AI news · Friday, June 5, 2026

Google will pay SpaceX $920 million a month for compute and 110,000 GPUs

The AI gold rush is hitting a massive wall of reality as tech giants dump billions into infrastructure while scrambling to cut runaway costs. Google just committed to paying SpaceX a staggering $920 million every single month for compute power, including 110,000 NVIDIA GPUs, to keep up with the explosive demand for its Gemini agent platform. This massive spending spree is happening as firms like Anthropic—which is already paying SpaceX $1.25 billion monthly—push for a global pause on frontier AI development to allow societal guardrails to actually catch up to the technology. While leadership at these labs claims they want to slow down, investors see it as a self-serving move to lock in their lead just before their massive IPOs. Meanwhile, companies are starting to realize their 'tokenmaxxing' addiction is backfiring. Priceline and other major firms are seeing token usage explode by nearly 20x, forcing them to scramble for basic cost controls because their AI budgets are being vaporized by unpredictable agentic workloads. Some startups are even hacking their own bills by having employees use personal accounts to bypass massive enterprise surcharges. The industry is desperate for a common language to measure value, with the Linux Foundation launching the Tokenomics Foundation to try and bring order to this financial chaos.

Behind the billion-dollar boardrooms, the human side of the AI revolution is getting messy. Goldman Sachs is signaling that entry-level hiring will contract as AI begins to automate grunt work that historically taught junior staff how to think critically. While some experts argue this isn't a 'jobpocalypse,' the reality on the ground is that white-collar productivity is being fundamentally altered. Companies are struggling to manage these AI agents, with Microsoft CEO Satya Nadella publicly rebuking a VP’s plan to make their AI agent 'addictive,' insisting instead that agents need to be managed like employees with strict permissions and audits. Security is also becoming a nightmare as the AI gets smarter. Hackers are now moving beyond digital tricks, with gangs physically impersonating IT support to steal data, while others are exploiting Meta’s own AI support bots to hijack high-value social media accounts. Even the intelligence community is getting involved, with the NSA reportedly using Anthropic’s powerful Mythos model for offensive cyber operations against adversaries. As companies rush to launch these tools, they are discovering that moving fast often means leaving the door wide open for disaster.

The quick hits

Google is paying SpaceX $920 million monthly for massive compute access — this insane spending shows just how desperate the biggest tech giants are for chips.
Companies are seeing AI token bills explode by nearly 20x — businesses are in a full-blown panic as they realize their AI usage is burning money uncontrollably.
The NSA is reportedly using Anthropic's 'Mythos' model for offensive cyberattacks — this marks a major shift where frontier AI is now being weaponized by national intelligence.
Hackers are now using Meta’s own AI support agents to hijack Instagram accounts — this proves that AI agents are easy targets if developers don't build proper security guardrails.

Google will pay SpaceX $920 million a month for compute and 110,000 GPUs

The quick hits

Sources