Here’s what’s on our plate today:

• 🧪 OpenAI’s ad pivot and AI’s real business model.

• ⚡ Quick Bits, No Fluff: Meta’s AI sales, visas, space burials.

• 🧰 3 Things Worth Trying: Tools for ad-light, privacy-conscious AI use.

• 🗳️ Poll: Should ChatGPT show ads to stay mostly free?

Let’s dive in. No floaties needed…

The Laboratory

What OpenAI’s ad pivot reveals about the real economics of AI

Back in the late 19th century, when the internal combustion engine first appeared on the scene, it required significant capital expenditure, largely on factories, tooling, fuel networks, roads, and standards.

However, even with this capital influx, adoption was initially limited to well-capitalized players, and many pioneers failed before winners emerged. At the time, the cost curves looked irrational, and it was only decades later that the investments began to yield returns.

Even then, the biggest returns did not go to the investors, but to those who scaled the technology through manufacturing, infrastructure, and optimization.

In contemporary times, this trajectory is being closely mirrored by the AI industry. Only, this time it is happening at a much quicker pace.

Infrastructure before ROI

While the technology driving AI models is advancing rapidly, they still require data centers, energy, cloud infrastructure, and enterprise integration before they can deliver sustained ROI.

And since OpenAI has long been associated with the central figure in the world of AI, its revenue reports are of particular interest for many.

Recently, OpenAI Chief Financial Officer Sarah Friar shared the company's annualized revenue. According to her blog post, the company's annualized revenue surpassed $20 billion in 2025, up from $6 billion in 2024, with growth closely tracking an expansion in computing capacity.

She further shared that OpenAI’s computing capacity rose to 1.9 gigawatts (GW) in 2025 from 0.6 GW in 2024.

However, the most interesting piece of the announcement was neither the revenue nor the growth figures. Rather, it was that in 2026 the company planned to prioritize “practical adoption,” particularly in health, science, and enterprise.

OpenAI’s advertising pivot

While OpenAI has plans to focus on practical adoption of its models, the company has also announced it will begin testing advertisements in ChatGPT for U.S. users on free and low-cost tiers.

For an industry watching to see if AI can justify its spending spree, this development offers a critical data point about sustainable business models.

The decision reverses CEO Sam Altman's previous stance. In 2024, he called ads a "last resort" business model and described "ads plus AI as uniquely unsettling." By 2026, economic realities had shifted the calculus.

Despite achieving $20 billion in annualized revenue, Microsoft's SEC disclosures revealed OpenAI's quarterly losses exceeded $13.5 billion in the first half of 2025. The company operates with approximately 95% of users free, creating a fundamental mismatch between usage and revenue capture.

The advertising model should help to address this imbalance directly.

OpenAI's approach, internally labeled "intent-based monetization," will place clearly marked sponsored content at the bottom of ChatGPT responses when relevant products or services are available.

Premium tiers ($20/month Plus, $200/month Pro) remain ad-free. The company projects up to 20% of future revenue could come from advertising.

And this will not be the first time a major enterprise has relied on advertisements to sustain growth.

Google generates over 80% of its $200+ billion annual revenue from advertising, while Meta introduced ads to WhatsApp in 2025 to monetize its massive free user base.

The advertising pathway offers a proven path to revenue that could also subsidize future infrastructure costs.

Competing paths to monetization

However, advertising is not the only way for AI labs to monetize their products. Companies are testing fundamentally different approaches to the same economic challenge.

Anthropic has pursued enterprise-first revenue without consumer advertising. The company grew from roughly $1 billion to $4 billion in annualized revenue by mid-2025, driven primarily by API contracts with cloud providers and large organizations.

By positioning Claude as the safer, enterprise-friendly alternative to ChatGPT, Anthropic captured a large share of the enterprise market and reportedly achieved profitability before OpenAI, despite operating at a smaller scale.

Mistral AI represents a third pathway that questions whether infrastructure lock-in is necessary at all. The French company develops open-weight models that run on a single GPU, enabling deployment across hardware ranging from on-premises servers to laptops and edge devices.

For organizations with data sovereignty requirements or those unwilling to accept recurring API costs, Mistral's approach provides genuine technological independence.

CEO Arthur Mensch frames the value proposition around reliability and control: "Using an API from our competitors that will go down for half an hour every two weeks — if you're a big company, you cannot afford this."

While Mistral still offers commercial licenses and hosted API access, the open-source foundation ensures organizations don't become permanently dependent on the company's infrastructure.

However, while OpenAI is looking to advertising to increase revenue, it won't be easy.

The trust question

One of the major hurdles OpenAI faces is the trust deficits that advertising may amplify. Unlike the ICE, which had decades to prove its value, AI is moving so fast that consumer sentiment can shift within quarters rather than over generations.

Research data reveals significant skepticism around AI and advertising. Studies show that 57% of consumers globally believe AI poses a significant threat to their privacy, while 63% express concern that generative AI could compromise their personal data.

More concerning for OpenAI's strategy, research on AI-generated advertisements found that disclosure of AI involvement actually decreased trust in both the advertisement and the organization behind it.

This puts companies in a difficult spot. They are required to clearly say when ads are present and when AI is involved, but doing so often makes people more skeptical and less receptive to the message. OpenAI has to be transparent about its advertising while also hoping users trust the experience enough not to leave for other options.

At the same time, OpenAI says ads won’t shape ChatGPT’s answers and that user data will never be sold to advertisers. Those promises are meant to draw clear lines and build trust, but trust here isn’t granted once; it has to be earned every day.

One visible misstep could quickly confirm people’s fears and push users toward ad-free options like Claude or self-hosted solutions like Mistral.

Who ultimately wins

The internal combustion engine ultimately proved its value, but not on the timelines or in the ways its earliest backers expected. Capital flowed first into invention, then bled through years of infrastructure build-out, consolidation, and failure.

When returns finally arrived, they accrued less to those who built the engines than to those who made them reliable, affordable, and indispensable to everyday life.

AI now stands at a similar inflection point, only compressed into a fraction of the time. The technology clearly works, as OpenAI’s revenue growth demonstrates. But the turn toward advertising underscores a deeper truth: the economic model for large-scale AI is still unsettled.

Like early automakers experimenting with luxury sales, fleets, or mass production, AI labs are testing multiple paths to sustainability under intense cost pressure.

OpenAI’s advertising experiment is not a sign of weakness so much as a signal of maturation. It suggests the industry is moving from proving what AI can do to figuring out how it pays for itself.

As with the automobile, the ultimate winners may not be the model builders, but the companies that master infrastructure, integration, trust, and distribution. The engine is running. The road, however, is still being built.

Quick Bits, No Fluff

• Meta’s AI Pitch: CNBC says Meta is still struggling to turn its AI hype into paid enterprise deals as buyers stick with incumbent cloud providers.

• Visa Chaos Shift: Bloomberg reports H-1B chaos & higher costs are pushing top tech talent and U.S. roles to relocate to India instead.

• Ashes To Orbit: TechCrunch profiles a startup planning low-cost memorial flights, sending ashes of 1,000 people into orbit starting in 2027.

Thursday Poll

3 Things Worth Trying

• Mistral AI’s Le Chat / Playground: Use web chat or dev console to see what “run it on your infra” economics look like versus renting everything from a single lab.

• Kagi Search (subscription, no ads): Test-drive a search + AI assistant that lives off subscriptions instead of tracking and ads.

• Ollama (run models locally): If you want to escape both ads and cloud lock-in, this app lets you run open-weight models (like Llama, Phi, etc.) directly on your laptop or desktop. 

Rate This Edition