The competition in AI is getting intense-and expensive. Google is stepping in with a huge commitment, pledging up to $40 billion in cash and computing power to Anthropic as companies race to secure the resources needed to build cutting-edge AI.

What’s happening:

Google is putting in $10 billion upfront, valuing Anthropic at around $350 billion. On top of that, another $30 billion could follow but only if Anthropic hits certain performance milestones.

At the same time, Anthropic has quietly rolled out Mythos, its most advanced AI model so far, though it’s only available to a small group of partners for now.

Investor interest is also surging. Reports suggest Anthropic’s valuation could climb as high as $800 billion, with a potential IPO possibly arriving as early as October.

What’s going on behind the scenes:

This deal comes during a hectic stretch for Anthropic. Users have been frustrated with usage limits on Claude, highlighting just how stretched the company’s computing capacity has become.

To keep up, Anthropic has been racing to secure more infrastructure. It recently partnered with CoreWeave, raised another $5 billion from Amazon, and even committed to spending up to $100 billion to access roughly 5 gigawatts of compute power.

Now, Google is adding even more firepower another 5 gigawatts of TPU-based capacity over the next five years. This builds on an earlier agreement involving 3.5 gigawatts of capacity tied to Broadcom, expected to come online around 2027.

Why this matters:

In today’s AI landscape, the biggest bottleneck isn’t ideas it’s compute.

Anthropic’s massive spending (over $140 billion across deals with Amazon and Google) shows just how costly it is to stay competitive at the frontier of AI. Interestingly, Google is effectively funding a rival, likely because the alternative is losing influence if Anthropic relies entirely on other providers.

At this scale, AI is no longer just about algorithms it’s about access to power, chips, and cloud infrastructure.

The White House has raised concerns that some Chinese companies may be quietly copying American AI models using a technique known as “distillation.”

Key Points:
The U.S. government is planning to work more closely with domestic AI companies to counter what it sees as large-scale copying efforts. Officials claim that some groups are creating thousands of fake accounts to interact with U.S. AI systems, gradually extracting useful insights. Companies like Anthropic and OpenAI have reportedly flagged firms such as DeepSeek, Moonshot, and MiniMax as possible participants.

What’s Going On:

In a recent internal memo, tech policy chief Michael Kratsios described these activities as “industrial-scale” attempts to replicate American AI. The method, distillation, involves repeatedly querying AI models to uncover patterns or responses that can help train competing systems. Over time, this information can be used to build similar models without the same level of investment. In response, the White House is looking to share intelligence with AI companies and strengthen protective measures. Meanwhile, China’s embassy has rejected these claims, calling them unfair and politically motivated.

Why It Matters:

Developing advanced AI requires massive investment, often running into billions. If others can replicate similar capabilities at a fraction of the cost, it creates a serious competitive imbalance. While the memo highlights the issue, it doesn’t outline clear consequences yet-leaving U.S. companies in a position where innovation may be quickly mirrored without much deterrence.