Enterprises seeking on-prem AI deployments should look at the Mac

If you’re looking for an AI PC, why wouldn’t you be looking for a Mac?

After all, Apple’s systems are built for privacy and security and, thanks to Apple Silicon, are now quite capable of running big generative AI (genAI) models swiftly in-house — and at a fraction of the energy costs of Windows-based systems. While a move to Apple won’t suit every company or even every situation, it’s absolutely true that the platform is a viable one for AI, and has remained so for some time.

To understand how viable Macs are for running AI in your business, consider the recently-launched M3 Ultra Mac Studio. Apple made impressive claims for these systems on launch. It told us its highest-end Mac Studio was 16.9 times faster than the M1 Ultra Mac Studio at generating tokens using a large language model (LLM) with hundreds of thousands of parameters. That neat claim was recently put to the test when YouTuber Dave Lee showed us that the most highly specified M3 Ultra Mac Studio is capable of running a 671-billion-parameter AI model natively on the system at under 200 watts of energy. (He used DeepSeek RI, of course.)

A supercomputer on your bookshelf

That’s an amazing statistic; it means you can run a powerful genAI models using DeepSeek on a system that fits on a shelf above your desk, which must be a compelling proposition to any enterprise purchaser considering deployment of an in-house self-hosted system for AI. It’s even more impressive when you consider that comparable performance on traditional PCs would require multiple GPUs and consume 10 times as much energy. 

(You might be able to find some way to put together a PC-based cluster to deliver the same kind of performance, but it is arguable that you’ll still spend the $14,000 you’d need to drop to get the most powerful Mac desktop. (In some cases, you might not need quite as much of a system.)

What this all means — or should mean — is that the enterprise landscape is ripe for Apple. After all, the company’s devices are private by design, secure by nature, and deliver performance and energy consumption advantages other PCs at the same price points don’t match. Not only do they deliver this, but they also have the processor horsepower it takes to run AI natively.

If you just need to run smaller AI models, then an M4 MacBook Air also has you covered, allowing most businesses to now consider Mac-based and self-hosted AI deployments for a range of use cases.

Apple, the perfect endpoint AI solution

While buffeted by bad news elsewhere in its business, the Mac has become a bright star for Apple, thanks to Apple Silicon. While Apple Intelligence might have let the company down, Apple Silicon has not. That matters quite a lot, given a recent report from IDC Research claiming that four-in-five (80%) of PCs sold this year will be AI PCs, though there are some challenges to that.

Apple helps meet those challenges, of course. IDC’s survey revealed that while nearly all organizations are using or intend to use cloud-based AI platforms, the security and data privacy inherent to these systems remains the biggest challenge to broad deployment, even more than cost.

While most business users are looking at the ever-expanding plethora of enterprise-focused cloud-based AI systems, they are really concerned about what happens to their data when shared in the cloud, particularly in regulated industries where compliance matters, such as healthcare.

That’s driving many enterprises seeking to deploy AI to search for solutions that let them deploy genAI tools on endpoints — self-hosted AI that runs natively on a device they own and have in their own office, in other words.

That, of course, is what Apple’s Macs — and certainly its Mac Studio — are more than capable of doing. 

AI models for the few

There’s a lot more to running AI in business than text summaries, automated writing tools, Microsoft Copilot, or even super-smart Webex AI bots; enterprise users are seeking to build or buy their own bespoke AI packages capable of learning from their own heavily curated business data and helping them with their business. That’s not software that necessarily comes in a box (digital or otherwise), but is software that, at best for ultimate privacy and security, runs natively on the device.

Apple Silicon is more than up to the task. (And you get an annual free OS upgrade of what is arguably the most secure operating system, too.)

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.