prevent Celonis from competing with SAP’s own process mining company, Signavio.”
Strava tightened its own data policies in late 2024, limiting the export of user data and blocking AI training.
Each case is distinct, but the general trend is towards data-holders looking to yank revenue from the information in question, even if it belongs to or was generated by the user. In the case of JPMorganChase and the fees it wants to charge fintech companies, the risk is that the smaller companies’ business models become defunct when facing a huge increase in their cost of revenue (or: a collapse in their gross margins).
Salesforce’s actions paint a picture of a future in which startups struggle to win over customers because incumbents build walls around customer data and, thus, market access. SAP falls into the same bucket, while Strava is a smaller case.
Sounds like we need some sort of ‘data bill of rights’ for customers and companies alike.
The return of the press release
Yesterday, while chatting with a close friend — cough — we discussed press releases. Anyone versed in the journalist-PR debates of yesteryear knows that the venerable press release is thought to have gone the way of the manual toothbrush. A relic. Something past-tense. A holler from before.
But despite the fact that PR folks don’t like to write press releases and journalists love to ignore them, they have thus far failed to die. Hell, I linked to the Ambiq press release above, to pick a recent example.
More than simply dying, however, it appears that we’re on the precipice of a press release renaissance. Why? Simple:
Companies make lots of news, and want people to know about it.
Press releases were once a good way to get human information aggregators and disbursers to help spread your news.
That’s no longer the case, but artificial information aggregators and disbursers are hungry for information, and face an increasingly walled-off internet.
AI companies do not want to pay for access to information to answer user queries, as paying for data would harm their future profits.
But users want to ask AI questions, so where can AI models source the needed information to respond to user queries?
The bevy of freely accessible press releases on the Internet, baby!
You can’t read most of the hard news on the Internet today without a subscription. But PR wires? Free and all-you-can-eat.
So! As startups are formed to help companies rank more prominently in AI search results and folks worry about the end of the SEO era, what’s a company to do? Release a lot of press releases that contain a lot of approved information. From that moment forward, AI models will ingest and/or ping those releases for information, data that just happens to come from a particular company!
Reach, sans the need to get a stinky human to actually write up whatever it is your company did most recently.
The second-order effects of press releases becoming more important in the AI era than they were before should include:
Longer press releases (why not share more information, and thus earn potentially more inclusion in AI search results?)
Press releases tuned for robots, not humans (why write for an audience that isn’t your audience?)
Bolder claims, as there won’t be a layer of human skepticism between the company posting news and the audience that may consume it.
I suppose this means that the PR game just became more valuable. Godspeed, everyone working in an agency who now has to write more releases. Perhaps AI can help?