Over the next year, Roku predicts that 100% of the streaming audience will see ads. For growth marketers in 2026, CTV will remain an important “safe space” as AI creates widespread disruption in the search and social channels. Plus, easier access to self-serve CTV ad buying tools and targeting options will lead to a surge in locally-targeted streaming campaigns.

Read our guide to find out why growth marketers should make sure CTV is part of their 2026 media mix.

Learn more.

Here's what you need to know: Amazon just announced they're dropping $50 billion—that's billion with a B—to build AI supercomputing infrastructure exclusively for the U.S. government. And if you think this is just another tech company throwing money at data centers, you're missing the point entirely.

This is AWS building an unbeatable moat.

Here's the play. The federal government is sitting on a massive AI adoption problem. They've got 11,000 agencies that need to process classified intelligence, run defense simulations, and analyze satellite imagery—tasks that currently take weeks but could take hours with AI. The catch? You can't just spin up ChatGPT on a government laptop and call it a day. This stuff requires Top Secret and Secret-level infrastructure that literally no one else has at scale.

AWS already dominates government cloud with roughly 65-70% market share. Now they're adding 1.3 gigawatts of computing capacity—enough to power 750,000 homes—across three security tiers: Top Secret, Secret, and GovCloud. By the time this infrastructure goes live in 2026, the switching costs for government agencies will be astronomical.

The timing is perfect. The White House just released its "AI Action Plan" explicitly calling for domestic AI infrastructure to maintain dominance over China. Federal AI spending is projected to grow from $2.7 billion in 2026 to $3.1 billion by 2028—but that's likely understated given this infrastructure buildout suggests the real number could hit $10 billion annually by 2030.

What's brilliant is AWS isn't just selling compute power. They're deploying their proprietary Trainium chips alongside NVIDIA GPUs, which gives them 30-40% better price-performance and reduces dependency on NVIDIA's constrained supply. They're pre-loading the infrastructure with pre-authorized AI models like Anthropic's Claude and Amazon's Nova—all FedRAMP and DoD certified. Every layer of this stack makes it harder for Microsoft Azure or Google Cloud to compete.

The payoff timeline? Probably 8-12 years to break even on the $50 billion, which is perfectly acceptable for infrastructure with a 20-30 year useful life. But the real return isn't just revenue—it's becoming the canonical platform for U.S. government AI across multiple administrations. That's the kind of strategic position you can't buy at any price.

Microsoft and Google now face an uncomfortable choice: match AWS's $50 billion bet or concede the government AI market entirely.

Most will choose the latter.