$150 billion in one year. 800 million devices with Gemini baked in. And Sam Altman casually mentioning trillions—with a T—in data center spending.

If you took the week off, you missed a lot. Here’s what actually matters.

The Money Is Real (Even If the Hype Isn’t)

AI startups raised a record $150 billion in 2025—63% more than 2024. But here’s what caught my attention: that money is concentrated in a handful of players. OpenAI alone pulled in $40 billion. Anthropic grabbed $13 billion. xAI secured $10 billion.

Everyone else? Fighting for scraps.

This isn’t “AI is hot” money. This is “we’re picking winners” money. And the bets are getting placed.

Morgan Stanley is now projecting AI will drive 20% of global economic growth this year. That’s not a prediction you make lightly. They’re talking about a $10 trillion corporate investment cycle.

Meanwhile, Nvidia has quietly become the most important investor in the space—67 venture deals in 2025, many of which include commitments to buy more chips. They’re not just selling shovels during the gold rush. They’re investing in the mines too.

The Real Infrastructure Race

Everyone’s talking about AI agents. But the real story? Power.

Tech giants added over $121 billion in new debt last year building data centers. Google just acquired Intersect Power for $4.75 billion—not for their AI tech, but for their renewable energy projects. OpenAI’s “Stargate” initiative in West Texas? $850 billion total project, 1+ GW capacity per site.

Meta’s “Hyperion” campus in Louisiana will use more power than New Orleans.

Read that again.

The constraint on AI isn’t talent or algorithms anymore. It’s electricity. The companies that control their energy pipeline will control the AI future.

Meta Makes a $2B Bet on Agents

Meta acquired Manus—a Singapore-based AI agent startup with Chinese roots—for over $2 billion. The deal happened in 10 days.

Why the rush? Because agentic AI is the battleground now. Not chatbots. Not copilots. Full autonomous agents that can execute complex tasks across platforms.

Manus technology is heading into Facebook, Instagram, and WhatsApp. If you’re building on those platforms, pay attention. The rules are changing.

Samsung + Google: 800 Million Devices, One AI Layer

Samsung announced they’re doubling their Galaxy AI footprint to 800 million devices in 2026. All powered by Gemini.

That’s not a partnership announcement. That’s a distribution moat. Gemini just became the most widely deployed AI model on the planet—baked into phones, tablets, and even refrigerators.

T M Roh, Samsung’s Co-CEO, said it plainly: “We will apply AI to all products, all functions, and all services as quickly as possible.”

This is what AI-native distribution looks like. Not apps. Infrastructure.

Claude Just Built a Year’s Worth of Engineering in an Hour

A Google Principal Engineer—Jaana Dogan—posted that she gave Claude Code a multi-paragraph prompt to build a distributed agent orchestration system. The kind of system her human team had been exploring for nearly a year.

Claude delivered a functional version in about an hour.

Not production-ready. But functional. In one hour.

If you’re running engineering teams and ignoring AI coding tools, you’re already behind. The economics of software development just shifted fundamentally.

Regulation: The Patchwork Problem

California’s SB 53 went into effect January 1st. Large AI model developers now need to publish safety frameworks, report critical incidents, and protect whistleblowers. Fines run up to $1 million.

But here’s the tension: the Trump administration just issued an executive order trying to preempt state-level AI laws. DeSantis is already pushing back, claiming Florida’s right to regulate independently.

We’re headed for a legal mess. 38 states passed AI-related legislation in 2025. No federal framework exists. Every company is now navigating a regulatory patchwork that makes GDPR compliance look straightforward.

If you’re in a regulated industry, this is your headache now.

The Grok Problem

I wasn’t going to mention this, but it matters.

xAI’s Grok got caught generating inappropriate images—including of minors. The company’s initial response to Reuters was literally “Legacy Media Lies.” They later admitted the problem.

Here’s why this matters for GTM: enterprise buyers in regulated industries won’t touch products with safety incidents like this. While everyone’s racing to ship features, trust becomes the real differentiator.

Anthropic’s bet on “do more with less” and disciplined safety investment? Looking smarter by the day.

The 2026 Shift: From Hype to Pragmatism

Every exec I talk to says the same thing: 2026 is about ROI, not experimentation.

Business Insider surveyed ten tech executives. The consensus: the era of AI pilots is ending. Companies want measurable productivity gains. They’re preparing for workforces where humans orchestrate AI agents, not replace them.

But there’s a warning buried in those predictions too—at least one foundational model company will face cash shortfalls this year. Market consolidation is coming.

The question isn’t whether you’re using AI. It’s whether your AI strategy can survive a correction.

What I’m Watching

Three patterns from this week worth tracking:

1. Vertical integration is accelerating. Google buying power companies. OpenAI building custom chips. The winners are controlling their entire stack.

2. Agentic AI is mainstream now. Meta’s $2B acquisition, Google’s Scheduled Actions feature, Samsung’s device-wide Gemini. We’ve moved from “talk to AI” to “AI does things for you.”

3. The efficiency vs. scale debate is real. OpenAI is spending trillions. Anthropic is focused on capability per compute dollar. Both strategies can work—but they require different execution paths.

The $150 billion that flooded into AI last year isn’t a bubble. It’s a rebuild. And the companies that understand that distinction are the ones making the bets that matter.

What’s the biggest shift you’re seeing in how AI is actually being deployed in your organization? I’m genuinely curious where the rubber is meeting the road.

GTM AI Podcast & Newsletter

The Week That Proved AI Isn’t a Bubble—It’s a…