Sam Altman’s vision for AI is huge – but there’s just one thing standing in his way

• New blog post for Sam Altman outlines the road ahead for AI
• Altman wants to create a factory that can produce a gigawatt of new AI infrastructure every week
• Altman suggests that AI may become something we consider a fundamental human right

In a new blog post called Abundance Intelligence, Sam Altman, CEO of OpenAI, lays out the benefits of more computing power for AI and calls for increased investment in AI infrastructure. “If AI stays on the trajectory that we think it will, then amazing things will be possible", he writes.

While Altman makes no definite predictions about what these amazing things will be, he is willing to muse on the potential benefits of increased computational power could mean for the future. "Maybe with 10 gigawatts of compute, AI can figure out how to cure cancer. Or with 10 gigawatts of compute, AI can figure out how to provide customized tutoring to every student on earth.”

He also issues a stark warning: “if we are limited by compute, we’ll have to choose which one to prioritize; no one wants to make that choice, so let’s go build."

Let's go build

Altman doesn’t write blog posts just to pontificate; he generally uses them to outline the direction for the next phase of OpenAI’s expansion, and it’s clear that it's now all about increasing "compute". Compute is the word Altman uses as a shorthand for the raw horsepower necessary to run and train LLMs like ChatGPT.

In the real world, that horsepower equates to data centers – vast warehouse-sized facilities full of servers, networking gear, and cooling equipment, and as you can imagine, they need large amounts of electricity to run and function.

Just yesterday on X.com, Altman tweeted a video showing progress on OpenAI and Oracle's latest massive data centers in Abilene, Texas. Part of the $500 billion Stargate Project, with five more data centers opening in the US soon. As you can see from the video, the sheer scale of it is impressive.

As we reported on Tuesday, Nvidia is investing $100bn in OpenAI, and will start by deploying as much power as 10 nuclear reactors.

In his blog post, Altman lays out exactly what the goal for OpenAI is when it comes to data centers: “Our vision is simple: we want to create a factory that can produce a gigawatt of new AI infrastructure every week. “

That’s an astonishing ambition, and one he realizes will be challenging: “The execution of this will be extremely difficult; it will take us years to get to this milestone and it will require innovation at every level of the stack, from chips to power to building to robotics. But we have been hard at work on this and believe it is possible.”

Reflecting the new political desire for homegrown technology in the U.S., Altman writes: “We are particularly excited to build a lot of this in the US; right now, other countries are building things like chips fabs and new energy production much faster than we are, and we want to help turn that tide.”

As AI gets smarter

Altman’s vision of the future is clearly going to require some incredible infrastructure building to achieve, with even more data centers than we currently have in production. The massive consumption of power required has attracted its fair share of criticism as well. Partly, this is due to the sheer environmental impact of building increasingly large data centers, but also because scaling AI compute power has so far failed to produce AGI, and there’s no indication that it will.

While there is no mention of AGI in his most recent missive, it has been a popular theme of Altman’s previous blog posts. He does, however, talk about what is going to happen as AI gets smarter: “access to AI will be a fundamental driver of the economy, and maybe eventually something we consider a fundamental human right. Almost everyone will want more AI working on their behalf.”

While it seems that achieving AGI remains as elusive as ever, there’s no reason to think that OpenAI’s plans for the future won’t be as innovative as we've come to expect, and Altman is keen to reveal them soon. “Over the next couple of months, we’ll be talking about some of our plans and the partners we are working with to make this a reality”, he writes, before ending on the enigmatic: “we have some interesting new ideas”, and I can’t wait to see what those will be.

You might also like

• ChatGPT’s 'compute-intensive' next act better be dazzling, because the competition is outshining its recent efforts
• YouTube is finally letting you hide one of its most annoying features – and you’re going to be thankful for this
• Microsoft unveils advanced AI cooling which lowers heat, cuts energy use - and could lead to more powerful data centers