The Great Unraveling Disguised as Business as Usual

---

The hacks are getting boring.

That’s the problem. When Russia’s military compromises thousands of consumer routers, when Iran-linked hackers disrupt US critical infrastructure, when someone steals $15 million from a currency exchange and blames “unfriendly states”—the headlines stack like email in an ignored inbox. We’re past the shock phase. We’re into the inevitability phase.

Meanwhile, the AI industry is having a very different conversation. Cerebras is filing for IPO on the back of deals with Amazon and OpenAI worth north of $10 billion. Tesla’s robotaxi service is now live in Dallas and Houston. The App Store is booming. By the metrics that matter in venture capital—growth, revenue, deals—everything is humming.

Except none of these things are as isolated as they look.

Photo by Vlad Bagacian / Pexels

The Math That Doesn’t Add Up

Here’s what strikes me: we’re watching the infrastructure that powers modern AI get systematically compromised while simultaneously betting hundreds of billions that this infrastructure is secure enough to control autonomous vehicles and critical systems.

Tesla has robotaxis operating without safety drivers in Austin, Dallas, and Houston. These are real cars on real roads making real decisions about acceleration, braking, and collision avoidance. The system depends on cloud connectivity, neural networks trained on massive datasets, and—critically—infrastructure that’s not being attacked by peer-state militaries.

Except it is being attacked.

When Russia’s military hacks consumer routers, they’re not just stealing bandwidth. They’re establishing persistent presence in home networks that could, theoretically, intercept or manipulate traffic from connected devices. When Iran disrupts critical infrastructure sites, they’re probing for weaknesses. When someone moves $15 million across borders through a sanctioned exchange using what amounts to state-sponsored theft, it’s not a crime. It’s practice.

I’m not saying Tesla’s robotaxis are about to be remotely hijacked tomorrow. I’m saying the threat model has fundamentally shifted, and the industry’s response has been to… price in the risk and move forward anyway.

That’s either bold or delusional. I haven’t decided which.

Photo by UMA media / Pexels

The Timing Problem Nobody’s Discussing

Anthropic just got designated a supply-chain risk by the Pentagon. This is not a compliment. It means the US government’s own security apparatus believes there’s a material risk to national security in how the company operates.

And yet—Anthropic is “thawing” its relationship with the Trump administration.

This is where the geopolitical game gets interesting. The Pentagon says you’re a risk. The administration says maybe we can work together anyway. Both statements are probably true. Defense institutions move slowly. Political administrations move fast. The gap between those two timescales is where actual leverage lives.

Here’s my read: Anthropic (and by extension, the entire frontier AI sector) is being offered a choice. Play ball with the administration’s vision for AI development, or face the full weight of supply-chain designations, export controls, and the slow strangulation of access to critical components and talent. The “thawing” isn’t reconciliation. It’s negotiation.

Meanwhile, Cerebras is going public while its chips are being integrated into Amazon’s data centers. That’s not just a win. That’s a statement. American AI infrastructure, American chips, American cloud providers, American profits. That narrative wins political allies fast.

But here’s what worries me: we’re building this entire stack—chips, cloud, models, applications—while under active adversarial pressure. The Q-Day references in the headlines aren’t about some distant quantum apocalypse. They’re about the current state where breakthrough capabilities in cryptography or model compression could render entire security architectures obsolete. We’re not ready for that. Not even close.

The Quiet Exodus

VMware migrations are happening because of “negative views of Broadcom.” That’s corporate-speak for “we don’t trust this company anymore.”

Broadcom owns VMware. And when thousands of enterprises start moving their virtualization workloads away from a vendor, it signals something deeper than technical dissatisfaction. It’s about control. About whether you want your infrastructure in the hands of a company that might be subject to different regulatory pressures, different ownership structures, different incentives.

This matters because it’s happening while Cerebras is going public, while AI infrastructure is consolidating around specific vendors, while geopolitical pressure is increasing. The smart money is diversifying. The dumb money is trusting that infrastructure concentration = efficiency gains.

I think the smart money is right, and the next three years will prove it. Companies with single-vendor dependencies in critical AI infrastructure will get squeezed. Not because of technical failure. Because of regulatory or geopolitical events that change the rules mid-game.

Photo by Denys Gromov / Pexels

The Personal Dimension

Ron Conway, one of Silicon Valley’s most active early-stage investors, just announced he has a rare form of cancer and is stepping back.

I want to be careful here. This isn’t a dunk or a callous observation. But it matters in context. Conway represents a generation of venture capitalists who built their networks during an era of relative stability. They made bets on founders, not on geopolitical posturing. That generation is aging out, and the one replacing it doesn’t have the luxury of ignoring nation-state actors, regulatory capture, or supply-chain warfare.

The AI founders getting funded today are inheriting a fundamentally different world than the ones Conway backed in 2005. They’re not just building products. They’re navigating between superpowers, managing their relationship with governments, and operating under the constant assumption that their infrastructure could be compromised or nationalized or sanctioned at any moment.

That changes everything about how you build, who you hire, and where you locate servers.

The App Store’s Deceptive Boom

New app launches are spiking. AI tools are fueling a mobile software boom.

Fine. But here’s what I want to know: how many of these apps are producing venture-scale returns? How many are just capitalizing on the AI hype cycle, building lightweight wrappers around existing models, and hoping for acquisition before the music stops?

The App Store boom is real. But it’s probably not the indicator people think it is. It’s not evidence that AI is entering the mainstream in a sustainable way. It’s evidence that there’s a gold rush mentality, and people are staking claims fast before the land fills up.

That’s cyclical behavior, not structural change.

My prediction: 60% of these new AI-powered apps won’t exist in 18 months. The ones that survive will be the ones solving genuine distribution or user experience problems that couldn’t be solved without AI. Everything else is noise.

What I’m Watching

1. Cerebras’ stock performance in Q2 2026 and AWS’s actual integration timeline. If the Amazon deal is real and delivering, we’ll see it reflected in usage metrics and revenue growth within two quarters. If it’s more of a strategic partnership meant to signal confidence, the actual shipment numbers will lag. Watch the earnings call language closely.

2. Whether a major robotaxi incident correlates with a coordinated cyberattack. This is morbid, but necessary. Tesla, Waymo, Cruise—any of them could experience a failure that originated from compromised infrastructure. The question is whether the industry has the forensic capability (and transparency) to actually detect it.

3. The next round of VMware migration patterns. Which vendors are absorbing the refugees? If they’re consolidating around open-source platforms or edge infrastructure, that suggests enterprises are intentionally de-risking vendor concentration. If they’re just moving to the next big vendor, it’s theater.

4. How Anthropic’s regulatory status actually changes under the thawed relationship. Will they get carve-outs from export controls? Will they need to integrate US government requirements into their model development? Watch for quiet policy shifts, not press releases.

The infrastructure’s under attack. The industry’s consolidating. The geopolitical game is tightening.

And everyone’s acting like we’re still in a winner-take-all tech boom.