From AWS to Coldcart: How Cloud Computing Principles Unlocked a Perishable Shipping Revolution

From AWS to Coldcart: How Cloud Computing Principles Unlocked a Perishable Shipping Revolution

The pattern was obvious once you saw it. Jason Park spent his consulting years at Bain watching tech companies struggle with the same problem: massive capital expenditures on physical infrastructure they didn’t understand, couldn’t optimize, and had to manage alone.

Then AWS changed everything.

In a recent episode of Category Visionaries, Jason Park, CEO of Coldcart, revealed how that transformation shaped his entire approach to building a perishable logistics platform. A decade after watching cloud computing centralize tech infrastructure, he recognized the same pattern waiting to unfold in frozen food shipping, temperature-controlled pharmaceuticals, and industrial chemicals.

The question wasn’t whether centralization would happen. It was who would build it.

The Infrastructure Tax Before Cloud

Jason’s consulting work in the early 2010s gave him a front-row seat to enterprise transformation. “The big thing happening in tech and to tech companies was the cloud, quote unquote,” he recalls. Companies were trapped in a familiar cycle: “Having giant physical server boxes and data centers that companies didn’t really understand. And they just had to have them. And they cost a lot of money.”

The problem wasn’t that these companies were incompetent. They were rational actors making locally optimal decisions. If you needed compute power, you bought servers. If you needed more capacity, you built data centers. Every company solved the same problems independently because no alternative existed.

But this created systematic inefficiency. Small companies over-provisioned capacity because they couldn’t predict demand spikes. Large companies under-utilized resources because coordination across business units was impossible. Everyone paid the infrastructure tax.

Then came the transition “to a world in which we sign up for an Amazon Web Services AWS account and all that centralized in a way that’s more efficient than any of these individual companies can achieve on their own.”

The key phrase: more efficient than any individual company can achieve. Not just cheaper. Not just easier. Fundamentally more capable through centralization.

Recognizing the Same Pattern in Perishable Logistics

Years later, Jason reconnected with Matt Salzberg, his friend from Harvard who had founded Blue Apron. The meal kit company had reached $2 billion valuation, and Jason had a simple question: how did you figure out shipping boxes of vegetables to apartments without everything spoiling?

Matt’s answer was stark: “It was prohibitively hard.” But the way he described the challenge triggered Jason’s pattern recognition. “The way you describe it is always just too much overhead, too much infrastructure, too much specialization.”

This was the same phenomenon. Every company shipping perishable products was solving identical problems in isolation. They were building their own “data centers”—specialized fulfillment infrastructure that was too expensive, too complex, and fundamentally inefficient at the individual company level.

Jason connected the dots: “That is the same phenomenon that happened in tech industry in 2010s with data centers to centralized infrastructure.”

Why Individual Optimization Fails at Scale

The parallel runs deeper than cost reduction through volume. The real advantage of centralized infrastructure comes from what Jason calls network effects across companies—efficiencies that are literally impossible for individual companies to achieve, regardless of their size.

His example illustrates the dynamic precisely: “1000 companies shipping frozen product in New York City shipping via UPS…UPS is running 85% on time delivery. That’s fine. No alarm bells are going off, no one’s concerned. But those thousand companies are all refunding 15% of their sales because as soon as it delivers late, it’s dead.”

Each company is trapped in information poverty. They can only learn about carrier problems through their own failures. By the time they know UPS is having issues, products have already spoiled and customers have already churned.

But aggregate those same companies on a shared platform, and the information asymmetry disappears. “Through cold cards platform, we would see that happening in real time, say across a couple of companies. And then for 998 other companies we would reroute those shipments to ship out of different warehouses.”

The transformation is dramatic: “For 998 of those companies, their refunds go from 15% to 2%.” This isn’t incrementally better. It’s structurally different.

The Paradox That Makes Centralization Necessary

What makes perishable logistics particularly ripe for the AWS treatment is a fundamental paradox that individual optimization can’t solve. “You can’t just lower costs. You have to solve for cost and spoilage at the same time. And the more you improve one, the worse the other one gets,” Jason explains.

This creates a brutal trade-off. Add more insulation and ice to prevent spoilage? Costs go up. Reduce packaging to cut costs? Spoilage increases. Choose faster carriers to ensure delivery? Pay premium rates. Use cheaper carriers? Risk late deliveries and total product loss.

Every company faces this paradox alone. They pick a packaging configuration that works for worst-case scenarios and accept the waste. “Traditionally, you can’t change your SOPs. You can’t change your fulfillment process on every shipment…so you pick a packaging configuration, you stick with it.”

Centralized infrastructure solves this differently. With real-time data across weather patterns, carrier performance, and destination conditions, Coldcart can optimize each individual shipment. The same company might use different packaging for Phoenix in summer versus Seattle in winter, different carriers based on real-time performance data, different warehouse locations based on transit time predictions.

This is only possible through centralization. The data required to make these decisions only exists by aggregating across companies. The operational flexibility to execute only works with a shared network of warehouses. No individual company can build this, regardless of resources.

Why the Largest Companies Struggle Most

Jason notes something counterintuitive: “Ironically the largest companies with the largest supply chains actually tend to be the worst at this. And they will self admit that.”

This violates the usual logic of economies of scale. Shouldn’t larger companies with more volume optimize better? In traditional logistics, yes. In perishable logistics with competing constraints, no.

Large companies have massive sunk costs in existing infrastructure. They’ve built specialized facilities, established processes, trained teams. This creates organizational inertia that prevents the dynamic optimization centralized platforms enable.

They also lack the critical data advantage. A company shipping 10 million parcels annually still only sees their own performance data. They can’t learn from carrier issues affecting other companies. They can’t benefit from weather pattern analysis across multiple brands. They optimize in isolation.

This is why Jason emphasizes that network effects create efficiencies “not available for any one of those companies individually, no matter how big they are.” Size doesn’t solve the fundamental information problem.

The Category Creation Playbook

The AWS parallel reveals a repeatable pattern for finding category creation opportunities in traditional industries:

Look for infrastructure that companies must build individually despite solving identical problems. When every company builds their own version of the same complex system, centralization opportunities exist.

Find categories where individual optimization creates systematic inefficiency. The best opportunities aren’t just expensive—they’re structurally inefficient when done in isolation.

Identify problems where aggregated data creates compounding advantages. Network effects in physical operations emerge when shared intelligence enables capabilities impossible at the individual company level.

Recognize where the largest players have the least flexibility. When market leaders are trapped by sunk costs and organizational inertia, centralized platforms can serve everyone including leaders.

Jason’s AWS insight wasn’t about copying cloud computing. It was about recognizing the conditions that make centralization inevitable. When he looked at perishable logistics, he saw companies trapped in the same infrastructure tax that enterprises paid before AWS—solving identical problems alone because they had no alternative.

Coldcart became that alternative. Not by building better individual solutions, but by creating the infrastructure layer that makes individual solutions obsolete.