AI Infrastructure Under Pressure: Data Centers, Crypto Heists, and Who Pays the Bill

The question of who actually pays for the AI boom is getting harder to ignore. From jet engines powering data centers to a record-breaking crypto theft, this week's tech news draws a straight line between AI's enormous appetite and the real-world consequences for companies, learners, and the broader industry.

Data centers are running out of power — and patience

Building the hardware backbone of AI is turning out to be much messier than the glossy announcements suggest. A striking report reveals that data center operators are now repurposing old aircraft jet engines as makeshift power sources because waiting years for grid connections and new turbines is simply not an option. These engines are less efficient and dirtier than conventional alternatives, but speed wins.

Meanwhile, Anthropic finds itself in an uncomfortable position. Even as demand for its Claude models grows, the company's compute costs are rising faster than its margins can absorb, which is a reminder that building frontier AI is an enormously capital-intensive bet with no guaranteed payoff.

On the other side of the ledger, AMD just reported its largest and most profitable quarter ever, with $13.4 billion in revenue driven heavily by data center chip sales. Its next quarter guidance points even higher. The company selling shovels in a gold rush is, predictably, doing very well.

OpenAI, perhaps sensitive to the growing environmental and community backlash around data centers, is reportedly lobbying to rebrand them as "AI compute campuses". The idea would be to frame these facilities as economic assets rather than energy-hungry warehouses. Whether the new name changes anything on the ground is a different question.

The $6.3 billion Binance hack explained

The largest cryptocurrency theft on record did not happen overnight. Binance has published a postmortem explaining that the attackers spent months working their way through the company's defenses, starting with a developer who downloaded a compromised installer promoted through a fake Google ad. The full chain of events shows how patient, methodical social engineering can defeat even large security teams.

Separately, cybersecurity researchers are pointing out that warnings about these exact tactics from North Korean hacking groups had been circulating for years before Binance was hit. For anyone studying cybersecurity, this is a textbook case in why threat intelligence needs to translate into actual operational changes, not just reports on a shelf.

The scale of the event is already reshaping investment: Ted Schlein's Ballistic Ventures is now seeking close to $490 million across two new cybersecurity-focused funds, betting that demand for AI-driven security tools will keep climbing.

CrowdStrike is also moving fast, announcing it will spend nearly $500 million to acquire two AI security startups — one focused on monitoring AI agents and another on automating security testing. As AI agents become a standard part of software stacks, the tools that keep them in check are becoming valuable very quickly.

Can you trust that an AI response is real?

A quieter but genuinely interesting development: Anthropic is building a system called FlashProve that would let anyone verify that an AI-generated response actually came from the model it claims to come from. In a world where AI-generated text is everywhere and impersonation is easy, being able to cryptographically confirm an output's origin matters more than it might seem at first glance.

Canva cuts staff as it restructures around AI

Canva announced changes affecting 22 percent of its workforce, including layoffs and role consolidations into central hubs. CEO Melanie Perkins framed it as a structural shift rather than a purely financial one. For anyone watching the design tools space, the Canva restructuring is worth following: the company has aggressively added AI features to its platform, and this reorganization likely reflects what it thinks its product will look like in the next few years.

Google puts schools in a difficult position

Google has updated its policy in a way that forces schools to make an uncomfortable choice: opt into AI tools or lose access to Google Search entirely. Educators who want students to use standard search without AI-generated answers now have to work around a policy that effectively bundles both. Experts quoted in coverage of the story raise concerns about student privacy and the academic integrity implications of AI in classroom search.

The parallel Ars Technica piece confirms that Google does offer a path to skip the AI features, but the trade-off is losing search access altogether, which is not a realistic option for most schools. This story sits at the intersection of edtech, AI policy, and digital literacy — all things that matter to anyone learning or teaching in a technical field.

A botnet hiding in plain sight

Google has taken down a botnet that had been silently running on 1.3 million Android TV boxes. The devices were compromised without their owners knowing, turning cheap consumer hardware into a distributed network for malicious activity. It is a sharp illustration of why supply-chain security and firmware integrity are not just enterprise concerns.

AI inference startup Fal reaches a $16 billion valuation

Fal, a five-year-old startup specializing in AI inference — the process of actually running AI models after they have been trained — has doubled its valuation to $16 billion after raising $1.3 billion. Inference is the less glamorous but more commercially intensive side of AI: every time someone uses an AI product, inference is what happens. As AI gets deployed at scale, the infrastructure that handles that workload becomes increasingly valuable.

Nexus Mods requires AI disclosure for game mods

The largest repository for PC game mods is now requiring creators to disclose whether AI tools were used in their work. The move reflects a broader debate in creative communities about transparency and attribution. For aspiring developers building games or tools, this is a useful signal: users increasingly want to know how software was made, and disclosure norms are starting to formalize across different platforms.

The coming months will push these tensions further into the open. AI infrastructure costs are not falling fast enough for some companies to stay comfortable, security threats are growing more sophisticated, and the platforms people use every day are quietly rewriting their rules around AI. Keeping track of where the pressure is building is a good habit for anyone building a career in tech.