The Week the AI Compute Race Went Nuclear
On August 10, Nvidia gathered six of the world’s largest financial institutions—Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—and announced a plan to mobilize over $500 billion for AI infrastructure. Not in five years. Not in ten. In the next few years.
Two weeks later, Anthropic signed a $45 billion compute deal with Nscale—a company founded only in 2024.
And on August 26, Nvidia agreed to acquire Hugging Face for $12.9 billion.
This is not incremental growth. This is an industry rewriting its own economics in real time. The numbers are so large they stop meaning anything—until you realize what they represent: a bet that AI infrastructure is the most valuable asset class on the planet.
The $45 Billion Compute Deal That Changes Everything
Anthropic has been on a compute-buying spree. Over the past eight months, the company has aggressively scaled up its capacity to compete with OpenAI. Earlier this month, it signed a $10 billion deal with AI cloud startup Volta. In July, a $5 billion deal with AMD. In May, a deal with SpaceX reportedly providing $1.25 billion worth of capacity each month. In April, an expansion with Amazon for an additional 5 gigawatts of compute. And then, in August, the $45 billion Nscale deal spanning six years, powered by Nvidia’s Vera Rubin chips.
The message is unmistakable: the AI labs are not just building models. They are building empires. And the cost of entry is measured in tens of billions.
Nvidia’s $500 Billion Infrastructure Gambit
On August 10, Nvidia announced strategic partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish a “compute financing platform”. The goal: mobilize over $500 billion in third-party capital for AI infrastructure.
Nvidia CEO Jensen Huang called it a way to help customers access “scarce computing resources” and build the “AI factories” that will drive the AI era. Apollo’s president called computing resources “a critical asset class with highly attractive investment characteristics”. BlackRock’s CEO Larry Fink noted that “AI infrastructure construction will require investment on an unprecedented scale”.
Nvidia is also reportedly in talks with OpenAI to provide about $250 billion in backup financing for data center projects, while Meta is working with BlackRock on off-balance-sheet financing for its Texas data centers. This is not just technology. This is finance. And the scale is staggering.
The $13 Billion Hugging Face Acquisition
On August 26, Nvidia agreed to acquire Hugging Face for $12.9 billion. The deal values the open-source AI platform at nearly $13 billion.
Why does this matter? Because every major closed-source AI lab—OpenAI, Google, Amazon, Anthropic—is now building its own chips to reduce reliance on Nvidia. A thriving open-source ecosystem gives customers more alternatives to those closed labs, which keeps more of the market dependent on Nvidia’s hardware.
Hugging Face has also been expanding into physical AI. On August 27, the company unveiled Microduck, a $399 open-source robot that can waddle, pick things up with its beak, get back up when it falls, crouch, and even roller skate. “Welcome to the era of open-source affordable robots to democratize physical AI and world models,” said CEO Clem Delangue.
The Telecom Industry’s Awkward Reality
While the infrastructure players were making headlines, the telecom industry was quietly admitting it’s not ready.
A new HCLTech survey of nearly 200 senior executives across network operators, MVNOs, and CSPs worldwide found that while 60% of telecom leaders see AI as a key driver of future revenue, only 25% believe their organizations are ready to operationalize AI at scale. Nearly 70% agree that connectivity services are increasingly commoditized, reinforcing the need for differentiated digital services. Yet nearly 80% have launched fewer than five new digital products in the past year.
The gap between ambition and execution is wide. And it’s widening.
The Chip Design Arms Race
Anthropic is now hiring a team to design its own custom chips. The company is planning to co-design hardware and models to help its technology run faster and more efficiently.
This follows OpenAI’s unveiling of its Broadcom-built Jalapeño chip in June, designed specifically for inference workloads. Google DeepMind has long relied on Alphabet’s TPU chips. Meta has been developing its own MTIA accelerators.
The message is clear: the AI labs are no longer content to be customers. They want to be chip designers. They want to control the entire stack. And they are willing to spend whatever it takes to get there.
AI is no longer about models. It’s about infrastructure. It’s about capital. It’s about who controls the supply chain.
The companies winning aren’t the ones with the best models. They’re the ones building the factories that make the models possible. They’re the ones financing the data centers, designing the chips, and controlling the open-source ecosystems.
This was the week the AI industry stopped being a technology sector and started being an infrastructure sector. The numbers are staggering. The stakes are higher than ever. And the race is just beginning.
#AI, #Nvidia, #Anthropic, #HuggingFace, #AIIinfrastructure, #Telecom, #AIChips, #EnterpriseAI, #DigitalTransformation, #Innovation, #TechNews, #Compute, #OpenSource, #AgenticAI