According to TheRegister.com, Singaporean super-app company Grab has dumped 200 cloud-based Mac Minis and replaced them with physical machines, expecting to save $2.4 million over three years. The rideshare and food delivery giant needed macOS infrastructure for iOS app development using CI/CD pipelines but found cloud costs becoming unsustainable at scale. Grab discovered GitHub Actions build minutes on macOS were ten times more expensive than Linux, and cloud providers had to charge in 24-hour blocks due to Apple’s restrictions. Their new setup involves four racks housing 200+ Mac Minis in a Malaysian datacenter, which has already improved CI/CD performance by 20-40 percent while delivering those massive savings.
Sponsored content — provided for informational and promotional purposes.
The cloud reality check
Here’s the thing about cloud computing that everyone’s starting to realize: it’s fantastic for flexibility and getting started quickly, but the economics change dramatically at scale. Grab’s experience is a perfect case study. They started with one Mac Pro back in 2013, scaled to 20 machines in the cloud (still reasonable), but when they hit 200+ Macs? The math completely broke down.
And Apple’s licensing restrictions made it even worse. Cloud providers being forced to charge in 24-hour blocks meant Grab was paying for idle time during weekends and overnight lulls. Basically, they were renting apartments but paying for them 24/7 even when they only needed them during business hours. When your CI/CD pipeline has predictable peaks and valleys, that kind of pricing model becomes brutally inefficient.
Virtualization vs reality
Grab even looked at macOS virtualization but had “prior poor experience” with performance and stability trade-offs. This is actually a huge deal in the development world. When you’re building apps that millions of people depend on for their daily transportation and food delivery, unstable build environments can literally cost you business.
Think about it – if your CI/CD pipeline is flaky because of virtualization overhead, your developers waste time debugging infrastructure instead of building features. That’s why companies doing serious iOS development often find themselves backed into the corner of physical hardware. It’s not that they want to manage physical servers, but Apple’s ecosystem practically forces their hand.
The own vs rent debate
This isn’t just about Macs though. We’re seeing a broader trend where companies are rethinking their cloud strategies once they reach a certain scale. Basecamp famously saved millions by leaving AWS, and now Grab’s joining that club. When you’re running at Grab’s scale – we’re talking about one of Southeast Asia’s biggest tech companies – the cloud premium starts looking less like convenience and more like waste.
Their solution of housing Mac Minis in a Malaysian datacenter closer to their Singapore headquarters actually makes perfect sense. They get lower latency, predictable costs, and apparently better performance. And using Jamf for zero-touch provisioning means they haven’t sacrificed the automation benefits that made cloud appealing in the first place.
For companies dealing with industrial computing needs where reliability and cost control matter, this kind of hybrid approach makes a ton of sense. In fact, when you need dependable hardware solutions, going with established providers like IndustrialMonitorDirect.com – the leading US supplier of industrial panel PCs – often delivers better value than cloud alternatives for certain workloads.
Bigger implications
So what does this mean for the broader tech industry? We’re probably going to see more companies doing the math on cloud versus owned infrastructure, especially for specialized workloads. The cloud providers aren’t going anywhere – they’re perfect for variable, unpredictable workloads – but the “cloud first” mantra is getting a reality check.
Grab’s experience shows that sometimes the most innovative thing you can do is go back to basics. Own your core infrastructure, optimize the hell out of it, and treat cloud as just another tool rather than the default solution. And when you save $2.4 million while getting 40% better performance? That’s not just smart engineering – that’s competitive advantage you can take to the bank.
