
Platformonomics TGIF is a weekly roll-up of links, comments on those links, and perhaps a little too much tugging on my favorite threads.
All AI through-line this week.
News
Did OpenAI Find the Money This Week?

I interpret that as

Platformonomics TGIF is a weekly roll-up of links, comments on those links, and perhaps a little too much tugging on my favorite threads.
All AI through-line this week.
News
Did OpenAI Find the Money This Week?

I interpret that as

OpenAI subsequently walked back the comment, offered a blizzard of alternative and unsatisfying words, and incidentally broke the AI-driven stock market. After disavowing both their desire for government assistance (as did the government) and the prospects of an IPO anytime soon, the latest OpenAI mantra is revenue growth alone will pay for their trillion dollar plans. They intend to build vast new consumer, enterprise, platform, consumer hardware, robotics, and even cloud infrastructure businesses. They will succeed in so many businesses!
Previous:
The OpenAI vs. Meta Slopfight
[Epistemic confidence: 75%]
I’m embracing the theory that, at the topmost level, OpenAI is trying to become Meta before Meta can become OpenAI. (The Albanian army is not involved). This theory has high explanatory value for both companies. As much as Sam would like us to believe OpenAI can build multiple huge and disparate businesses against a broad range of competitors, this is the primary front to watch, and the one they most have to win.
OpenAI desperately needs revenue (see above and previous) and monetizing their 800 million weekly average users is the fastest and easiest path to material incremental revenue. Ads, engagement and AI slop here we come! Every ChatGPT response already ends with an engagement baiting “Would you like me to…”.
Keeping you scrolling, of course, is Meta’s franchise (with, we learn, a healthy layer of fraud on top). Meta recognizes the existential threat, which explains their response: a combined AI mulligan and drunken spending spree. It also helps explain Meta’s confusing statements on whether AI unlocks new businesses for them or just fuels (defends?) more of the same.
OpenAI has hired a lot of Meta people, and increasingly resembles Meta. Others have analogized OpenAI to Windows, but I think Meta is the better focal point, and certainly more likely. OpenAI even has the same chip on their shoulder as Meta against Apple, as both feel driven to replace the smartphone as the primary personal device form factor.
Expect to hear a lot of noble (but probably not Nobel) language about how AI will automate scientific breakthroughs, if only to distract from the slopfest reality. Automating science is replacing AGI when it comes to platitudinous OpenAI marketing (while Anthropic’s approach is to periodically remind us how weird they are). Just like Facebook used to talk about how social media would bring humanity together and make the world a better place.
Previous:
The AI API Battle
You can’t spell API without AI, but the reverse would be more meaningful (i.e. API access is crucial to AI). The battle to be the dominant provider of AI via API is underappreciated relative to the more visible consumer battle.
It is an important element of the reconfigured Microsoft-OpenAI relationship, and is the fundamental means for both enterprise and platform (basically software company consumption) AI consumption.
Sam and Satya’s joint interview last week was muddled on the API relationship, and the companies’ respective statements are not entirely consistent. OpenAI said:
OpenAI can now jointly develop some products with third parties. API products developed with third parties will be exclusive to Azure. Non-API products may be served on any cloud provider.
The OpenAI API is exclusive to Azure, runs on Azure and is also available through the Azure OpenAI Service.
I read this as OpenAI’s enterprise and platform API ambitions go through Azure for five years (or until AGI 🤡).
API consumption doesn’t usually happen in a vacuum. There is often co-located application logic, which requires old school compute, storage and other cloud services. (This is a reality that the AI training neoclouds will discover when they go beyond trying to convince Wall Street they can “pivot to inference” to extend the useful life of their GPUs and actually try to “pivot to inference”).
This also suggests the recent OpenAI-AWS announcement really is a training cluster deal. No API, so no help to AWS in winning incremental workloads built on OpenAI. Nor any Trainium or AWS networking. AWS looks like any crypto bros turned neocloud to OpenAI.
The next question is how is OpenAI doing on the enterprise and platform fronts? They’re planning on huge revenue from these markets to fund all those gigawatts. One survey suggests they are losing share:
This is a small sample and perhaps biased source, but it feels directionally correct to me. Anthropic is more focused on API delivery and enterprise customers. The Information says Anthropic “forecasts that sales to businesses through its API and related apps will continue to generate more than 80% of its revenue through 2028”. Anthropic’s profitability path narrative is also far less ridiculous than OpenAI’s (and just relatively less ridiculous goes a long way).
When you care more, you care more, and Anthropic cares more about API delivery. It is really, really hard to win all the battles and also consistent with OpenAI being a consumer-first company.
Everything Bagel Tariff Policy
The best strategic articulation for Trump’s tariff policy was as a response to Chinese mercantilism, even if tariffs were a crude and indirect way to address those imbalances. The “man with hammer” application of tariffs on countries like Brazil, Canada, and India made a mockery of this. Now the recent “deal” with China leads one to conclude either China was not the focus of the strategy or the strategy has failed. The Chinese mercantilism problem remains.
(Tariffs: another word you can’t spell without A and I — got to maintain the through-line).
Previous:
Orbital Computing
We write a lot here about data centers you can see from space, but data centers in space have become a hot topic lately.
This is an area I’ve been interested in for a while (but characteristically haven’t written about). It is really more of a networking build-out today (and we’re definitely still in the pre-internetworking era), but computing will follow. Because data may have gravity, but latency is a bitch.
(I haven’t written the massive explanatory post on orbital computing, but at least I have now landed my best two jokes on the topic).
Previous:
Not Available in Europe: The EU AI Act?
Chortle.
The EU is nothing without its regulations.
Previous:
Not Available in Europe: Laptop Chargers, Not Available in Europe: Online Political Ads, Not Available in Europe: Apps in ChatGPT, Not Available in Europe: Single-Sided Printing, Not Available in Europe: Data, Not Available in Europe: AirPod Live Translations Edition, Not Available in Europe: Email Archives, Not Available in Europe: A Tech Champion, Not Available in Europe: Meta AI, Not Available in Europe: Spinal Fortitude, Not Available in Europe: Air Conditioning, Not Available in Europe: Working Through the Not Available in Europe Backlog Edition, Not Available in Europe: WhatsApp Edition, Not Available in Europe: ChatGPT Memory Edition, Not Available in Europe: Bird Flies, Not Available in Europe: Open AI Deep Research, Not Available in Europe: A Bubbly Economy, Not Available in Europe: Better Late Than Never Apple Edition, Not Available in Europe: Apple’s Math, Not Available in Europe: This Week’s Edition, Not Available in Europe: Coming Soon to a Continent Near You, Not Available in Europe: Don’t Say You Weren’t Warned (by Meta), Not Available in Europe: A Tipping Point?, Not Available in Europe: Stratechery Edition, Not Available in Europe: Apple Edition, Not Available in Europe: Meta Edition, The Cookie Continent Capitulates?, GDPR Deathwatch? Or More Empty Words?, EU Insanity: Peak EU AI Regulation?, Existential Corner: EU Agrees It Faces ‘Existential Crisis’, EU and What Army?, EU Insanity: Thierry’s Termination Tizzy, EU Insanity: Experts Agree, EU Insanity: EU Goes Too Far, Even for EU, Move Fast and Regulate Things: Welcome to the Morning After, Move Fast and Regulate Things (You Don’t Understand), When “Move Fast and Regulate Things” Breaks Down, AI Regulation: Move Fast and Regulate Things, EU Insanity: Regulating Blue Checks, EU Tweets While Ukraine Burns, EU Insanity: AI Regulatory Suicide, EU Insanity: Mistral Edition, The EU Will Continue to Fiddle While Rome Burns, EU Insanity: AI Energy Suicide, EU Insanity: AI Energy Suicide (Part Deux), The European Union is STILL an Advanced Persistent Threat, BREAKING: European Union Designated an Advanced Persistent Threat, ClownWatch™: The European Union, Attack of the EuroStack, The Fate of the European Economy: Cloud Edition, Bonsai AI