Final week, OpenAI introduced that it has employed Fidji Simo, the CEO of Instacart, to turn out to be the CEO of its purposes division. From Reuters:
Instacart head Fidji Simo will step down to hitch OpenAI as its chief of purposes, CEO and co-founder Sam Altman stated on Thursday, including she’s going to report back to him and that he’ll stay head of the Microsoft-backed ChatGPT maker … The announcement comes two days after OpenAI dialed again a big restructuring plan, with its nonprofit dad or mum retaining management in a transfer that’s more likely to restrict Altman’s energy.
The information is noteworthy for a number of causes. The primary is that Instacart is a profitable enterprise: its inventory is up 34% on a year-over-year foundation, and the corporate beat analyst expectations on each income and earnings in Q1 2025. Whereas OpenAI lately raised cash at a $300BN valuation — in comparison with Instacart’s $11BN market capitalization — it’s nonetheless uncommon for the CEO of a publicly traded firm to decamp for a non-CEO function in an organization working in a very unrelated class.
The second is that it signifies that the enterprise fashions of Instacart and OpenAI are more likely to turn out to be rather more comparable. As I remarked when Instacart filed its S-1, Instacart is an promoting and subscription enterprise: the corporate’s promoting income progress outpaced that of transactions in Q1 2025 (14% vs. 8%, respectively), with transaction income capturing the Instacart+ subscription enterprise. Instacart operates a $1BN promoting enterprise that doubtless contributed disproportionately to the corporate’s $106MM of web revenue in Q1.
Fidji Simo was recruited to guide Instacart from Meta, the place she served as the overall supervisor of the Fb app and reported on to Mark Zuckerberg. However extra considerable than having run the Blue app on the finish of her tenure with the corporate, Simo is credited with having constructed the corporate’s income engine: advertisements within the cellular Information feed. As Antonio Garcia Martinez, a former product supervisor on the advertisements crew at Meta (then, Fb) remarks in his memoir, Chaos Monkeys (emphasis mine):
For all of the Sturm and Drang, Fb’s quarter-saving gold mine, the factor that catalyzed the inventory out of the post-IPO doldrums, wasn’t Customized Audiences or FBX. A 3rd product, the one different novel advertisements product Fb launched throughout its harried IPO interval, code-named “Neko,” was that savior. The product itself, like so many, was merely the mix of two in any other case disparate domains: Fb’s ever-addictive Information Feed and advertisements stock on the Fb cellular app, as an alternative of the desktop web site. That’s it: advertisements in Information Feed, whereas the consumer was on his or her cellular system—that’s what saved Fb. The individual most answerable for this coup was a product supervisor with the unbelievable identify of Fidji Simo. She was one in every of two workplace wives of mine (sure, I used to be an workplace Mormon), who had began her profession in Fb Adverts as a lowly product marketer. She had in a short time and elegantly navigated herself up the Fb company ladder, touchdown herself the product supervisor job the place Adverts and the remainder of the corporate overlapped, putting herself within the bigger Fb (and Zuck) highlight.
Again in March, in an interview with Ben Thompson, Sam Altman, OpenAI’s CEO, remarked that the corporate wasn’t at present severely contemplating promoting as a monetization technique:
At the moment, I’m extra excited to determine how we will cost folks some huge cash for a extremely nice automated software program engineer or different form of agent than I’m making some variety of dimes with an promoting primarily based mannequin.
However OpenAI was nearly actually within the final levels of negotiations with Simo on the time of the interview. And it appears unbelievable — frankly, inconceivable — that Altman might concurrently consider that promoting represented a chance sized at “some variety of dimes” and that Simo, who deployed some of the profitable promoting merchandise ever conceived, was one of the best individual to run OpenAI’s purposes division. These two notions are incompatible.
Clearly, OpenAI will monetize with advertisements. Hiring Simo represents such an on-the-nose acknowledgment of that proven fact that I nearly didn’t write this piece. Besides that OpenAI’s admission that promoting is its path ahead on monetization serves to dispel a typical false impression (actually, a fallacious, superstitious tech dogma) that promoting is however one in every of many monetization methods which are all equally able to attaining optimum income for scaled client know-how merchandise. This isn’t true. If maximizing income is a company’s goal perform, and its product can probably attain a scale of billions of customers, then promoting stands alone as its optimum monetization technique.
That is the crux of the freemium mannequin: a product with probably billions of customers ought to be given away totally free so as to attain them. Consideration is an asset. The product use instances able to agglomerating the eye of a whole lot of hundreds of thousands or billions of individuals are, nearly as a bi-product of their accessibility and broad breadth, doubtless not those greatest positioned to derive most worth from that spotlight.
Fortunately, promoting exists. Companies can bid on that spotlight: not within the combination, for a complete product’s consumer base, however for people, primarily based on what is thought about them. The asset of consideration might be atomized all the way down to a single consumer and offered to the product that may greatest monetize it. Promoting is an identical market that allocates consideration to the agency positioned to greatest activate it. I focus on this concept in Digital Promoting, Demand Routing, and the Millionaires’ Mall:
One problem with optimizing for conversions is that, in an ecosystem as huge and principally heterogeneous because the web — even within the context of particular, scaled merchandise — the presence at any given second of a consumer that 1) has an curiosity in some product and a couple of) possesses the disposable revenue to buy that product is uncommon. Digital promoting doesn’t create client demand; slightly, digital promoting ought to search to route present demand to the merchandise that greatest serve it. An environment friendly digital promoting channel matches client demand for a product with essentially the most satisfying and fulfilling variant of that product. An promoting channel is just not a requirement manufacturing facility however a requirement freeway: the extra effectively an advert channel can route client demand to merchandise, the extra financial worth it produces.
OpenAI will introduce promoting. It has constructed a generational product, ChatGPT, that has succeeded in aggregating consideration at a rarefied scale: the corporate revealed in April that it has 500MM weekly customers (WAU). However simply 4% of these customers are subscribers. The rest will probably be monetized with advertisements. This isn’t a design resolution; it’s an optimization train and a business necessity.