In the AI landscape, computing power reigns supreme—those who command electricity wield the 'sword of life and death'.

A longstanding maxim echoes through the corridors of the AI world (or 'jianghu' for cultural flair): To engage with AI is to dance with electricity.

Today, this adage has been starkly validated by a burgeoning AI contender, Anthropic, which has poured a staggering $45 billion—approximately RMB 320 billion—into securing computing power, a sum that could acquire a leading internet giant.

The announcement sent shockwaves through Silicon Valley, prompting a collective gasp: What level of profitability justifies such a colossal gamble?

01 Where Did the Money Vanish? Land Leasing, Chip Acquisition, and Power Securing

Anthropic's latest move targets a 'goldmine' held by cloud infrastructure provider Nscale—a sprawling data center park in West Virginia, USA.

In essence, Anthropic isn't merely renting server space; it's acquiring the 'brains' and 'hearts' within.

The 'brains' are the chips: NVIDIA's next-gen, most potent Vera Rubin chips will be deployed, maximizing computational might. The 'heart' is the electricity: A six-year contract guarantees up to 460 megawatts of power!

What does 460 megawatts entail? It's comparable to energizing 345,000 American households. Some netizens quipped: 'This isn't a data center—it's a metropolis in miniature!'

Yet, this is merely the tip of the iceberg in Anthropic's computing power empire. Previously, it inked deals with AMD, Google, and Broadcom. Even more extravagant is its agreement with Musk's SpaceX, shelling out $1.25 billion monthly to utilize SpaceX's two training clusters, dubbed Colossus.

Such profligate spending can only signify one thing: To avoid being 'strangled' by computing power scarcity, Anthropic is going all-in.

02 Why Such Expenditure? Because 'High Traffic Leads to Collapse'

Many are puzzled: As an AI large model company, why not focus solely on coding? Why splurge on infrastructure?

Anthropic shares the same frustration: User growth is skyrocketing, and servers are struggling to keep pace!

Earlier this year, Anthropic voiced its concerns publicly. They revealed that their Claude AI model's popularity had surged to the point where, during peak hours, user experience suffered, impacting 'reliability and performance'.

It's akin to having a 100-megabit internet connection at home, only to have the entire neighborhood install smart home devices. At night, when everyone streams videos, your web pages won't even load.

In the AI realm, without a robust computing power foundation, even the most sophisticated algorithms are mere castles in the sky. To prepare for its legendary 'epic IPO' and convince Wall Street it merits a $965 billion valuation, Anthropic must demonstrate: 'I can craft poetry, and I can generate power!'

03 Industry Upheaval: AI Competition Enters the 'Capital-Intensive' Second Half

Anthropic's $45 billion spending spree has sounded the alarm across the industry.

Firstly, small firms are now out of the race. Developing large models today isn't just about algorithms—it's about who possesses more GPUs and larger power plants. This is a nuclear war of capital and energy. Can an average person turn the tables with just an idea? The barrier is dauntingly high.

Secondly, those selling 'shovels' are reaping greater profits than those digging for gold. Interestingly, Nscale, the company that clinched this deal, was only established last year and was relatively obscure. By merely holding a park capable of supplying 1.35 gigawatts of power, it instantly became a hot commodity and may even go public on the U.S. stock market as soon as next month. In this era, whoever controls the distribution of energy and computing power reigns supreme.

Thirdly, power poses an even greater challenge than chips. We used to fret over chip shortages, but now, a stable power supply is equally crucial. To cater to these electricity-guzzling 'behemoths,' Nscale even constructed its own power plant, utilizing natural gas units to generate electricity on-site, independent of the public grid.

In the end, AI boils down to power!

Epilogue

Some deem Anthropic's move as madness, pushing forward despite incurring over $42 billion in losses.

But in my view, this is a calculated, bold maneuver.

The AI industry's knockout phase has commenced. Today, Anthropic is employing the 'dumbest' method to construct the deepest moat.

This $45 billion gamble isn't merely about Anthropic's destiny—it's about the exorbitant 'ticket price' for humanity's quest toward Artificial General Intelligence (AGI).

Regardless, a new era of computing power and electricity infrastructure has arrived with a bang.

And we are all witnesses to this historic transformation.