Hi there,

While everyone's been watching the big model releases, this week delivered something more revealing: the cracks are starting to show in AI's foundation. From political pressure campaigns to infrastructure failures, October 21st marks a moment where the industry's vulnerabilities are becoming impossible to ignore.

The OpenAI-Microsoft Divorce Gets Messier

Reports emerged this week that OpenAI has been quietly downgrading Microsoft's role in its operations—and the numbers tell a brutal story. OpenAI will share just 8% of its revenue with Microsoft by 2030, down from 20% today. According to The Information, OpenAI leadership told staff that Microsoft "isn't moving fast enough" to supply the compute infrastructure needed for training massive models.

The pivot is already happening: OpenAI's $100-$500 billion Stargate joint venture with Oracle and SoftBank bypasses Microsoft entirely. They're shifting workloads to Oracle's cloud and exploring deals with regional providers. Meanwhile, Microsoft's $56 billion in 2024 capital expenditures are yielding solid Azure growth, but margins are straining under power-hungry GPUs—and AI data centers now consume 2% of global electricity.

My take: This isn't just a partnership dispute—it's a warning about the massive upfront costs AI demands for uncertain payoffs. When your biggest investor can't keep pace with your infrastructure needs, that tells you something about the sustainability of the current AI buildout.

Anthropic Under Political Fire

In a development that could reshape AI regulation, Trump's AI czar David Sacks launched a public campaign against Anthropic this week, accusing the company of "running a sophisticated regulatory capture strategy based on fear-mongering." The attack came after Anthropic's head of policy published an essay on "technological optimism and appropriate fear."

The political dynamics are stark: OpenAI has positioned itself as a White House partner since day one of Trump's second term, while Anthropic—founded on a mission to build safer AI—finds itself painted as the opposition. With OpenAI commanding a $500 billion valuation versus Anthropic's $183 billion, the political pressure adds another dimension to an already lopsided competitive landscape.

My take: This goes beyond typical startup rivalry. When government officials start publicly attacking AI companies for advocating safety measures, we're entering dangerous territory. Anthropic's entire value proposition is building trustworthy AI for regulated industries—exactly the approach that makes some politicians uncomfortable.

The AWS Outage Nobody Talks About

Yesterday's massive AWS outage disrupted Snapchat, Alexa, Venmo, Robinhood, Fortnite, Zoom, and dozens of other services—and it perfectly illustrates AI's infrastructure problem. The outage started around 7:40 AM BST due to an "operational issue" at a regional gateway in North Virginia. Recovery took several hours.

Here's the uncomfortable truth: as companies race to build AI infrastructure, they're all betting on the same small number of cloud providers. The centralization creates massive single points of failure. Elon Musk seized the moment to promote X Chat's "independence" from AWS—a reminder that when critical infrastructure fails, it becomes a political and competitive weapon.

The Reality Check on AI Employment

A Yale Budget Lab study released this week provides crucial context: AI hasn't yet caused widespread job losses since ChatGPT's 2022 release. Using labor market data through July 2025, researchers found that occupational shifts remain minimal compared to previous technological transitions.

Only 1% of services firms reported AI as the reason for recent layoffs (down from 10% in 2024), while 35% used AI to retrain employees and 11% hired more as a result. Critics argue companies are "scapegoating" AI to justify challenging business decisions—particularly those who "significantly overhired" during the pandemic.

What This Means for You

This week's developments point to three critical dynamics:

1. Infrastructure is the new battleground - The OpenAI-Microsoft split shows that compute availability matters more than model capabilities. Companies that control infrastructure hold the power.
2. Political risk is now existential - Anthropic's experience shows that AI companies can't stay neutral. Your political positioning matters as much as your technology.
3. The "AI replaced my job" narrative is overblown - Companies are using AI as cover for conventional restructuring, but the data doesn't support mass technological unemployment—yet.

The real question for 2026: Will infrastructure constraints and political pressures slow AI development, or will they accelerate consolidation around a handful of politically-connected players with deep pockets?

One more thing: As cloud outages become more common and AI systems become more critical, we're due for a serious conversation about infrastructure resilience. Yesterday's AWS outage was a preview of much bigger problems to come.

P.S. - Are we building AI too fast on infrastructure that's too fragile? Or is this just growing pains? I'd love to hear your take—hit reply.